OpenEvidence
面向医生决策支持的循证临床 AI
OpenEvidence 已拿到少见的临床医生采用率,也建起真正的高端内容护城河;但 2026 年 1 月 $12 billion 估值远远跑在公开证据前面,收入质量、广告主耐久性、企业端深度和治理披露都还撑不起这个价格。
封面要素
公司概况
OpenEvidence 是一家快速扩张的医疗 AI 公司,面向临床医生的免费产品能用带引用、经同行评审的证据回答诊疗现场问题。公司由 Daniel Nadler 和 Zachary Ziegler 于 2022 年创立,把快速医生采用与授权医学内容、出版方合作和正在成形的医疗体系工作流集成结合起来。截至 2026 年 1 月,OpenEvidence 已完成 $250 million Series D,估值 $12 billion;不到一年前,它的 Series A 估值还只有 $1 billion。
- 成立时间
- 2022-01-01
- 创始人
- Daniel Nadler, Zachary Ziegler
- 创立地点
- Miami, FL, USA
- 总部
- Miami, FL, USA
- 产品
- OpenEvidence 提供循证临床 AI 平台,覆盖网页、原生移动端,并越来越多嵌入工作流。核心模块包括快速临床咨询、Deep Consult 长篇研究、语音交互、文档与编码支持、沟通工具,全部都绑定来自 NEJM、JAMA、NCCN、Cochrane 等授权医学内容的引用。
- 客户
- 已验证身份的临床医生是主要用户,也是分发切入口;企业级医疗体系正在成为机构买方,可把使用范围扩展到护士、药师和多学科护理团队。
- 商业模式
- 核心访问对已验证临床医生免费,主要靠药企和医疗器械广告变现。管理层和第三方分析都指向正在发展的企业级与生命科学变现层,但公开收入结构和合同经济性仍不透明。
- 阶段
- Series D private company
- 融资情况
- OpenEvidence 于 2026 年 1 月以 $12 billion 估值完成 $250 million Series D;此前 2025 年快速连续融资,包括 $75 million Series A、$210 million Series B 和 $200 million Series C。公开来源称,公司过去 12 个月累计融资约 $700 million。
执行摘要
主要优势
- 据称美国医生日使用率超过 40%,触达 10,000+ 家医院和医疗中心
- 授权证据护城河覆盖 NEJM、JAMA、NCCN、Cochrane 等医学出版方
- 免费临床医生产品叠加强医生口碑,降低获客摩擦
- 投资人阵容一线,包括 Sequoia、GV、Kleiner Perkins、Thrive 和 DST
- 在 Cedars-Sinai、Mount Sinai 和 Sutter Health 拿到早期企业工作流牵引力
主要风险
- $12B 私募估值隐含倍数远高于上市医疗软件和医生网络可比公司
- 商业化仍似乎由广告驱动,但广告主集中度、续约和政策隔离均未披露
- 产品更深进入诊疗工作流后,FDA/CDS 分类漂移、医疗过失暴露和幻觉风险都会上升
- OpenEvidence 高度依赖授权内容和出版方关系,才能守住质量护城河
- 既有参考工具和通用模型平台可能压缩差异化和定价权
- 治理仍集中在创始人手里,烧钱速度、股权结构表和收入组合的公开披露有限
未决问题
- 按商业化来源拆分的经审计 FY2025 收入、毛利率和贡献利润率
- 头部广告主集中度、续约队列,以及隔离赞助内容和临床答案的防护机制
- 已签企业合同管线、落地时间和非广告 ARR 贡献
- 临床质量治理指标、幻觉事件日志和升级处理流程
- $12B 估值下的股权结构偏好、老股定价和清算悬垂
- 按专科、医院类型和部署批次拆分的留存与队列分析
目录
01公司概况
1.1 公司定位与产品模式
OpenEvidence 是一家总部在 Miami 的医疗 AI 公司。公司由 Daniel Nadler 和 Zachary Ziegler 于 2022 年创立,面向临床医生的平台在 2023 年推出。公司把自己定位成医疗知识平台,以及服务高风险诊疗现场决策的 AI 副驾,而不是通用聊天机器人。这个差异对其 GTM 叙事很关键:OpenEvidence 称答案基于经同行评审的医学文献,会引用底层来源,并用授权医学内容训练,而不是开放互联网。产品界面如今覆盖核心 Ask 体验、用于就诊捕捉和病历生成的 Visits、可免手操作问答的 Voice Mode,以及用于更长篇智能体研究的 DeepConsult。核心访问仍对已验证身份的美国临床医生免费,这帮助了分发,但免费模式由广告买单。最终形成的平台更像工作流基础设施,而不是单一搜索工具;但它的商业耐久度仍取决于医生是否同时信任答案质量和变现护栏。[CO001, CO002, CO003, CO004, CO005, CO006]
| 指标 | 数值 / 状态 | 日期 | 置信度 | 缺口 / 备注 |
|---|---|---|---|---|
| 总部 | 佛罗里达州迈阿密 | 2026 | 高 | 官网未突出总部信息;当前地点由 CNBC 报道佐证。 |
| 成立 | 2022 | 历史 | 高 | 产品随后于 2023 年上线。 |
| 平台上线 | 2023 | 历史 | 中 | 公司宣传把两周年对应到 2025 年春季。 |
| 最新估值 | $12B | 2026-01 | 高 | Series D 轮估值。 |
| 过去 12 个月融资额 | ~$700M | 2026-01 | 高 | 公司披露的滚动期间数字,不是累计融资额。 |
| CNBC Disruptor 列示融资额 | $795.4M | 2026-05 | 中 | 可能是累计融资额;尚未与滚动期间披露口径勾稽。 |
| 美国医生日使用率 | >40% | 2025-07 to 2026-05 | 高 | 日用指标;NBC 另按不同分母报道了更大的活跃用户数。 |
| 临床咨询 | 2025 年 12 月 18M;迄今 >200M | 2025-12 / 2026 | 中 | 所有咨询量数字均由公司披露。 |
| 年化收入 | >$100M | 2025 | 中 | CNBC 报道;没有经审计的公开申报。 |
| 安全态势 | 符合 HIPAA;SOC 2 Type II | 2025-04 / 2026 | 高 | 尽管公司如此表述,一些医疗系统仍限制输入 PHI。 |
| 董事会公开披露 | 2026 | 低 | 留存来源未披露董事席位和控制权。 |
数值混合了公司披露与独立报道。Null 表示未公开披露;公开来源之间的融资总额尚未完全勾稽。
[CO001, CO002, CO003, CO018, CO025, CO027]展示授权内容、工作流产品、免费分发和商业化如何串联,形成 OpenEvidence 的采用飞轮。
[CO004, CO005, CO006, CO007, CO008, CO009]1.2 创始人、领导层与治理透明度
OpenEvidence 明显由创始人驱动。Daniel Nadler 仍是公司的公众面孔,并反复以创始人兼 CEO 身份出现;Zachary Ziegler 则是最常与平台 AI 研究背景绑定的联合创始人。Nadler 此前创办 Kensho,在 OpenEvidence 出现之前就已获得投资人和企业买方的信任;S&P Global 于 2018 年收购 Kensho,也把他确立为有大型 AI 退出经历的连续创业者。相比技术或商业动能,当前公开记录对正式治理的披露薄得多。官网突出展示了一个异常广泛的医学顾问网络,覆盖美国主要医疗体系和学术中心,这强化了临床可信度;但留存材料没有披露董事会构成、投资人观察员权利或控制架构。这种不对称带来真实尽调风险:顾问深度能帮助验证产品方向,但 Nadler 的关键人依赖看起来仍高,因为融资、媒体叙事和平台定位都由他同时锚定。[CO012, CO013, CO014, CO015, CO016, CO017]
| 人物 | 职务 | 背景 | 创始人-市场契合 / 覆盖面 | 关键人物依赖 |
|---|---|---|---|---|
| Daniel Nadler | 创始人兼 CEO | Harvard 博士;创办 OpenEvidence 前创立 Kensho | 连续创业的垂直 AI 创始人;锚定融资、产品愿景和公开叙事 | 高 |
| Zachary Ziegler | 联合创始人 | Harvard AI 研究员 / 博士生 | 核心 AI 和模型开发背书 | 高 |
| Mondira Ray, MD, MBI | 医学顾问 | Boston Children’s Hospital / Harvard Medical School 背景 | 提供临床输入和儿科工作流背书 | 低 |
| Ania Bilski, MD | 医学顾问 | UCSF 与 Kaiser 助理临床教授 | 为床旁产品设计提供临床工作流背书 | 低 |
| Paul E. Sax, MD | 医学顾问 | Brigham and Women’s Hospital 感染病科临床主任 | 感染病与证据解读领域背书 | 低 |
公开材料清楚识别了创始人和庞大的顾问阵容,但没有披露正式董事会或更广泛的运营层级。
[CO012, CO013, CO014, CO015, CO016]1.3 融资历程与投资方图谱
即使按 2025-2026 年 AI 公司的标准看,OpenEvidence 的融资曲线也异常陡峭。公司在 2025 年 2 月完成 Sequoia 领投、估值 $1 billion 的 Series A,7 月升至 $3.5 billion 的 Series B,10 月完成估值 $6 billion 的 Series C,2026 年 1 月又完成估值 $12 billion 的 Series D。每一轮都增加了一层赞助方质量:Sequoia 锚定机构入场,GV 和 Kleiner Perkins 共同领投 Series B,Thrive 与 DST 领投 Series D。公司披露称过去十二个月融资约 $700 million,而 CNBC 2026 年 5 月的 Disruptor 画像列出累计融资 $795.4 million。这些数字未必矛盾,但公开材料也尚未把它们对齐。明确的是,OpenEvidence 现在的股权结构叙事建立在顶级风投、来自 Mayo Clinic 的战略医疗信号,以及反复跟投之上,说明投资人把采用数据看得异常可信。[CO017, CO018, CO019, CO020, CO021, CO022]
| 利益相关方 | 角色 | 轮次 | 控制权 / 经济重要性 | 尽调问题 |
|---|---|---|---|---|
| Sequoia Capital | 领投方 | Series A;Series B 跟投 | 以 $1B 估值锚定首轮机构融资,并可能在入场时塑造治理结构 | 确认当前持股、董事会权利以及任何保护性条款。 |
| Google Ventures (GV) | 联合领投 / 领投方 | Series B 联合领投;Series C 领投 | 在 Nadler 多个项目上持续下注;对 AI 人才和战略释放强信号 | 确认 GV 是否持有董事或观察员席位,以及任何信息权。 |
| Kleiner Perkins | 联合领投方 | Series B 联合领投;Series C 参与方 | 高信号软件投资机构,影响扩张叙事 | 确认治理权利,以及 Kleiner 是否有正式董事会代表。 |
| Thrive Capital | 成长阶段投资者 / 联合领投 | Series B 参与方;Series D 联合领投 | 当前 $12B 估值轮的领投方;可能重大影响下一轮融资或退出预期 | 确认持股集中度、清算优先权和治理权利。 |
| DST Global | 成长阶段联合领投 | Series D | 后期全球资本,可能推动规模化和流动性结果 | 厘清持股、跟投能力和任何否决权。 |
| Nvidia | 战略投资者 | 公司与轮次披露列为主要投资者 | AI 基础设施信号价值,以及潜在算力关系杠杆 | 评估股权投资之外是否存在任何商业或供应承诺。 |
| Blackstone | 后期参与方 | Series C / 后续投资者名单 | 非传统跨界资金增强后续轮估值支撑 | 判断该持仓是战略性、财务性,还是与老股交易有关。 |
| Mayo Clinic | 战略投资者兼验证伙伴 | 投资者名单;研究合作方 | 医疗品牌信号与临床验证杠杆 | 评估 Mayo 是否在股权持仓之外拥有优先研究、数据或商业权利。 |
投资者角色汇总自轮次公告和官方关于页面。持股比例、董事席位和附属协议仍未公开。
[CO017, CO018, CO021, CO024, CO026, CO027]1.4 采用率、内容护城河与公开里程碑
OpenEvidence 的护城河叙事建立在两个相互强化的支柱上:医生采用和独家内容供给。公司披露与独立报道中最一致的牵引力数字是,超过 40% 的美国医生每天使用该平台,覆盖逾 10,000 家医院和医疗中心。公司公告称,系统在 2025 年 12 月支持了约 18 million 次美国临床咨询,2026 年 3 月单日咨询量突破 1 million 次,迄今已支持超过 200 million 次咨询。CNBC 另行报道称,公司 2025 年年化收入超过 $100 million,说明它已把注意力转化为真实变现。供给侧,OpenEvidence 组建了差异化授权语料库:1990 年以来的 NEJM Group 内容、JAMA 期刊家族全文、Wiley 400 多种期刊和图书、Cochrane 综述与临床答案,以及 NCCN 肿瘤学算法。Visits 等产品里程碑,加上 2026 年与 Sutter 和 Cedars-Sinai 的医疗体系合作,显示公司正从搜索走向嵌入工作流的临床基础设施。[CO029, CO030, CO031, CO032, CO033, CO034]
| 日期 | 事件 | 类型 | 金额 / 估值 / 状态 | 参与方 | 含义 |
|---|---|---|---|---|---|
| 2022 | Daniel Nadler 和 Zachary Ziegler 创立 OpenEvidence | 创立 | Daniel Nadler;Zachary Ziegler | 公司围绕仅限临床医生使用的医疗 AI 逻辑成立,并承接 Nadler 过往 AI 运营经验。 | |
| 2023 | 面向医生的平台上线 | 产品 | OpenEvidence | 后来 2025-2026 融资中强调的采用曲线由此开始。 | |
| 2025-02 | 宣布 Series A 与 NEJM Group 内容协议 | 融资 | $75M,估值 $1B | Sequoia;NEJM Group | 机构融资和优质内容供给同步到位。 |
| 2025-04 | 宣布符合 HIPAA | 监管 | 符合 HIPAA 状态 | OpenEvidence | 使公司能更积极地营销可处理 PHI 的临床医生工作流。 |
| 2025-04 | 宣布 Mayo Clinic 研究 | 规模 | 在常见场景中可比医生 CDM | OpenEvidence;Mayo Clinic | 为答案质量提供早期公开验证叙事。 |
| 2025-06 | 宣布 JAMA Network 内容协议 | 合作 | 13 份期刊,包括 11 份专科期刊 | OpenEvidence 与 American Medical Association / JAMA | 扩展多专科全文和多媒体访问。 |
| 2025-07 | 宣布 Series B 与 DeepConsult | 融资 | $210M,估值 $3.5B | GV、Kleiner Perkins、Sequoia 与 OpenEvidence | 估值跃升叠加更高算力强度的智能体产品发布。 |
| 2025-08 | 面向患者就诊场景推出 Visits | 产品 | 实时就诊智能 | OpenEvidence | 从搜索扩展到文档和工作流支持。 |
| 2025-10 | Series C 被报道 | 融资 | $200M,估值 $6B | GV、Sequoia、Kleiner Perkins、Thrive、Coatue、BOND、Blackstone、Craft 等投资方 | 估值在 Series B 三个月后几乎翻倍。 |
| 2025-06-20 | OpenEvidence 在 D. Mass. 起诉 Doximity | 不利 | Case 1:25-cv-11802 | OpenEvidence、Doximity、Jey Balachandran、Jake Konoske 等相关方 | 法律风险和竞争性安全冲突公开化。 |
| 2026-01 | 宣布 Series D | 融资 | $250M,估值 $12B | Thrive Capital 与 DST Global | 使 OpenEvidence 成为全球估值最高的医疗 AI 公司之一。 |
| 2026-02 | 宣布 Sutter Health 工作流合作 | 合作 | Sutter Health;OpenEvidence | 表明医疗系统嵌入正走出消费者式搜索。 | |
| 2026-03 | 宣布 Wiley/Cochrane 合作 | 合作 | 400+ 份期刊 / 图书,加 Cochrane 综述 | Wiley;Cochrane;OpenEvidence | 加深长尾专科内容和系统综述覆盖。 |
| 2026-03-10 | 单日完成 100 万次临床咨询 | 规模 | 24 小时 1M 次咨询 | OpenEvidence;NPI 验证临床医生 | 显示临床医生使用正从新鲜尝试转向规模化日常行为。 |
| 2026-04 | 宣布 NCCN 肿瘤算法合作 | 合作 | 标准治疗算法 | NCCN;OpenEvidence | 强化肿瘤床旁相关性。 |
| 2026-05 | 宣布 Cedars-Sinai 患者感知 AI 合作 | 合作 | Cedars-Sinai;OpenEvidence | 指向具备患者上下文的企业级临床智能。 |
时间线覆盖公开宣布的创立、融资、产品、合作、规模和不利里程碑。公开来源看不到私人治理和内部运营里程碑。
[CO002, CO003, CO018, CO020, CO023, CO025]| 合作伙伴 / 资产 | 贡献 | 公开日期 | 战略意义 | 尽调问题 |
|---|---|---|---|---|
| NEJM Group | NEJM 系列出版物 1990 年以来全文和多媒体内容 | 2025-02 | 顶级文献护城河和信任信号 | 厘清经济条款、排他性和续约权。 |
| JAMA Network | JAMA、JAMA Network Open 和 11 份专科期刊的全文与多媒体 | 2025-06 | 床旁多专科证据深度 | 厘清使用权、排序待遇和内容更新延迟。 |
| Wiley / Cochrane | 400+ 份期刊 / 图书,加 Cochrane 综述和 Clinical Answers | 2026-03 | 扩展长尾专科覆盖和黄金标准证据综合 | 确认学会拥有的内容是否并入同一协议。 |
| NCCN | 标准肿瘤治疗算法 | 2026-04 | 让肿瘤建议更贴近工作流、更具指南针对性 | 要求说明算法更新节奏和展示权。 |
| Sutter Health | 面向医生洞察的工作流合作 | 2026-02 | 显示医疗系统嵌入超出独立搜索 | 评估部署深度、用户数和 EMR 集成条款。 |
| Cedars-Sinai | 患者感知临床智能合作 | 2026-05 | 指向上下文感知企业级临床 AI | 厘清数据访问、隐私护栏和商业化范围。 |
本表聚焦已公开宣布的内容合作和医疗系统协作,它们看起来与护城河形成和工作流嵌入最相关。
[CO037, CO038, CO039, CO042, CO043, CO044]从创立到 2026 年 5 月,覆盖融资、内容、工作流、规模和负面进展的关键已定日期里程碑。
时间线优先呈现具有持久战略意义的公开里程碑;未披露的董事会或人员事件未纳入。
[CO002, CO003, CO018, CO020, CO023, CO025]1.5 风险、商业化张力与待解问题
让 OpenEvidence 强大的那些特征,也构成本章最主要的尽调旗标。公开报道显示,临床医生认可速度,但 NBC News 发现,幻觉、不完整答案、患者结局证据有限、削弱低年资临床医生批判性思维等担忧仍持续存在。隐私接受度也不均衡:虽然 OpenEvidence 宣布符合 HIPAA,并称已获得 SOC 2 Type II 认证,但 NBC 报道称 MaineHealth 仍要求临床医生不要输入 PHI。商业化是另一处张力。CNBC 和 NBC 都把核心产品描述为广告支持,包括药品或医疗器械推广;一篇 2026 年 medRxiv 预印本显示,在证据本来均衡时,广告会改变药物选择。最后,OpenEvidence 已与 Doximity 陷入现场诉讼,争议涉及涉嫌提示词攻击和商业秘密提取,反诉则指向错误信息和挖角。这些风险都没有否定公司的非凡采用率,但它们意味着,治理透明度、隐私护栏和广告模式控制,值得获得与使用增长和头部估值同等强度的尽调。[CO007, CO011, CO035, CO036, CO040, CO046]
1.6 展示项
02市场分析
2.1 市场边界、纳入口径与现状替代品
对 OpenEvidence 而言,相关市场不是“所有医疗 AI”,甚至也不是临床软件的每个类别。AHRQ 把临床决策支持定义为通常在诊疗现场提供的及时信息,用来帮助有关患者护理的决策;例子包括建议、数据库、提醒和警报。FDA 的临床决策支持指引进一步收紧商业边界,把若干面向临床医生的非设备 CDS 软件功能,与受设备监管的软件功能区分开来,尤其是后者面向患者,或仍处在 FDA 数字健康监管范围内。落到实际支出口径,这意味着市场包括面向临床医生的知识订阅、证据数据库、AI 辅助临床问题回答,以及嵌入 EHR 的推荐层;但大块环境式文档、面向患者的症状分诊、纯编码或收入周期工具、影像 / 信号分析产品则被排除在外,除非它们作为床旁或会诊工作流里的临床医生决策支持出售。 现状替代品也不只是某一家既有供应商。临床医生在采用新的专用工具之前,经常靠在线数据库、免费互联网搜索和同事回答问题;UpToDate 与 DynaMed 则设定了参照标准:快速、基于文献、并集成进移动端和 EHR 工作流的答案。Epocrates 和 AMBOSS 说明,该类别的一部分仍可用自助或小团队订阅跑通;Glass Health 和 Pathway 等更新的 AI 原生产品则显示,替代集合如今既包括 LLM 形态界面,也包括传统参考库。因此,从市场定义看,OpenEvidence 位于一个“临床医生信任 + 工作流”类别中;这里的买方已经期待带引用的证据、快速检索和低摩擦部署,而不是消费者聊天体验。[CM001, CM002, CM003, CM004, CM005, CM006]
| 层级 | 纳入支出 | 排除或相邻支出 | 典型买方 / 付款方 | 为什么重要 |
|---|---|---|---|---|
| 核心市场——面向临床医生的床旁 CDS | 证据数据库、AI 辅助临床问答、药物 / 指南查询,以及临床医生诊疗时使用的建议层 | 无——这就是核心品类 | 临床医生用户;付款方可能是个人、科室或医疗服务机构 | 这是 OpenEvidence 实际切入的市场 |
| 嵌入式工作流 CDS | 集成进 EHR 的提醒、医嘱指导和嵌在临床工作流中的决策支持模块 | 不归因于决策支持的独立 EHR 平台支出 | 医疗系统或医疗集团 | 企业扩张中,工作流适配比新奇感更重要 |
| 现状替代品 | UpToDate、DynaMed、Epocrates、AMBOSS、PubMed / 免费网页搜索、同事会诊 | 无——这些替代品定义了要替换的既有行为 | 个人临床医生、图书馆、GME 或医疗系统 | OpenEvidence 首先要抢的是既有问答习惯,而不是空白市场 |
| 排除 — 类器械或患者端 AI | 除非明确作为临床医生 CDS 销售,否则不计入 | 面向患者的症状工具、仍按器械监管的影像或信号分析软件功能、护理者工具 | 消费者、器械预算,或受监管器械买方 | 这些类别会把泛医疗 AI 的 TAM 撑大,但采购路径不是一回事 |
| 相邻 — 文档与编码自动化 | 只计入同时提供临床医生证据或治疗支持的部分 | 环境式病历记录、编码优化、不含诊疗现场知识检索的收入周期 AI | 运营、收入周期或行政预算 | 有参考价值的相邻赛道,但不是测算 OpenEvidence 核心市场最干净的基准 |
边界逻辑先按 AHRQ 和 FDA 定义划线,再叠加临床医生调研文献中观察到的替代行为,以及在位厂商的产品定位。 纳入和排除的行是类别判断,不是经审计的会计科目。
[CM001, CM002, CM003, CM004, CM005, CM006]2.2 TAM、SAM 和 SOM——自上而下估计矛盾,自下而上只给出下限
公开自上而下估计分歧足够大,TAM 必须呈现为区间,而不是单一口号。Business Research Company 把 2026 年全球临床决策支持系统市场定在 $3.96 billion,而 Fortune Business Insights 给出 2026 年 $4.45 billion。这两个数字方向相近,但仍相差约 $0.49 billion;相对这个规模的市场,差异很大,也反映了纳入口径、供应商集合和地理假设不同。Grand View Research 对更窄的 CDSS 中人工智能细分市场给出的 2026 年估计只有 $1.5 billion;这与更广义 CDSS 总量并不冲突,更像是证据:诊疗现场医疗 AI 是嵌在更大证据与警报生态中的子集。Research and Markets 通过把 CDSS 按组件、模型、交付模式、应用和终端用户细分至 2035 年,也在结构上强化了同一点,而不是把它当作单一整体 SKU。 对 OpenEvidence 来说,更有决策价值的测算视角是自下而上、以美国为中心。BLS 报告 2024 年约有 839,000 名医生、162,700 名医师助理和 320,400 名执业护士,也就是约 1.322 million 个经常作出诊断或治疗决策的高级临床医生席位。套用 Epocrates(每年 $179.99)和 AMBOSS(每年 $259)公开自助价格,美国高级临床医生循证支出代理区间约为每年 $238 million 至 $342 million。如果初始商业切口只看医生,同一方法给出医生优先子市场下限约 $151 million 至 $217 million。这是保守的 SOM 代理,而不是总市场天花板:企业打包、治理工作、实施服务和报价制合同都可以放大实际 ACV,但公开来源不支持更干净的机构价格基准。最终得到的是本章特定的 TAM/SAM/SOM 栈:全球 CDSS 是外层 TAM,美国高级临床医生诊疗现场席位市场是证据最充分的 SAM,医生优先部署是最清晰的初始 SOM。[CM009, CM010, CM011, CM012, CM013, CM014]
| 口径 | 地域 | 数值 | 增长 / 份额 | 方法 | 置信度 | 局限 |
|---|---|---|---|---|---|---|
| 全球 CDSS(TBRC) | 全球 | 2026 年 $3.96B | 到 2026 年 CAGR 12.9%;2030 年 $6.49B | 出版方自上而下的临床决策支持系统市场模型 | 中 | CDSS 大类较宽;不专指诊疗现场 AI |
| 全球 CDSS(Fortune) | 全球 | 2026 年 $4.45B | 到 2034 年 CAGR 9.49%;北美占 2025 年市场 35.77% | 出版方自上而下的 CDSS 市场模型,带详细细分 | 中 | 高于 TBRC,说明定义敏感 |
| CDSS 中的 AI(Grand View) | 全球 | 2026 年 $1.5B | 到 2033 年 CAGR 17.1% | 更窄的 AI 赋能 CDSS 子赛道 | 中 | 不能和全 CDSS 估算直接对比 |
| 类别宽度(Research and Markets) | 全球 | 抓取页面未披露单一 2026 年数字 | 预测表延伸至 2035 年 | 按组件、模型、交付、应用和终端用户切分 | 低 | 来源更能说明类别很宽,而不是给出一个醒目的总量数字 |
| 自下而上的美国高级临床医生 SAM 替代测算 | 美国 | 每年 $238M–$342M | 基于 1.322M 个医生、PA 和 NP 席位 | BLS 席位数 × 公开自助年费($179.99–$259) | 中 | 下限,因为企业定价不透明 |
| 自下而上的医生优先 SOM 替代测算 | 美国 | 每年 $151M–$217M | 基于 839k 个医生席位 | BLS 医生数 × 同一公开年费区间 | 中 | 假设先打医生切口,忽略服务和企业溢价 |
美元数字刻意混合自上而下的市场报告和自下而上的公开价格代理,以展示边界分歧。 自下而上测算来自 BLS 席位数和公开价格,不是已披露的企业合同金额。
[CM009, CM010, CM011, CM012, CM013, CM014]OpenEvidence 的决策最需要的视角,是从宽泛的全球 CDSS 收窄到医生优先的下限支出代理;这一口径匹配公司面向临床医生的定位。
上层使用出版商市场报告,下层用计算出的美国席位数乘以公开自助价格。图中刻意强调医生优先切入点,这是最贴合 OpenEvidence 的商业视角,而不是重新渲染完整方法表。
[CM022, CM023, CM041, CM043]独立 2026 年市场估计从较窄 AI-CDSS 视角延伸到更宽的全 CDSS 定义;另有美国公开价格下限,远低于企业级 TAM 口号。
所有数值均为十亿美元。低值 / 高值相同表示出版商点估计,而不是明确置信区间。
[CM009, CM010, CM011, CM012, CM022]2.3 买方、用户与付款方分层
终端用户通常是带着紧急问题的临床医生,但付款方会随合同形态大幅变化。Epocrates 和 AMBOSS 的公开价格页显示,个人医生和小团队仍可直接购买;这对自下而上采用很重要,因为它降低了初次试用摩擦。但 BLS 就业数据显示,绝大多数医生、PA 和 NP 都在诊所、医院和门诊机构内执业,因此规模化收入仍取决于医疗服务机构,而不是原子化消费者需求。这使真实商业地图呈三角形:临床医生是用户,医疗服务机构或科室通常付款,临床和数字化领导层的混合体担任买方与守门人。 企业预算所有者越来越不再是单个图书馆或教育管理员。Healthleaders 和 HIMSS 都描述了一种采购动作:当 AI 输出触及决策,或最终可能嵌入 EHR 时,流程会经过 CMIO、CIO、首席数字官、信息学、合规和治理委员会。医生自己也预期参与其中:85% 的 AMA 受访者希望被咨询,或直接参与采用决策。实践中,这创造了几条并行预算路径。自付和科室购买可以打开入口,学术中心和培训项目仍可赞助参考访问,大型医疗体系则会在相信工具能降低认知负荷、加快文献综合,并满足验证、隐私和监督预期时批准企业级交易。因此,对 OpenEvidence 来说,最现实的早期买方,是由医生主导或 CMIO 支持的内部支持者;用户是临床医生,但付款方是诊所、医疗体系或培训组织。[CM025, CM026, CM027, CM028, CM029, CM030]
| 细分市场 | 主要买方 | 主要用户 | 主要付款方 | 预算负责人 / 审批方 | 采用触发点 |
|---|---|---|---|---|---|
| 个体医生 | 本人 | 医生 | 个体临床医生 | 个人 CME / 自由支配预算 | 诊疗中需要更快检索证据 |
| 小型诊所或科室 | 诊所负责人或专科主任 | 医生和 APP | 诊所或科室 | 诊所管理员加医生倡导者 | 小团队标准化和共享访问 |
| 医疗系统 / IDN 企业 | CMIO / CIO / 数字化或信息学负责人 | 医院和诊所内的临床医生 | 医疗服务机构 | 临床治理、安全、合规和高管发起人 | 工作流集成、验证和 ROI 论证 |
| 学术医疗中心 / GME | 图书馆、GME、CMIO 或服务线发起人 | 主治医生、住院医师、专科 fellow、学生 | 机构 | 图书馆 / 教育预算,临床签字 | 培训一致性和诊疗现场文献访问 |
| 医生优先 AI 试点 | 医生倡导者,配合 IT 和合规审查 | 部分医生或服务线用户 | 科室或创新预算 | CMIO / 首席数字官 / 创新委员会 | 已证明搜索时间或认知负担下降 |
预算负责人列综合了公开订阅路径、临床医生调研数据和治理文章。 公开证据足以证明这些渠道存在,但对 CMIO、CIO、图书馆和 GME 之间的准确预算份额拆分支持较弱。
[CM025, CM026, CM027, CM028, CM029, CM030]诊疗现场 AI 要从临床医生需求变成预算,必须先经过医生背书、治理审查、验证和合同模式选择。
这是概念性购买路径,不是按时间缩放的实施计划。它强调每个关口谁重要,而不是给出公开来源无法支撑的周期长度估计。
[CM026, CM027, CM028, CM029, CM030, CM031]2.4 增长驱动、采用约束与真正释放预算的条件
增长逻辑很直接。医生使用 AI 已经从好奇跨过门槛,变成常规职业行为:超过 80% 的受访医生如今在专业场景使用 AI,平均使用场景升至 2.3,39% 已经用 AI 总结医学研究和照护标准。这些统计很关键,因为它们直接重叠了诊疗现场医疗 AI 要完成的任务。与此同时,临床医生仍有真实工作流痛点。PCP 在带薪休假日的中位数 39% 时间会使用 EHR,并把约 39.5% 的 PTO EHR 时间花在收件箱任务上;另有文献显示,临床问题高频出现,时间是这些问题未解决的最大原因,超过一半的诊疗现场问题可能从未被追问。因此,能减少检索时间、综合证据,并降低文档或收件箱负担的诊疗现场 CDS,有可信需求驱动。 约束端同样重要。警报疲劳仍可量化:近期警报暴露最高的医生,对新的 CDS 警报反应概率低得多,这提醒我们,更具侵入性的自动化并不天然更有价值。同一份 AMA 调查既显示强采用,也说明企业预算所有者为什么犹豫:88% 担心技能流失,86% 认为隐私保护至关重要,88% 希望有扎实的安全性和有效性验证,而清晰责任框架在会提升信任的监管行动中排位最高。HIMSS 也呼应这些担忧,明确要求 ROI、可追溯性、持续监控和重新验证。既有厂商也没有停下:UpToDate 和 DynaMed 已经占据受信任的证据工作流,EBSCO 的 Dyna AI Mode 显示既有厂商可以把 AI 叠加到既有参考内容业务之上。净结果是,市场需求真实存在,但转化取决于有证据支撑的信任,而不是新奇感。OpenEvidence 受益于临床医生使用 AI 的强劲长期顺风,但规模化预算释放只会发生在产品证明透明来源、工作流集成,以及相对根深蒂固替代品可量化降低认知负荷或提高决策质量之后。[CM032, CM033, CM034, CM035, CM036, CM037]
| 驱动或约束 | 方向 | 证据 | 时点 | 商业含义 | 尽调问题 |
|---|---|---|---|---|---|
| 医生 AI 采用率已越过 80% | 驱动 | AMA 显示 >80% 医生用于专业场景,平均 2.3 个用例 | 当前 | 市场不再需要基础认知教育 | 哪些临床医生群体每周使用 OpenEvidence,哪些只是试用? |
| 研究摘要已是主流 AI 用例 | 驱动 | 39% 医生用 AI 做研究和照护标准摘要 | 当前 | 与诊疗现场证据检索的核心任务直接重合 | 使用中有多少是床旁决策支持,多少是异步研究? |
| 职业倦怠和 EHR 负担仍然尖锐 | 驱动 | 70% 认为自动化可缓解倦怠;PTO 期间 EHR 工作仍在 | 当前 | 能省点击和搜索时间的工具有可信 ROI 叙事 | OpenEvidence 能否量化节省的时间,或避免的收件箱 / 搜索负担? |
| 治理、隐私、验证和责任门槛 | 约束 | 86% 关注隐私、88% 关注验证,责任框架是最主要监管诉求 | 当前 | 没有信任控制,企业买方会挡住规模化 | OpenEvidence 提供什么验证包、审计轨迹和法律定位? |
| 警报疲劳和工作流过载 | 约束 | 高警报暴露会显著降低对新 CDS 警报的响应 | 当前 | 打扰式 UX 即使底层智能提升,也会压低采用率 | OpenEvidence 如何避免沦为又一个打断式警报源? |
| 在位厂商装机基础和 AI 延伸 | 约束 | UpToDate 和 DynaMed 已掌握可信工作流;Dyna AI Mode 把基础继续延伸 | 近期 | 新进入者必须打赢信任加工作流,而不只是模型能力 | 什么切换触发点能证明替换或叠加在位产品值得? |
这张表把需求侧驱动和卡口约束放在一起,因为同一股工作流压力既创造需求,也抬高采购标准。 时点指因素何时影响采用,不是底层来源何时发布。
[CM027, CM029, CM030, CM031, CM032, CM033]证据出版商、供应商编选、AI 综合、治理和工作流交付必须对齐,临床医生才会改变诊疗现场行为。
这条流展示价值链依赖,而不是收入分成。节点代表功能阶段,每一段都可能成为采用瓶颈。
[CM003, CM006, CM007, CM029, CM030, CM036]2.5 展示项
03竞争格局
3.1 竞争版图与直接同业
OpenEvidence 并不是在与单一可比产品竞争。最接近的直接重叠来自 AI 原生临床参考工具,它们承诺在床旁快速给出答案;但这个类别很快扩展到相邻的医生工作流套件,以及正在把生成式 AI 嫁接到精选内容上的既有证据产品。OpenEvidence 自己围绕来自授权期刊和指南、带引用链接的答案来描述产品;Doximity 则营销一个免费的 Ask 产品,并把 Pathway 的参考语料吸收入更宽的医生工作流套件。Glass Health 也是直接同业,不过其公开界面更轻地把自己定位在环境式记录加临床决策支持,而不是宽医院平台。 直接同业证据很重要,因为价格和分发已经开始分化。OpenEvidence 对临床医生仍免费并由广告支持;Pathway 原来付费的高级档,在 Doximity 以 $63 million 收购该公司后被折叠进 Doximity Ask。这一组合把一个曾经独立付费的临床参考产品,变成了 Doximity 网络里的免费功能;Doximity 称该网络触达超过 80% 至 85% 的美国医生。对 OpenEvidence 来说,近期竞争重点不再是证明临床医生想要 AI 搜索,而是当更大的工作流平台决定以零边际价格打包类似参考能力时,一个独立目的地能否守住注意力。 [CP001, CP002, CP003, CP005, CP008, CP009]
| 竞争对手 | 类别 | 规模 / 融资信号 | 目标细分 | 差异化 | 局限 / 风险 |
|---|---|---|---|---|---|
| OpenEvidence | 直接 AI 临床搜索 | D 轮估值 $12B;过去 12 个月约融资 $700M;迄今 >200M 次临床医生咨询 | 需要快速床旁证据查询的已认证临床医生 | 授权期刊 / 指南合作,加上带引文链接的答案和免费访问 | 广告支持模式、缺少明显付费席位锁定;必须防住捆绑和内部自建 |
| Doximity Ask (+ Pathway) | 直接 AI 临床参考 + 工作流套件 | 美国医生中 >80%-85% 是 Doximity 用户;Pathway 收购价 $63M | 已在使用 Doximity 工作流工具的美国认证临床医生 | 免费 Ask 产品把 Pathway 语料库与 Scribe、Dialer 打包 | 未来或可做企业变现,但当前免费捆绑会压缩类别定价 |
| Glass Health | 直接 AI CDS / 环境式病历记录同业 | 早期私营产品;抓取页面披露的公开规模有限 | 需要 AI 笔记记录加 CDS 的临床医生 | 官方定位覆盖环境式病历记录和临床决策支持 | 与更大对手相比,公开定价和信任细节较薄 |
| UpToDate Expert AI | 带 GenAI 的在位精选证据工具 | UpToDate 获 >3M 医疗专业人士信任;Wolters Kluwer 2024 年收入 €5.9B | 已标准化使用 UpToDate 的医院、集团和个体临床医生 | 专家撰写内容、来源 / 理由链接、企业部署、CME 和 EHR 场景 | 企业包装偏重报价,产品界面迭代慢于消费者 AI 工具 |
| DynaMedex | 带 AI 层的在位精选证据工具 | 2026 年 Best in KLAS;结合 DynaMed 与 Micromedex | 需要在一个工作流中同时查疾病和药物证据的护理团队 | Dyna AI,加上来自 Micromedex 的剂量和用药安全集成 | 公开定价仍不透明;分发依靠企业 / 图书馆渠道 |
| ClinicalKey AI | 企业证据套件 | Elsevier 企业产品;公开页面强调全组织部署 | 医院、临床医生、药师、学生 | 基于可信书籍、期刊、指南和 EHR 集成的负责任 AI | 相比 Doximity 或 OpenEvidence,自助属性更弱,医生网络效应也不明显 |
| Micromedex | 药物侧更重的替代 / 附加工具 | 受 >80 个国家信任 | 药房、毒理学、用药安全、药品目录团队 | 2,500+ 篇专论、700+ 个计算器、毒理学、RED BOOK 定价 | 床旁答疑范围窄于通用临床搜索工具 |
| Epocrates+ | 轻量替代工具 | $24.99/月或 $179.99/年 | 关注药物信息、ICD-10、相互作用和指南的个体临床医生 | 低价、自助、熟悉的移动端工作流 | 不是广义证据综合平台 |
| OpenAI 内部自建 | 潜在进入者 / 自建底座 | 已发布 API 定价;ChatGPT for Healthcare 有具名医院落地 | 自建 AI 副驾驶的医疗系统和厂商 | 透明引用、BAA、客户托管加密密钥、模板、API | 不掌握临床参考入口或独家医学内容 |
| Anthropic 内部自建 | 潜在进入者 / 自建底座 | 已发布席位和 API 定价;医疗产品支持 HIPAA | 自建工作流智能体的支付方、服务方、初创公司和企业 | CMS / ICD-10 / NPI / PubMed 连接器和面向 FHIR 的工作流 | 通用模型仍需证据锚定和治理,才能获得临床信任 |
这些行保留了截至 2026-05-25 可从抓取公开来源观察到、最影响决策的直接同业、在位厂商、替代品和潜在进入者。 AI 原生和通用模型厂商的公开融资或定价细节,明显比在位企业套件丰富。
[CP005, CP006, CP007, CP008, CP010, CP013]既有厂商在工作流触达上最靠右;OpenEvidence 和 UpToDate 在临床出处上得分最高;通用模型内部自建方案在即用分发上较低,但买方选择权更高。分数是有来源支撑的序数判断,不是厂商披露指标。
坐标轴为序数刻度,依据已获取证据中关于内容来源、可审查性、企业控制、临床医生覆盖和工作流嵌入的信息推导。图中有意比较相对竞争姿态,而不是精确市场份额或准确率分数。
[CP001, CP004, CP010, CP016, CP018, CP021]3.2 既有循证工具与现状替代品
对许多临床医生而言,现状替代品仍是一组精选参考资料,而不是单一 AI 答案引擎。Wolters Kluwer 的 UpToDate 营销个人、团队和企业套件,称其获得超过 3 million 名健康专业人士信任;其 Expert AI 发布也明确把生成式 AI 包在专家撰写、经同行评审的 UpToDate 内容外,并提供透明来源和推理链接。DynaMedex 同样把 DynaMed 疾病证据与 Micromedex 药物证据结合起来,并依托 KLAS 认可,以及 Dyna AI 向用药剂量和安全性的扩展。Elsevier 的 ClinicalKey AI、Micromedex 和 Epocrates 各自覆盖同一待完成任务的相邻切片:快速证据检索、药物安全、指南查找和工作流安全的参考访问。 这一点重要,因为机构医疗里的切换成本不只关乎答案质量。既有厂商已经嵌入 EHR 链接、CME 习惯、处方集工作流、采购预算和医院图书馆合同。它们也在防守更窄但黏性很强的使用场景。Micromedex 的药物相互作用、毒理学和 RED BOOK 定价数据足够专业,通用 AI 副驾不会自动替代它们。Epocrates 更便宜、更轻量,因此即便不是完整百科式证据工具,也能成为药物和指南查找的实用替代。OpenEvidence 因此竞争的不只是“更好的搜索”,还包括已经确立的临床习惯:这些习惯把受信参考、运营性药物数据库和本地政策混在一起。 [CP016, CP017, CP018, CP019, CP020, CP021]
| 购买标准 | OpenEvidence | Doximity Ask | UpToDate Expert AI | DynaMedex | ClinicalKey AI | OpenAI 内部自建 | Anthropic 内部自建 |
|---|---|---|---|---|---|---|---|
| 带引文链接的答案 | 强 | 强 | 强 | 中等 | 强 | 强 | 中等 |
| 独家授权医学内容 | 强 | 中等 | 强 | 中等 | 强 | 有限 | 有限 |
| 药物数据库 / 处方集深度 | 中等 | 中等 | 中等 | 强 | 中等 | 有限 | 有限 |
| 行政工作流自动化 | 有限 | 强 | 中等 | 有限 | 中等 | 强 | 强 |
| 医院治理 / HIPAA 姿态 | 强 | 强 | 强 | 中等 | 中等 | 强 | 强 |
| 企业 / EHR 嵌入 | 中等 | 中等 | 强 | 强 | 强 | 中等 | 中等 |
| 自定义连接器 / 构建界面 | 有限 | 有限 | 有限 | 有限 | 有限 | 强 | 强 |
| 自助价格透明度 | 强 | 强 | 有限 | 有限 | 有限 | 强 | 强 |
单元格是基于抓取到的公开产品页面、有证据支撑的分档判断。 '有限' 可能表示厂商没有公开强调该能力,并不等于合同下不可能提供。
[CP001, CP003, CP011, CP012, CP016, CP018]| 工具 | 公开套餐 | 公开定价信号 | 信任 / 控制信号 | 含义 |
|---|---|---|---|---|
| OpenEvidence | 免费临床医生产品 | 医生免费;广告支持 | 符合 HIPAA 和 SOC 2 Type II | 临床医生采用快,但显性的付费席位切换成本较弱 |
| Doximity Ask | 已验证临床医生免费层 + 企业询价 | 已验证的美国临床医生免费;企业需联系销售 | 符合 HIPAA;用户可包含 PHI | 激进免费捆绑可能压低独立付费工具 |
| Pathway(并入前) | 收购前的高级版应用 | 并入 Doximity Ask 前,高级版每年 $300 | 主打循证临床参考 | 说明参考功能可以从付费应用迁入免费捆绑包 |
| UpToDate | 个人、团体和企业订阅 | 年付、30/90 天循环订阅和团体选项;商店定价因国家而异 | 专家精选内容和企业治理 | 套餐结构清晰,但多数企业经济条款仍靠谈判 |
| DynaMedex | 机构 / 企业套件 | 抓取页面未列自助标价 | Best in KLAS 认可,并整合疾病 + 药物证据 | 主要靠采购渠道和可信度竞争,而不是公开价格透明度 |
| Epocrates+ | 月付或年付自助订阅 | 每月 $24.99 或每年 $179.99;医学生免费 | 知名药物参考和指南工作流 | 轻量查询任务里最便宜的自助替代品 |
| Micromedex | 企业 / 机构部署 | 抓取页面未列自助标价 | 药价、毒理学、处方集和计算器深度 | 价格不透明,但在药房和安全工作流里粘性仍高 |
| ClinicalKey AI | 企业级证据套件 | 抓取页面未列自助标价 | 基于可信 Elsevier 内容的负责任 AI,并接入 EHR | AI 作为更大内容合同里的升级项出售 |
| OpenAI | API + 医疗企业工作区 | GPT-5.5 每 1M tokens 输入 $5 / 输出 $30;网页搜索每 1K 次调用 $10 | BAA、客户托管加密密钥、不用客户内容训练 | 让医院和供应商看清内部自建经济账 |
| Anthropic | 消费者席位、Team/Enterprise 和 API | Team 每席每月 $20-$25;Opus 4.7 API $5 / $25 MTok | 提供 HIPAA 就绪产品;具备企业控制 | 自助试用和企业扩张路径并行 |
这些定价行把标价、打包方式和明确的询价制信号放在一起,因为多数既有 CDS 厂商不会在抓取到的公开产品页面披露企业合同经济条款。
[CP003, CP011, CP014, CP017, CP024, CP028]存量厂商的机构控制力得分最高,Doximity 的医生分发杠杆最高,OpenAI / Anthropic 的内部自建可选性最高。本图有意分析竞争控制力,而不是重新呈现产品能力表。
各单元格是序数判断,依据已获取证据中关于采购渠道、捆绑、定价透明度、监管可审查性和人工核验负担的信息。价格压力或核验负担列的“高”是风险分,不是优势分。
[CP004, CP015, CP020, CP029, CP031, CP033]3.3 相邻 AI 进入者与自建替代
OpenAI 和 Anthropic 在传统意义上不是精选医学参考资料,但它们是可信竞争进入者,因为现在都围绕通用模型销售医疗专用包装。OpenAI for Healthcare 强调对同行评审研究、临床指南和公共卫生来源的透明引用,以及 BAA、审计日志、客户管理加密密钥、客户内容不用于训练等企业控制。Anthropic 的医疗产品同样被包装为 HIPAA 就绪,并且越来越原生嵌入工作流:CMS 覆盖范围连接器、ICD-10 和 NPI 查询、PubMed 访问、FHIR 开发帮助、预授权技能和索赔申诉支持。两家供应商也都发布自助价格,而不是迫使每个买方进入不透明企业采购。 这个组合降低了内部自建门槛。已经拥有 IT 人员、治理和本地政策内容的医疗体系,越来越可以在基础模型之上拼出自己的临床医生 AI 副驾,而不是采用独立参考目的地。取舍在于信任和专用性。通用模型在连接器、自动化和行政工作流上可以推进更快,但要被临床接受,仍需要证据锚定、监督和机构护栏。对 OpenEvidence 来说,威胁不在于 OpenAI 或 Anthropic 成为最好的疾病百科,而在于它们让医院或工作流供应商能以低成本、行政上更方便的方式,在更大系统里复刻足够多的床旁搜索工作。 [CP027, CP028, CP029, CP030, CP031, CP032]
3.4 切换成本、护城河耐久度与反向证据
证据支持一个更细的护城河判断。OpenEvidence 通过官方内容合作、面向临床医生的定位,以及已公布的 HIPAA/SOC 2 控制,拥有真实信任优势。这一点重要,因为 FDA 对非设备临床决策支持的指引仍以临床医生可审阅性为中心:建议必须支持而不是取代判断,用户必须能够独立审阅答案依据。OpenEvidence 重引用的设计与这一要求方向一致;UpToDate Expert AI 通过暴露假设和来源推理,也在表达同一点。换句话说,来源可追溯性是竞争门槛,不只是产品功能。 但反向证据同样清晰。Doximity Ask 自己的 FAQ 称输出可能包含不准确之处,应始终验证;学术文献也仍把幻觉视为活跃的患者安全问题,即便在检索增强系统里也是如此。MDPI 的本地部署 RAG 研究显示缓解是可能的,但仍需要来源标签、审计轨迹和验证层。这使商品化风险在两类场景最高:工作流能容忍“足够好”的有证据锚定答案,且更大平台可以补贴分发。OpenEvidence 的护城河在授权内容访问和临床医生心智份额上最耐久,在通用问答机制上最不耐久,因为 Doximity 可以打包,既有厂商可以延伸既有合同,通用模型供应商也能持续把自建还是购买的边界推向医院和其他工作流所有者。 [CP004, CP018, CP027, CP030, CP033, CP034]
| 护城河主张 | 威胁 | 严重性 | 重要性 | 缓释措施 / 尽调问题 |
|---|---|---|---|---|
| 授权期刊和指南访问 | 内容所有方也向既有厂商或工作流平台授权类似访问 | 高 | 内容访问具备独家性或深度异乎寻常时,OpenEvidence 差异化最强 | 核验续约条款、独家边界和出版方集中度 |
| 临床医生采用和习惯养成 | Doximity 内免费捆绑包或嵌入 EHR 的既有厂商先抢走工作流 | 高 | 如果临床医生在既有工具里拿到“够用”的答案,流量可能转移 | 衡量 Doximity/UpToDate AI 上线后保留下来的日活使用 |
| 引用背书的信任定位 | 所有厂商的 LLM 幻觉仍需要核验 | 高 | 临床场景里的信任很脆弱,并直接关联医疗法律可辩护性 | 跟踪错误率、来源覆盖和机构级治理结果 |
| 免费产品分发 | 如果使用量转移或广告需求走弱,广告支持模式的经济性可能弱于企业捆绑 | 中 | 低价有助采用,但本身不能形成锁定 | 厘清广告集中度、变现深度和企业追加销售路径 |
| 独立入口体验 | 医院可以基于 OpenAI 或 Anthropic 自建内部 AI 助手 | 高 | 通用模型入局者降低自建成本、缩短上市时间 | 在并排工作流中测试专有内容和答案质量是否仍显著更好 |
| 药物查询广度 | Micromedex 和 Epocrates 在用药安全和处方集工作流里仍有粘性 | 中 | 通用医学搜索产品未必拿得下药房专项任务 | 评估 OpenEvidence 能否替代药物密集型工具,还是只能补充 |
| 既有厂商采购惯性 | UpToDate、DynaMedex 和 ClinicalKey 已在企业合同和 EHR 日常流程里 | 高 | 机构切换比消费者 AI 试用慢 | 量化医院内替代性中标与辅助性使用的比例 |
| 监管可解释性匹配 | AI 行为不透明或理由披露薄弱,可能把买家推向更可审查的工具 | 中 | FDA 指引强调临床医生可审查性和判断保留 | 审计引用质量、理由质量,以及日志 / 溯源深度 |
严重性是基于抓取公开来源的前瞻性投资判断,不是任何厂商模型输出。高表示未来一至三年内,对定价权或分发构成现实威胁。
[CP004, CP027, CP033, CP034, CP035, CP036]最影响决策的竞争数据点,结合了 OpenEvidence 的规模、Doximity 的触达、UpToDate 的存量覆盖,以及通用模型进入者公开的席位 / token 经济性。
这些项目有意混合规模、触达、打包和 API 经济性,因为这个市场的竞争就绪度取决于分发、信任和自建成本,而不是单一标准化收入倍数或基准分数。
[CP005, CP010, CP013, CP014, CP016, CP028]3.5 展示项
04财务情况
4.1 收入模式与商业化架构
OpenEvidence 的公开商业化模式对医疗软件而言并不常见,因为核心临床医生产品免费,而不是按订阅定价。官方和独立来源一致描述,该平台对已验证身份的美国医生免费,主要由答案加载时展示的药企和医疗器械广告资助。这个模式在财务上很关键,因为它把首次变现事件从席位销售转向查询量、医生专科结构和广告主需求。Sacra 的收入研究显示,广告引擎在规模化后可能具有可观经济性,估计 CPM 远高于一般消费者媒体,ARPU 约 $124;与公司相关的新闻报道则称 OpenEvidence 在 2025 年年度收入突破 $100 million。与此同时,公开来源也显示管理层正在尝试超越广告做多元化。Mount Sinai 的 Epic 部署、Sutter 的工作流合作,以及 Sacra 对未来非广告企业层的讨论,都暗示通向更高质量合同收入的路径;Veeva/Open Vista 合作则创造了另一条生命科学渠道。仍缺失的是定价透明度:没有留存公开来源披露企业 ACV、实施费、收入分成条款,或广告、企业和生命科学收入之间的拆分。[CI001, CI002, CI003, CI005, CI006, CI010]
| 收入流 | 机制 | 单位 | 当前数值 / 状态 | 收入质量 | 尽调问题 |
|---|---|---|---|---|---|
| 药企 / 医疗器械广告 | 在答案生成等待时间展示赞助位 | CPM / 活动花费 | 主要上线收入流;公司称产品对医生免费,靠广告支持 | 当前变现中高;可扩展性强,但受广告主预算和政策约束影响 | 索取广告主集中度、填充率、续约率,以及药企与器械花费结构 |
| 免费临床医生搜索量 | 咨询量创造展示库存和按专科定向的需求 | 每月咨询次数 | 2025 年 12 月约 18M 次月咨询;2026 年 1 月估计约 20M/月 | 本身不是收入,但支撑广告收益率的核心需求引擎 | 展示咨询到曝光的转化、每名临床医生平均会话数和未售库存 |
| 医疗系统企业部署 | 工作流集成、治理,以及潜在非广告合同访问 | 系统合同 / 席位 / 平台费 | 已宣布 Mount Sinai 企业级 Epic 部署;公开提及 Sutter 和其他系统 | 可能带来质量更高的经常性收入,但价格和合同结构未披露 | 提供标准合同计量、ACV 区间、实施负担和续约时间 |
| 工作流产品(Visits、Coding、Dialer) | 更宽的工作流模块把搜索之外的追加销售面做大 | 模块捆绑 / 企业附加项 | 2025-2026 年功能面扩张;无公开独立定价 | 战略扩张路径;变现未公开拆分 | 按模块拆分附加率、活跃用户和付费与免费使用 |
| Open Vista 生命科学渠道 | 与 Veeva 共建的 AI 平台,服务临床试验、药物发现和药物采用 | 战略合作 / 定制企业 | 首批 Open Vista 产品预计 2026 年推出;经济条款未披露 | TAM 可能很大,支付意愿高于临床医生 | 披露定价模型、商业归属和任何收入分成条款 |
| 二级市场载体 | 围绕 OpenEvidence 股权的集合投资基金活动 | 二级交易金额 | SEC Form D 显示 HII OpenEvidence-01 售出约 $5.76M;不是经营收入 | 不是核心收入流;只作为资本市场信号有意义 | 厘清二级流动性是否影响员工留任、内部人出售或股权结构压力 |
这些行把当前收入流、明确标注的新兴路径和资本市场信号放在一起。公开来源未披露实际收入结构或已实现定价。
[CI001, CI002, CI005, CI007, CI010, CI011]| 产品 / 合同界面 | 公开标价 | 买方 / 单位 | 已知情况 | 未知事项 | 来源视角 |
|---|---|---|---|---|---|
| 核心临床医生访问 | $0 | 已验证美国临床医生 | 官方和新闻来源称平台免费,由广告支持 | 广告负载、展示频率和每用户实际收入未披露 | BusinessWire, CNBC, Fierce, Sacra |
| DeepConsult | $0 | 已验证美国临床医生 / 每次运行 | 尽管计算量和成本超过标准搜索 100x,仍免费 | 使用上限、补贴经济性,以及重度用户是否触发内部限额均未公开 | 来源:PR Newswire、HLTH、HIT Consultant |
| Visits / 笔记生成 | 未披露 | 临床医生 / 诊所 / 医疗系统 | 工作流模块已存在,具备 HIPAA 和 SOC 2 姿态;定价未公开 | 独立定价还是纳入捆绑包未知 | OpenEvidence 用户指南 |
| Dialer / 消息 / 传真 / 编码智能 | 未披露 | 临床医生 / 诊所 / 企业 | 2026 年功能发布扩大工作流覆盖面 | 是靠广告、企业捆绑还是未来按席位层级变现,仍未知 | OpenEvidence 用户指南 |
| 医疗系统企业部署 | 未披露 | 医院 / 系统合同 | Mount Sinai 和 Sutter 证明企业销售已存在 | 席位计量、平台费、实施费和无广告高级版均未披露 | Mount Sinai, STAT, Sacra |
| 非广告支持企业层 | 未披露 | 大型医疗系统 | Sacra 称独立的非广告支持版本正在开发 | 上线时间、标价和利润率轮廓未知 | Sacra |
| Open Vista / Veeva 生态 | 未披露 | 生命科学公司 / 项目 | 合作面向试验访问、药物发现和采用分析 | 商业模式、买方和收入确认未公开 | PR Newswire Veeva 发布 |
公开标价只存在于 $0 的临床医生层。所有付费或新兴合同界面在留存来源中仍未披露。
[CI001, CI011, CI013, CI016, CI017, CI027]展示免费临床医生采用如何在当前转成广告收入,并打开企业与生命科学变现路径。
[CI001, CI005, CI010, CI011, CI012, CI017]4.2 公开牵引力与 GTM 质量
公开牵引力足以支撑真实收入基础,但仍高度依赖公司口径指标。BusinessWire、CNBC、Fierce 以及 2025 年 7 月相关融资报道都描述了非常快的临床医生采用:超过 40% 的美国医生每天使用产品,超过 10,000 家医院和医疗中心被产品触达,到 2025 年 12 月月咨询量约 18 million 次,2025 年中更早运行量为每月 8.5 million 次咨询,且每月新增 65,000 名已验证临床医生。Sacra 认为免费访问帮助 OpenEvidence 绕过了通常拖慢医院软件的约 18 个月采购周期。这个 GTM 优势可信:公司可以自下而上获取临床医生,证明工作流习惯,然后再销售治理和集成。问题在于,公开销售效率数学并不完整。没有披露 CAC,没有披露企业成交率,没有公开的医疗体系合同席位数,也没有经审计的从咨询量到广告库存或企业 ARR 的转化。因此,如果医院和生命科学合同放量,收入质量会改善;但今天的公开证据仍指向一个牵引力亮眼、变现结构却不透明的业务。[CI002, CI003, CI006, CI007, CI008, CI009]
围绕收入、利润率、广告收益和估值输入给出可由公开资料支撑的区间视图;每一项要么来自公司口径,要么由第三方明确估计。
低 / 基准 / 高值结合公司声明(收入 > $100M)与 Sacra 估计(收入 $150M、毛利率约 90%、ARPU 约 $124、CPM 为 $70-$1,000+)。本图是情景框架,不是经审计的公司预测。
[CI002, CI003, CI004, CI005, CI021, CI022]4.3 成本结构与利润率路径
相比硬件或医疗服务业务,OpenEvidence 看起来资本较轻,但运营并不便宜。已披露成本栈集中在算力、模型训练、内容授权、合规和商业化,而不是工厂或库存。管理层称 Series D 资金将主要用于 R&D、模型训练、算力和扩展内容合作;2025 年 7 月发布材料称,每次 DeepConsult 运行消耗的算力和成本超过标准搜索的 100 倍,尽管该功能对已验证临床医生仍免费。这形成熟悉的权衡:免费产品加速采用,但高算力高级功能最终必须由广告、企业合同或未来付费层补贴。授权依赖是第二个主要成本和利润率变量。OpenEvidence 的差异化依赖与 NEJM、JAMA、NCCN、Wiley/Cochrane 和其他内容所有者的协议;R&D World 和 Sacra 都把这些合作视为护城河核心。因此,公开毛利率证据很薄:Sacra 约 90% 的估计对软件加广告模式可能方向上合理,但它未经审计;如果授权费、推理强度或企业支持义务快于收入扩张,估计可能过高。反向来源还提出了更难的问题:药企资助模式可能有利润,但仍会面临发表偏倚、信任和赞助方冲突压力,最终限制付费意愿,或把医疗体系推向无广告配置。[CI004, CI005, CI013, CI014, CI015, CI016]
| 指标 | 数值 / 区间 | 置信度 | 重要性 | 尽调问题 |
|---|---|---|---|---|
| 2025 年年度收入(公司口径) | >$100M | 中 | 说明广告模式在规模化后已有实质意义 | 索取经审计 2025 年收入和月度期末收入运行率桥表 |
| 2025 年年化收入(Sacra 估计) | $150M | 中 | 为当前规模和估值倍数分析设定上限 | 要求管理层对公司口径和分析师估计做调节 |
| 估计毛利率 | ~90% | 低-中 | 如果算力和授权成本可控,经济性接近软件 | 提供经审计毛利率和按产品划分的直接成本分配 |
| 估计 ARPU | ~$124 | 低-中 | 衡量每名临床医生变现效率的有用代理 | 按专科和参与度队列披露临床医生变现 |
| 估计广告 CPM 区间 | $70-$1,000+ | 低-中 | 解释医生注意力为何能异常高效变现 | 按赞助商类型和库存类别提供已实现 CPM |
| 月咨询量 | 2025 年 12 月约 18M;2026 年 1 月估计约 20M | 中 | 推动广告库存和企业相关性的核心活动指标 | 按月和临床医生队列展示登录后的临床咨询 |
| 每日医生渗透率 | >40% 的美国医生 | 中 | 反映习惯强度和自下而上 GTM 效率 | 提供准确 DAU、WAU 和分母方法 |
| 新临床医生注册量 | 每月 65,000+(2025 年中) | 中 | 自然获客速度的代理指标 | 披露付费获客支出、推荐占比和激活率 |
| DeepConsult 计算负载 | >100x 标准搜索 | 中 | 重要,因为高阶功能如果保持免费,可能挤压利润率 | 提供每次 DeepConsult 运行的边际成本和使用上限 |
| 企业 ACV / 合同定价 | Unknown | None | 决定企业业务能否比广告改善收入质量 | 分享企业合同、续约条款和实施经济性 |
| 广告主集中度 / 留存 | Unknown | None | 评估广告支持收入波动性的关键 | 提供头部客户集中度和续约 / 追加销售数据 |
本表有意把公司宣称指标、分析师估计和未披露字段拆开。未知表示留存来源中未公开披露,不代表为零。
[CI002, CI003, CI004, CI005, CI007, CI008]定性单位经济模型桥接,说明只要广告收益和企业转化快过算力与授权成本,免费临床医生使用仍可增厚利润率。
公开来源只披露了部分数字。因此节点把少量公开估计(ARPU、毛利率代理、DeepConsult 计算倍数)与定性成本项合并展示。
[CI004, CI005, CI013, CI014, CI015, CI016]4.4 资本充足性与尽调阻塞项
OpenEvidence 的融资节奏显示其股权资本获取能力充足,但公开披露不足以让人有信心承销现金跑道。公司在 2025 年 7 月融资 $210 million,2025 年 10 月融资 $200 million,2026 年 1 月融资 $250 million;官方和新闻来源都称截至 2026 年初累计融资约 $700 million。相对于公开估计的收入规模,这是一笔可观资本,资金用途也指向增长投资,而不是紧急再融资。不过,同一组来源留下了核心偿付能力问题:没有公开现金余额,没有月度烧钱数字,没有现金跑道披露,也没有说明多少收入来自经常性企业软件、多少来自周期性广告支出。SEC 证据还增加一层细节:近期唯一与 OpenEvidence 相关的 Form D 结果,是与二级市场交易相关的集合投资工具;它能证明市场兴趣,但不能替代运营公司的透明度。财务上,这门业务看起来有前景,但仍高度依赖若干条件:它今天很可能能为产品扩张供血,但下一阶段承销取决于企业合同和生命科学产品能否在广告模式冲突、赞助方集中度或算力与授权成本压缩利润率叙事之前成熟。[CI019, CI020, CI021, CI022, CI023, CI024]
| 项目 | 数值 / 状态 | 含义 | 来源 |
|---|---|---|---|
| Series B 轮(2025 年 7 月) | 融资 $210M,估值 $3.5B | 在采用加速时,为早期规模化和 DeepConsult 发布提供资金 | 来源:PR Newswire、HLTH、HIT Consultant |
| Series C 轮(2025 年 10 月) | 融资 $200M,估值约 $6B | 显示一个季度内投资人胃口快速抬升 | TechCrunch、Sacra、市场数据摘要 |
| Series D 轮(2026 年 1 月) | 融资 $250M,估值 $12B | 当前基准轮;估值约三个月翻倍 | BusinessWire, CNBC, Fierce, Cooley |
| 已披露总融资 | 到 2026 年初约 $700M | 相对公开收入估计,显示股权资本获取能力强 | BusinessWire, CNBC, Fierce, Sacra |
| 计划资金用途 | 研发、模型训练、算力和内容授权 | 说明资金用于加深护城河,而不是购置重资产 | BusinessWire, Fierce, Cooley |
| 账上现金 / 烧钱 / 现金跑道 | 未知——未公开披露 | 论证资本充足性的主要阻碍 | 证据缺口 |
| 债务 / 项目融资义务 | 未发现公开披露;仅观察到二级市场集合基金申报 | 显示没有明显项目融资负担,但未披露不等于没有债务 | SEC EDGAR、留存来源梳理 |
融资时间线仅在其说明当前资本充足性时纳入。现金、烧钱、现金跑道和债务在公开材料中仍未披露。
[CI019, CI020, CI021, CI022, CI023, CI024]| 缺失指标 | 对分析的影响 | 尽调路径 |
|---|---|---|
| 账上现金、月度烧钱和现金跑道 | 阻断项——没有这些就无法论证资本充足性 | 索取董事会材料或 CFO 更新,包含当前现金、月度烧钱,以及基准 / 下行情景现金跑道 |
| 广告 vs 企业 vs 生命科学收入结构 | 阻断项——没有结构就无法判断收入质量 | 按收入流和客户类型索取 2025 年及 2026 年 Q1 收入桥表 |
| 企业合同定价和非广告层条款 | 重大——决定企业收入能否改善耐久度和毛利率 | 获取 MSA 样本、价目卡、实施 SOW 和续约历史 |
| 广告主集中度、续约和销售集中度 | 重大——广告依赖可能比表面收入更强 | 索取前 10 大广告主占比、留存、季节性和按赞助商类型划分的管线 |
| 内容授权费用和收入分成义务 | 重大——没有直接内容成本,利润率叙事不可靠 | 索取授权明细、最低保证,以及合作方分成 / 独家条款 |
| 经审计报表和产品级毛利率 / 算力成本 | 重大——公开主张依赖公司表述和分析师模型 | 索取经审计的 2025 年 P&L,以及搜索、DeepConsult 和企业部署的产品级单位成本报告 |
以下缺口是从叙事热情推进到可论证财务模型所需的最低尽调包。
[CI025, CI034, CI035, CI036, CI037, CI038]这张股权融资增长图展示公开资料显示 OpenEvidence 可能把现金投向哪里,以及哪些地方对投资者仍是黑箱。
公司未公开现金余额或烧钱速度,因此本图聚焦资金来源和用途,而不是量化现金跑道桥接。
[CI014, CI019, CI020, CI021, CI022, CI023]05产品与技术
5.1 用医生工作流定义产品
OpenEvidence 已经不只是医生搜索框。其公开用户指南界面覆盖快速咨询、指南、预授权和文书、药物信息与安全、临床试验招募、计算器、患者材料、医学教育、收件箱管理、Deep Consult、DotFlows、Voice Mode、Visits、笔记生成、账单编码、就诊后摘要、排程、模板、患者 Q&A、Voices、Dialer、Fax 和 Messages。按工作流看,产品想横跨三个反复出现的任务:快速床旁证据检索、复杂病例的更重文献综合,以及证据答案产出后的相邻文档 / 行政工作。Voice Mode 把咨询工作流延伸到查房或在患者之间移动等免手场景,Deep Consult 则把更难的临床问题包装成较长篇的研究智能体。移动端分发重要,因为平台现在有原生 iOS 和 Android 应用,承诺提供与网页产品相同的带引用答案体验;Google Play 还明确把该应用描述为仅面向已验证医疗专业人士的诊疗现场决策支持。[CE001, CE002, CE003, CE004, CE005, CE006]
| 模块 / 产品线 | 主要用户 | 状态 / 成熟度 | 差异化 | 尽调缺口 |
|---|---|---|---|---|
| Core Ask / 临床咨询 | 医生和其他已验证临床人员 | GA;核心产品界面 | 自然语言临床问答,答案附引用,并扎根于授权医学证据 | 复杂病例没有公开延迟基准,也没有引用忠实度审计 |
| Deep Consult | 处理复杂研究问题的医生 | GA;用户指南已有记录 | 高级智能体会生成长篇研究报告,而不是单轮答案 | 未公开披露完成时间、模型路由或失败率 |
| Voice Mode | 查房中或接诊间隙使用移动端和网页端的临床人员 | 2026 年 GA | 免手操作的语音工作流,会在会话内保留书面转录和参考文献 | 语音输入或嘈杂临床场景没有公开准确率基准 |
| Visits / 文档和编码工作流 | 记录就诊过程的医生和护理团队 | 正在扩展;2026 年已公开描述就诊转录和编码 | 把就诊工作流与 CPT、E/M 依据和 ICD-10 生成合在一起,在就诊结束时完成 | 公开材料没有展示独立的账单准确率验证或拒付影响 |
| Messages / Fax / Dialer 工具 | 管理患者随访和沟通的诊所 | 正在扩展;2026 年用户指南和媒体报道已露出 | 把患者消息和文档流留在同一个临床人员工作台内;Messages 增加了同意控制 | 审计日志、留存和投递可靠性的公开细节有限 |
| 企业级 Epic 部署 | 医院临床人员,包括护士和药剂师 | Mount Sinai 和 Cedars-Sinai 的生产级企业上线 | 把证据搜索嵌入 Epic;Cedars 还在 OpenEvidence 体验中加入患者感知上下文 | 公开材料没有披露 SLA 条款、实施周期或各站点使用结果变化 |
各行只反映公开记录中的产品界面。多项能力被归在一起,是因为公司在公开文档中把它们作为相邻工作流功能暴露,而不是单独定价的 SKU。
[CE002, CE003, CE004, CE005, CE014, CE016]一个临床问题如何从医生工作流进入带引用答案或下游动作。
[CE003, CE014, CE015, CE016, CE017, CE019]5.2 证据引擎、内容依赖与验证模型
公开材料把 OpenEvidence 描述得不像通用模型公司,更像围绕授权医学内容和临床医生验证搭建的证据交付层。官网和合作公告显示,语料库持续扩张:1990 年以来的 NEJM Group 内容和多媒体,JAMA 及其专科期刊全文和多媒体,NCCN 肿瘤指南加 JNCCN 内容,以及 Wiley 更广的组合,包括 Cochrane 系统综述、Cochrane Clinical Answers 和数百种期刊与图书。Wiley 自己的公告把运营原则说得异常明确:“好料进,好料出”;专门模型在经同行评审文献上训练,而不是开放互联网,每个答案都基于医生可下钻和验证的来源。这就是 OpenEvidence 愿意公开披露的核心产品架构:临床医生提出问题,中间是授权证据语料库加模型综合,输出带引用答案。同样的验证模型也出现在 Voice Mode 中,语音答案旁边保留书面转录和参考文献;NCCN 合作也强调来源文档和回链到底层指南材料。[CE007, CE008, CE009, CE010, CE011, CE012]
| 层 / 流程 / 组件 | 作用 | 依赖 | 风险 |
|---|---|---|---|
| 临床人员界面层 | 网页端、原生移动端和 Voice Mode 入口,用于自然语言提问 | OpenEvidence 应用界面和已验证临床人员访问 | 公司声称多界面一致,但缺少公开的设备级质量指标 |
| 研究编排层 | Ask 负责快速检索,Deep Consult 负责更长篇的智能体综合 | OpenEvidence 自己的工作流设计和提示词编排 | 公开来源没有披露模型路由、供应商组合或回退行为 |
| 授权证据语料库 | 提供支撑答案的文献和指南材料 | NEJM、JAMA、NCCN、Wiley/Cochrane、FDA、CDC 以及其他内容合作伙伴 | 覆盖广度取决于持续授权,在合作伙伴领域之外可能仍不完整 |
| 引用 / 验证层 | 给临床人员提供可下钻来源、转录、参考文献和来源文档 | 带来源链接的答案渲染和合作伙伴 QA 义务 | 出现引用本身不能证明生成答案具备因果层面的忠实度 |
| 企业上下文层 | 为医院工作流和患者感知使用场景注入 EHR 上下文 | Mount Sinai 和 Cedars-Sinai 的 Epic 集成 | 公开材料没有说明患者感知答案的数据流边界、留存窗口或安全审查 |
| 文档 / 沟通扩展 | 把证据检索转成就诊记录、编码、消息、传真和电话工作流 | 内部工作流模块加临床人员采用 | 每增加一个工作流,都放大了需要质量、隐私和可靠性保障的表面积 |
这是一张运营模型图,不是隐藏模型栈声明。它只反映公开来源中的界面、内容依赖和工作流层。
[CE003, CE004, CE010, CE011, CE012, CE013]公开披露的运营栈,从临床医生输入到授权证据和验证输出。
本图仅反映公开运营模型。OpenEvidence 未在本次审阅的来源中公开披露模型路由、基础设施供应商或底层推理架构。
[CE003, CE010, CE012, CE013, CE016, CE026]塑造 OpenEvidence 证据质量、工作流触达和产品风险的主要外部依赖。
[CE009, CE010, CE011, CE012, CE016, CE021]5.3 部署模型、企业集成与工作流延伸
OpenEvidence 的部署故事越来越绑定医疗体系工作流集成,而不是独立浏览器使用。Mount Sinai 2026 年 3 月推出时,让完整临床护理团队都能在电子健康记录内部访问产品,明确包括医生、注册护士和药师。OpenEvidence 与 Hit Consultant 的描述都强调,Epic 嵌入解决了“最后一公里”问题:临床医生可以在既有 Epic 工作流里用自然语言提出医学问题,并得到基于经同行评审文献和临床指南的答案,而无需切换到外部研究工具。Cedars-Sinai 又向前一步。其 2026 年 5 月合作被描述为感知患者上下文的临床智能,把来自 Epic 的患者上下文直接集成进 OpenEvidence,使答案能纳入既往手术、合并症、用药、过敏和纵向病史。在这一核心问答流周边,公司已经开始产品化相邻工作流延伸:就诊转录、编码支持、Voices、带同意控制的 Messages、Fax 和医生拨号器。这扩大了产品有用性,但也提高了证明可靠性的负担,因为它要支撑的临床触点远多于纯证据搜索引擎。[CE014, CE015, CE016, CE017, CE018, CE019]
| 用户任务 | 当前工作流 | OpenEvidence 方案 | 可衡量的公开证据 | 局限 |
|---|---|---|---|---|
| 床旁或走廊里的临床问题 | 临床人员本来要在时间压力下搜索 PubMed、指南或浏览器标签页 | Core Ask 加带引用答案,可在网页端、移动端或 Voice Mode 使用 | Voice Mode 可在网页端和移动端使用,并把书面参考文献与语音答案放在一起 | 公开来源没有按护理场景披露答案延迟中位数 |
| 罕见病例的复杂文献综合 | 手工搜索、筛选并综合大量研究 | Deep Consult 高级研究智能体 | 产品定位是为复杂问题生成博士水平研究报告 | 完成时间、引用错误率或临床人员接受度没有公开基准 |
| 查阅病历时的 EHR 内团队会诊 | 单独登录和外部检索拖慢采用 | Mount Sinai 为医生、护士和药剂师提供嵌入 Epic 的访问 | 2026 年 3 月上线明确把企业访问扩展到完整护理团队 | 公开来源没有给出部署成本、激活率或留存使用指标 |
| 患者感知证据检索 | 临床人员必须在脑中把文献与患者病历拼在一起 | Cedars-Sinai 把来自 Epic 的患者上下文直接整合进 OpenEvidence | 公开描述包括既往手术、合并症、用药、过敏和纵向数据 | 加入患者上下文后,引用忠实度没有外部验证 |
| 就诊后的文档和随访工作 | 编码、消息、传真和随访任务经常流入不同工具 | Visits、Coding Intelligence、Messages、Fax 和 Dialer 界面 | 2026 年公开材料描述了自动生成 CPT/E/M/ICD-10 和消息同意控制 | 临床、收入周期和沟通结果的公开证据仍然很薄 |
可衡量收益列记录的是公开文档已有内容,不是有保证的 ROI。大多数部署经济性和工作流节省主张仍需要客户尽调。
[CE003, CE004, CE015, CE017, CE019, CE020]5.4 隐私、访问控制与公开信任姿态
OpenEvidence 最清晰的公开信任承诺围绕 HIPAA 处理、临床医生准入和用户可控分享。公司在 2025 年 4 月宣布完全符合 HIPAA,并称输入 PHI 的受规实体会在业务伙伴协议下操作。公司还表示,对话默认私密,Share 按钮让用户控制对话访问权限。访问侧,Google Play 页面称产品仅面向医疗专业人士开放,使用前需要用户验证。这些控制很重要,因为产品正从通用文献搜索走向感知患者上下文和文档相邻工作流,PHI、沟通渠道和运营草稿输出都会进入范围。能见度较低的是企业买方通常会要求的更深技术信任栈:公开可用性承诺、延迟目标、状态历史、第三方安全架构细节,或当患者上下文和工作流自动化叠加之后,对引用忠实度的外部审计证据。因此,公开姿态在政策控制和来源锚定上较强,在可独立审计的运营透明度上较弱。[CE023, CE024, CE025, CE026, CE027, CE036]
| 控制 / 质量信号 | 状态 | 范围 | 缺口 |
|---|---|---|---|
| HIPAA 合规公告 | 已公开宣布 | 允许在产品中安全上传 PHI | 公开材料除了公告外,没有列举技术保障措施 |
| 商业伙伴协议(BAA) | 已公开说明 | 适用于选择输入 PHI 的受监管实体 | 公开文档没有完整展示条款、例外和审计流程 |
| 默认私密的对话 | 已公开说明 | 用户对话的默认设置 | 公开材料没有详细说明默认留存或管理员覆盖政策 |
| 分享控制 | 已公开说明 | 用户控制谁能访问已分享对话 | 未找到公开的企业审计日志或权限模型文档 |
| 已验证临床人员访问门禁 | 已公开说明 | 产品只向已验证医疗专业人士 / NPI 门禁用户开放 | 受训人员、国际用户或代理人的验证边界案例没有公开细节 |
| 来源文档和参考文献 | 已公开说明 | NCCN 内容扎根于来源文档;Voice Mode 会把书面参考文献与转录一起保留 | 复杂工作流或患者感知工作流中的引用忠实度未见外部审计 |
| 公开性能透明度 | 本轮审阅的公开来源未披露 | 应覆盖生产部署的正常运行时间、事故历史和延迟 | 缺少公开状态页、SLA 细节或基准材料,仍是重要尽调缺口 |
这里列出的控制是面向公开市场的声明或工作流能力。企业信任尽调仍应直接索取安全架构、管理员控制、事件处理和审计证据。
[CE023, CE024, CE025, CE026, CE027, CE036]5.5 差异化、路线图与技术风险图谱
OpenEvidence 的差异化并不是披露了新颖模型架构,而是一个原生嵌入工作流的证据栈:它把授权的同行评审文献和指南内容直接放进临床医生工作流,再在其上叠加带引用答案、语音、感知患者上下文的 EHR 上下文和文档工具。公开材料可见的产品路线图也体现了这一点:2025 年 HIPAA 与移动应用,2025 年末 NCCN 和更广肿瘤指南授权,2026 年 3 月 Coding Intelligence,2026 年 4 月 DotFlows,到 2026 年 5 月 Cedars 感知患者上下文能力和 Voice Mode。规模宣称很大——超过 40% 的美国医生、超过 10,000 家医院和医疗中心、860,000 名临床医生,以及单日 1 million 次医生-AI 咨询里程碑——但主要技术尽调问题都在下行场景。第一,证据语料库质量高,并不意味着幻觉风险消失;更广泛的医疗 LLM 研究仍发现推理失败时有患者伤害风险。第二,即便在带引用或类 RAG 系统中,可解释性仍不完整,因为引用本身不能证明生成答案在因果上忠实。第三,覆盖广度取决于出版方和医学会授权能否持续。第四,公开证据对可用性、延迟和复杂病例基准表现的支撑,仍显著薄于采用叙事。[CE012, CE021, CE022, CE028, CE029, CE030]
| 日期 / 阶段 | 功能 / 里程碑 | 状态 | 含义 | 来源 |
|---|---|---|---|---|
| 2025-02 | NEJM Group 内容协议 | 已宣布 | 加入 1990 年以来 NEJM 系列内容和多媒体,用来支撑答案 | OpenEvidence NEJM 公告;Medical Economics |
| 2025-04 | HIPAA 合规和安全 PHI 上传 | 已宣布 | 让平台更接近承载患者数据的工作流 | OpenEvidence HIPAA 公告 |
| 2025-06 | JAMA Network 战略内容协议 | 已宣布 | 加入 JAMA、JAMA Network Open 和 11 本专科期刊的全文与多媒体 | JAMA 媒体发布;OpenEvidence JAMA 公告 |
| 2025-11 | NCCN 指南授权和 JNCCN 集成 | 已宣布 | 加入肿瘤专科指南内容,并带来源文档和参考文献 | NCCN 新闻稿;OpenEvidence NCCN 公告 |
| 2026-03 | Wiley / Cochrane 内容扩展 | 已宣布 | 加深文献覆盖,并扩大可返回带引用答案的问题范围 | Wiley 新闻室;OpenEvidence Wiley 公告 |
| 2026-03 | Mount Sinai Epic 企业上线 | 已宣布 | 首个企业级规模部署,把访问扩展到医生、护士和药剂师 | OpenEvidence Mount Sinai 公告;Hit Consultant |
| 2026-03 | Coding Intelligence | 已在产品文档和媒体中宣布 | 把产品从搜索延伸到编码和笔记补全工作流 | OpenEvidence 用户指南;Fierce Healthcare |
| 2026-04 | DotFlows 定制 | 已宣布 | 加入可复用的自然语言提示词工作流和社区分享模式 | OpenEvidence DotFlows 公告 |
| 2026-05 | Cedars-Sinai 患者感知临床智能 | 已宣布 | 在企业工作流内,把 Epic 派生的患者上下文加入证据检索 | OpenEvidence Cedars 公告;Fierce Healthcare |
| 2026-05 | Voice Mode 全面可用 | 已公开上线 | 在网页端和移动端加入免手操作的语音到语音证据检索 | OpenEvidence 语音指南;Fierce Healthcare |
本表只捕捉公开宣布的里程碑,并不保证每项功能在所有客户部署中的成熟度完全相同。
[CE003, CE005, CE009, CE010, CE011, CE012]基于已披露功能面和证据,比较 OpenEvidence 主要公开能力的相对成熟度。
这里的成熟度指公开证据支持的成熟度,不是内部技术就绪度。公开性能透明度低,反映的是缺少公开 SLA、延迟和基准测试细节,而不是证明产品薄弱。
[CE003, CE004, CE014, CE016, CE019, CE021]06客户情况
6.1 客户分层与采用基础
OpenEvidence 最清晰的客户分层是个人临床医生,尤其是医生用户。访问仅限已验证的医疗专业人士;网页和应用商店界面反复把产品定位为免费的自助式诊疗现场工具,而不是先通过集中医院采购出售的软件。这个定位很重要,因为头部采用数字描述的是使用情况,不是已签约收入账户。官方界面称,超过 40% 的美国医生每天使用该产品,覆盖 10,000 多家医院和医疗中心;NBC 则报道称,2026 年 4 月约 65% 的美国医生、约 650,000 名活跃美国医生用户在使用。数字足够大,能证明医生采用有意义,但它们仍大多是公司定义指标,没有公开的分母桥接。 因此,经济地图是多边的。医生是主要用户,也是核心获客切口。足够多临床医生已经依赖工具之后,医疗体系会在漏斗后段成为机构买方或付款方。药企和医疗器械广告主似乎为今天的免费产品提供资金;ACOG、Wiley、NEJM、JAMA、NCCN 和 Cochrane 等内容合作伙伴提升产品质量,但不应被误认作客户验证。结果是分发很有吸引力,变现观感却复杂:OpenEvidence 可能在公开记录显示谁付款、付多少、关系多黏之前,就已经拥有异常深的医生触达。[CU001, CU002, CU003, CU004, CU008, CU010]
| 分层 | 买方 / 用户 / 付款方 | 主要使用场景 | 规模 / 证据 | 收入或战略价值 | 缺口 |
|---|---|---|---|---|---|
| 自助注册医生 | 用户:医生;买方:自助注册临床人员;付款方:通常未披露 | 即时护理证据搜索、治疗方案、指南回忆、研究支持 | 官方声称每日 40%+;NBC 称 65% / 约 650k 名活跃美国医生 | 主要分发切入点,也是最强采用证据 | 医生队列层面的付费转化、流失或专科结构没有公开 |
| 企业系统中的多学科护理团队 | 用户:医生、护士、药剂师、治疗师;付款方:卫生系统 | 嵌入 EHR 工作流内的临床决策支持 | Cedars-Sinai 和 Mount Sinai 部署 | 展示了从医生牵引采用走向机构预算线的路径 | 席位数和合同经济性未公开 |
| 卫生系统 / 医院 | 买方和付款方:卫生系统;用户:护理团队 | 企业授权、治理和工作流集成 | Cedars-Sinai、Mount Sinai,以及二级来源中的 Sutter 证据 | 可能是比广告支持的自助服务更高价值的变现层 | 公开客户名单很小,续约数据缺失 |
| 医学会和期刊 | 合作伙伴,不是客户 | 授权或供应临床内容和指南 | ACOG、Wiley、NEJM、JAMA、NCCN、Cochrane 等内容伙伴 | 提升信任、专科深度和工作流相关性 | 这些 logo 是合作伙伴证明,不是客户证明 |
| 药企和器械广告主 | 付款方 / 赞助方,不是终端用户 | 围绕临床研究工作流中的临床人员注意力投放广告 | NBC、Sacra 和 App Store 广告披露 | 免费医生访问的当前变现引擎 | 头部赞助商集中度和重复投放留存未披露 |
| 未来生命科学 / 企业相邻场景 | 潜在付款方 | 更高价值的工作流、分析或专科产品 | Sacra 和 STAT 把企业扩张描绘为下一阶段 | 如果签约,可能让收入跳出广告 | 当前证据偏方向性,还不是客户证明 |
各行区分用户、买方、付款方和合作伙伴角色。收入价值只指公开记录能支撑的内容;支出和续约留白是真实的尽调缺口,不是漏掉了分析。
[CU001, CU002, CU003, CU004, CU011, CU012]| 指标 | 数值 | 日期 / 期间 | 来源 | 置信度 | 含义 / 缺失分母 |
|---|---|---|---|---|---|
| 新增已验证临床人员注册 | 每月 65,000+ | 2025-07 新闻稿 | PR Newswire | 中 | 显示自助服务获客漏斗顶部增长很快,但没有扣除净流失 |
| 月度临床咨询量 | 8.5M+ | 2025-07 新闻稿 | PR Newswire | 中 | 证实到 2025 年中已有高频重复使用,但未披露独立用户数 |
| 月度临床咨询量 | 每月 ~20M | 2026-01 | Sacra | 中 | 暗示到 2026 年参与度增长强劲,但方法论不公开 |
| 单日咨询量 | 24 小时内 1M | 2026-03-10 | OpenEvidence / Newswise / Sacra | 中 | 峰值使用证明很强,但不等于每日活跃独立临床人员 |
| 美国活跃医生触达 | 约 65% / ~650k 名医生 | 2026-04 | NBC | 中 | 独立媒体印证了极大触达,但源头仍是公司 |
| 美国医生日触达 | >40% | 当前 2026 年公司界面 | OpenEvidence / Fierce | 中 | 使用强度看起来很高,但分母和均值方法未披露 |
| 移动生态代理指标 | 4.80 评分;3,706 条评价;100k+ Play 下载;~420k 生命周期下载 | 2026-05 | AppBrain | 中 | 独立移动分发代理指标,但应用下载不等于已验证临床人员 |
各行混合了公司声明、媒体采访和应用生态信号。它们支撑方向性的采用增长,但没有给出可调和的经审计仪表盘。
[CU003, CU004, CU005, CU006, CU007, CU008]| 实体 / 角色 | 客户 / 合作伙伴 / 付款方分类 | 公开证据 | 战略含义 | 局限 |
|---|---|---|---|---|
| 单个医生 | 用户与获客切入口 | 官网、应用商店、NBC、App 评分 | 核心分发引擎,也是最强采用信号 | 公开证据大多是使用,不是付费 |
| 医疗系统(Cedars-Sinai、Mount Sinai、Sutter) | 新兴买方 / 付款方 | 企业部署公告和二手报道 | 最可能通向可持续合同和更广席位扩张 | 收入范围和续约未披露 |
| 护士、药师、治疗师 | 企业交易内的用户 | Cedars-Sinai 和 Mount Sinai 部署描述 | 将组织内渗透从医生扩到其他角色 | 没有按角色披露的公开活跃用户数 |
| 医学会和期刊 | 合作伙伴证明 | ACOG、Wiley、NEJM、JAMA、NCCN、Cochrane 关系 | 提升信任、专科深度和工作流相关性 | 不能证明直接客户付费 |
| 药企和器械公司 | 经济付款方 / 广告主 | NBC、Sacra、App Store 广告披露 | 为医生免费访问提供资金,也可能带来高毛利变现 | 赞助商集中度和依赖未披露 |
| 未来企业 / 生命科学邻近业务 | 潜在付款方 | Sacra 和 STAT 将企业扩张描述为下一阶段 | 若能签约,可在广告之外实现多元化 | 当前证明是战略叙事,不是客户证据 |
这张表用于防止把合作伙伴或广告主证明误读为客户证明。分类只反映本轮可得证据。
[CU002, CU011, CU012, CU017, CU018, CU026]展示 OpenEvidence 如何先落地个体医生,再扩展到机构工作流和变现层。
[CU010, CU017, CU018, CU049, CU050, CU051]从广泛医生触达到更小规模具名机构部署的公开可见漏斗。
前两个阶段使用公开医生占比声称,最后阶段使用公开具名机构部署数量。本图是方向性判断,不是数学上连续的队列漏斗。
[CU003, CU008, CU019, CU022, CU025, CU049]6.2 已披露客户与用户验证
公开客户证据确实存在,但并不均衡。最强的组织级证据来自具名披露的医疗系统部署。 Cedars-Sinai 披露了一个具备患者上下文的企业级落地,医生、护士、药师和治疗师可以 结合特定患者的 EHR 数据查询文献;Mount Sinai 也披露了面向医生、护士和药师的 七家医院 Epic 部署,并称这是 OpenEvidence 与医疗系统签下的第一份企业协议。二级报道还 显示,Sutter Health 早在 2026 年已把 OpenEvidence 接入 Epic 工作流,不过其公开来源链比 Cedars-Sinai 或 Mount Sinai 更薄。合在一起,这三套系统说明 OpenEvidence 可以从个人设备使用, 推进到受治理的机构工作流。 有姓名背书的个人用户证据也能看到,但质量低于企业部署证据。John Lee 医生、Ram Dandillaya 医生 和 Antonio Jorge Forte 医生的官方推荐语都描述了反复出现的临床价值,NBC 报道又加入 Paul Sax、Jeremy Cauwels 等独立医生声音。不过,证据质量边界很关键。有姓名的医生引用能证明 产品有用;不能证明账户收入、合同范围或续约。同样,ACOG 与 Wiley 的期刊和学会协议证明内容深度 和分发相关性,但那是合作伙伴证据,不是客户证据。因此,本章把医院部署和具名医生推荐语作为 主要客户证据,同时明确把合作伙伴公告与付费客户证据分开。[CU019, CU020, CU021, CU022, CU023, CU024]
| 客户 | 分层 | 部署 / 使用场景 | 生产部署 / 试点 | 结果 / 证据 | 局限 |
|---|---|---|---|---|---|
| Cedars-Sinai | 卫生系统买方 + 多学科用户基础 | 面向医生、护士、药剂师和治疗师的 EHR 内患者感知证据搜索 | 生产级企业部署 | 官方客户公告和二级报道显示,机构级上线已经运行,并用于患者上下文化场景 | 未披露合同金额、续约状态或活跃席位数 |
| Mount Sinai Health System | 卫生系统买方 + 多学科用户基础 | 横跨七家医院、面向医生、护士和药剂师的 Epic 集成证据搜索 | 生产级企业部署 | 多个二级来源描述了全系统上线,并称它是 OpenEvidence 的首个企业交易 | 经济性、部署渗透率和续约条款仍未披露 |
| Sutter Health | 卫生系统买方 + 医生用户基础 | Epic 工作流集成,用来支持即时护理决策 | 据报道为生产部署 | Healthcare IT News 和 Sacra 均在 2026 年提到该集成 | 本轮中直接来源链比 Cedars-Sinai 或 Mount Sinai 更薄 |
| Dr. Ram Dandillaya (Cedars-Sinai) | 具名医生用户 | 针对具体患者事实模式使用 OpenEvidence,并将其与 UpToDate 比较 | 反复出现的临床人员使用证言 | 官方证言提供了具名专科用户证明 | 营销证言,不是独立结果或付费账户披露 |
| Dr. Paul Sax(Brigham and Women's 医生) | 具名医生用户 | 把 OpenEvidence 当作灵活搜索工具,用于有针对性的临床问题 | 媒体报道的反复临床人员使用 | NBC 认为它在工作流中反复产生价值,并且相较 UpToDate 搜索更有利 | 不能证明付费合同状态,也不能证明更广泛的组织部署 |
这是本轮审阅中具名部署和具名临床人员引用的部分公开枚举,不是完整客户名单。
[CU019, CU020, CU021, CU022, CU023, CU024]按具体性、部署成熟度和持续性可见度比较客户证明类别。
[CU034, CU038, CU040, CU043, CU047, CU048]6.3 满意度、重复使用和留存可见度
满意度和重复使用代理指标偏正面,但并不完整。App Store 显示约 10,000 个评分下得分 4.9, AppBrain 报告 3,706 条评论下得分 4.80,并有可观的 Android 下载量。具名医生推荐语很热烈, NBC 的报道显示医生会把产品用于真实的床旁问题,而不只是出于好奇或学习。这些信号支持一个判断: OpenEvidence 对一批规模可观的临床医生具有成习惯的工具价值。但它们不能替代留存披露。公开来源 没有披露自助临床医生或企业账户的 NRR、GRR、流失、续约率、合同期限或队列行为。 记录中的反向一面同样重要,因为大规模采用可能掩盖浅层参与或质量风险。Trustpilot 的存档页面 评价严厉,1.6 分主要集中在过时 ME/CFS 指引的投诉;NBC 记录了临床医生对幻觉、过度自信的综合、 隐私护栏和受训者技能退化的担忧。NBC 还指出,一些用户只有在答案看起来异常时才点开引用,这意味着 常规依赖可能跑在系统验证前面。正确解读不是 OpenEvidence 缺乏真实客户价值;它的使用足迹太大, 无法得出这个结论。正确解读是:公开证据支持有用性和满意度代理指标,但耐久性、质量控制和长期留存 仍明显披露不足。[CU034, CU035, CU036, CU037, CU038, CU039]
| 指标 | 数值 / null | 分群 | 置信度 | 尽调请求 |
|---|---|---|---|---|
| NRR / GRR / 流失率 | 所有客户类型 | 低 | 要求提供临床医生用户、企业客户标识和广告主的季度队列 | |
| 合同期限 / 续约率 | 企业级医疗系统 | 低 | 要求按医疗系统披露合同条款、续约日期和活跃席位利用率 | |
| 日常使用渗透率 | >40% 的美国医生(公司披露) | 自助式临床医生 | 中 | 要求同一时期的 DAU、WAU、MAU 及分母定义 |
| 更广泛的活跃使用 | 2026 年 4 月约 65% / ~650k 名美国医生 | 自助式临床医生 | 中 | 要求对齐每日使用说法,并说明重复账号处理 |
| 咨询量重复使用代理指标 | 每月 8.5M(2025-07)、每月 ~20M(2026-01)、2026 年 4 月 27M 次问诊 | 临床医生工作流 | 中 | 要求拆分独立用户、重复用户和专科构成 |
| 正向 App 满意度代理指标 | App Store 4.9/5,来自 10K 个评分;AppBrain 4.80/5,来自 3,706 条评论 | 移动 App 用户 | 中 | 要求产品内 NPS、周留存和临床角色拆分 |
| 反向质量代理指标 | Trustpilot 1.6/5,来自 23 条聚焦 ME/CFS 指导的存档评论 | 投诉队列 | 中 | 要求 QA 升级率、医学内容修正和问题关闭时长 |
公开留存证据大多是代理指标。null 值是真实披露缺口,不是建模遗漏。
[CU034, CU035, CU036, CU038, CU039, CU040]6.4 扩张循环与集中度风险
OpenEvidence 的扩张逻辑可信。平台似乎先靠免费医生采用切入,再借移动端和床旁工作流加深使用, 随后进入企业系统集成,把护士、药师、治疗师和其他护理团队角色纳入访问范围。与 ACOG、Wiley、NEJM、 JAMA、NCCN、Cochrane 的内容合作扩大了专科相关性,也让产品在受治理的临床环境里更有用。如果这一模式 延续,OpenEvidence 可以从广告资助的参考工具,走向更高价值的机构合同,并在 EHR 内部取得更广的工作流 所有权。 风险在于,公开记录还没有证明这种扩张在经济上是否耐久、是否集中。Sacra 和 STAT 都描述了一门 广告支持业务,现在想建立正式的医疗系统关系;但已审阅的官方来源和新闻报道都没有披露广告主集中度、 企业收入占比、头部客户敞口或合同续约。免费采用可能让漏斗在经济上显得更强,因为它绕过了采购摩擦, 却没有证明付费意愿或长期留存。因此还有两个待承保问题:企业部署规模是否足以让模型跳出广告主依赖, 广告主基础本身是否足够分散以避开集中度风险。管理层披露队列和渠道数据之前,上行情景可以成立, 但耐久性证据仍不足。[CU012, CU013, CU014, CU015, CU017, CU018]
| 扩张驱动因素 | 集中度 / 持久性风险 | 影响 | 尽调路径 |
|---|---|---|---|
| 靠免费自助访问拿下单个医生 | 高使用量未必能转成可持续付费 | 在采购前先做出快速分发和使用深度 | 要求提供高强度临床医生使用转化为企业交易或赞助支出的证据 |
| 借临床医生牵引向医疗系统增购 | 公开企业客户名单仍小,经济性不透明 | 若属实,可显著抬高 ARPU 和切换成本 | 要求按医疗系统提供销售管线、席位数和已签企业收入 |
| 从医生扩到护士、药师和治疗师 | 角色扩张可能加深工作流依赖,但会让采用率分母更复杂 | 扩大用户基础,也更容易证明企业预算合理 | 要求按角色和工作流拆分活跃用户构成 |
| 借 ACOG、Wiley、NEJM、JAMA、NCCN 和 Cochrane 加深专科效用 | 合作伙伴标识可能被误读为付费客户验证 | 提升信任和专科覆盖,支撑更多工作流 | 尽调中把合作伙伴经济性与客户经济性拆开 |
| 广告支持模式为免费访问买单 | 头部赞助商集中度和重复投放留存未披露 | 可能利润很高,但广告主依赖会带来波动 | 要求按垂直领域提供前 10 大广告主占比、续投曲线和 CPM 趋势 |
| 自下而上采用绕开采购周期 | 绕开采购也等于绕开早期合同锁定 | 解释采用速度和影子使用 | 要求提供临床医生队列留存,以及合同落地后的企业续约 |
扩张行反映公开的商业化证据。风险行强调公开记录尚未证明经济性可持续的地方。
[CU012, CU013, CU015, CU017, CU018, CU023]6.5 展品
07风险
7.1 监管、产品责任与隐私暴露
OpenEvidence 排名最高的风险,不是 FDA 目前把它作为设备监管,而是产品路线图始终贴近非设备临床决策支持 的边界。FDA 2026 年 1 月指南仅在医疗专业人员能够独立审查推荐依据、且并非意在主要依赖该推荐时, 才保留排除适用。OpenEvidence 的条款在这一点上有帮助:条款反复把责任推回临床医生,否认诊断用途, 并把软件定位为信息支持。这是实质缓释,不是样板话。但公司同时营销高风险床旁使用,如今支持 PHI 上传, 并披露录制式 Visits 等工作流。如果未来 UX、医院部署或智能体功能削弱医生实际检查答案依据的能力, 医院风险委员会和监管者可能很快重新审视分类问题。责任风险与分类问题并行。即便 FDA 从未把 OpenEvidence 视为设备,只要幻觉或不完整输出影响诊疗,临床医生、医院和供应商仍有敞口。隐私和 HIPAA 风险同样重要, 因为 OpenEvidence 用专业身份数据变现定制内容和广告,而 HHS OCR 明确表示,向追踪或营销供应商披露 PHI 需要适当许可和 BAA。公开控制在改善,但产品从文献搜索走向患者上下文工作流后,剩余暴露仍高。[CR001, CR002, CR003, CR004, CR005, CR006]
| 风险 | 风险成因 | 可能性 | 严重性 | 缓释措施 | 残余风险敞口 | 投资逻辑失效触发点 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| FDA / CDS 分类漂移 | FDA 2026 年指南要求临床医生能够独立审查依据,且不能主要依赖软件,才保留非器械状态;OpenEvidence 则在继续深入床旁使用和 PHI 处理工作流。 | 中 | 严重 | 条款把责任推回给临床医生;引用和仅限 HCP 的门槛维持辅助定位。 | 高——未来产品 UX 或医院工作流可能缩短「辅助」和「决定性」之间的实际距离。 | 医院合同、UI 或营销开始把输出视为主要诊断建议。 | 要求正式 FDA 分类备忘录、产品风险评估和任何医院委员会审查。 |
| 医疗事故 / 专业责任风险敞口 | OpenEvidence 已进入高风险临床问题;行业证据显示,幻觉叠加过度依赖会改变诊疗决策。 | 高 | 严重 | 合同约定临床医生负责、提供来源引用,并设定内部审查预期。 | 高——如果临床医生或实习人员过度信任输出,合同免责声明消不掉下游损害。 | 公开报道的患者安全事件,或与 OpenEvidence 指导诊疗相关的保险 / 法律索赔。 | 要求不良事件日志、赔偿条款和医院事件审查流程。 |
| HIPAA / OCR 隐私与广告技术风险敞口 | 产品现在支持 PHI 工作流,但隐私模型仍包含定向广告、赞助项目和受众扩展机制;HHS 表示,PHI 营销披露需要授权和 BAAs。 | 中高 | 高 | BAAs、HIPAA 合规姿态、SOC 2 Type II、加密和合同限制。 | 高——配置错误,或 PHI、职业身份和广告测量之间边界模糊,都会代价高昂。 | 任何 OCR 调查、泄露通知,或医院禁止涉及 PHI 的广告邻近工作流。 | 审查 BAAs、追踪器清单、广告技术架构,以及按工作流划分的权限。 |
| 国际 / EU AI Act 约束 | 条款把本地法律合规责任放在用户身上;Open Pharma 报道称,OpenEvidence 已因 AI Act 不确定性撤出 EU/UK 访问。 | 高 | 高 | 等法律立场更清晰后再启动;先把产品聚焦美国临床医生。 | 中高——合规策略、本地化和数据治理姿态成熟前,地域 TAM 受限。 | 到 2027 年仍没有可信的 EU/UK 重返计划,或合规解释变得明显更严苛。 | 要求司法辖区地图、外部律师备忘录和重返时间表。 |
| 商业秘密 / 提示词窃取诉讼分心 | OpenEvidence 已在就冒充、提示词窃取和抓取起诉直接竞争对手。 | 中 | 中高 | 寻求禁令救济、收紧身份控制,并监测滥用模式。 | 中——即便诉讼有依据,也会消耗管理层注意力,并强化敌意竞争环境。 | 证据开示暴露身份控制薄弱、大规模抓取,或削弱可信度的反诉。 | 要求诉讼预算、滥用遥测,以及投诉后完成的安全变更。 |
各行按 2026-05-25 的残余风险敞口排序;覆盖范围限于公开的公司特定争议,以及直接适用的美国 / EU 监管风险。
[CR001, CR002, CR004, CR006, CR009, CR010]截至 2026 年 5 月,OpenEvidence 在法律、质量、依赖与变现几类主要风险上的残余风险分布。
[CR002, CR006, CR016, CR021, CR031, CR037]7.2 幻觉、可审计性、安全与证据质量限制
幻觉和过度依赖部分是整个临床 AI 行业的风险,但对 OpenEvidence 尤其相关,因为产品明确被用于床旁高风险问题。 Voice Mode 和 Visits 把界面从输入式搜索扩展到实时对话,便利性更高,也让语音识别错误、上下文丢失或不完整摘要 有更多路径进入工作流。外部文献带有警示意义:Patient Safety & Quality Healthcare 概述了一项 JAMA Network Open 发现,AI 生成的出院摘要约 18% 情况下不完整或有误导性;Stanford 2026 年临床 AI 综述认为, 许多达到医生水平的声称仍建立在狭窄基准和稀疏真实患者评估上。OpenEvidence 可以用引用、授权内容和 Mayo 品牌挑战结果 作为缓释信号,但这些不等于经审计的真实世界结果。更深的运营问题是证据链。GLACIS 认为,当医疗系统无法重建哪段音频、 转录、模型版本和护栏状态生成了某个输出时,责任会上升。OpenEvidence 的安全页面显示可信的控制基础——HIPAA、 SOC 2 Type II、加密和年度渗透测试——但这些是平台控制,不证明就诊级失败容易调查。因此,除非真实世界错误率、 升级规则和取证审计轨迹得到独立证明,剩余风险仍高。[CR012, CR013, CR014, CR015, CR016, CR017]
| 故障模式 | 可能性 | 严重性 | 缓释成熟度 | 残余风险敞口 | 未解决缺口 |
|---|---|---|---|---|---|
| 幻觉或不完整临床回答影响诊疗 | 高 | 严重 | 中——引用和临床医生复核有帮助,但真实世界错误率没有披露。 | 高——床旁使用意味着少见故障也可能造成后果。 | OpenEvidence 部署没有公开事件率、升级率或医疗事故数据集。 |
| 即使有可见提示,临床医生或实习人员仍过度依赖 | 高 | 高 | 中低——合同免责声明存在,但工作流压力会压过良好意图。 | 高——过度信任是人加系统的问题,不只是模型问题。 | 需要医院政策证据,说明输出在使用前多常被核查。 |
| Voice / Visits 工作流中的 PHI 泄漏或同意失败 | 中 | 高 | 中——HIPAA 合规姿态、BAAs 和加密已公开;就诊级同意操作未公开。 | 中高——录音和转录文本扩大了出错面。 | 没有关于同意收集、转录文本留存或下游处理方边界的公开审计。 |
| 事件后的取证复盘能力弱 | 中 | 高 | 低——安全计划有文档,但就诊级可追溯性没有披露。 | 高——难以复盘的故障会放大法律和信任成本。 | 医院能否调取逐次推理日志、模型版本和护栏轨迹仍未知。 |
| 基准测试成功夸大真实世界质量 | 高 | 高 | 中低——公司引用优质内容和基准测试胜出,但独立真实世界研究仍少。 | 高——这会限制医院在受治理工作流中依赖该工具的意愿。 | 需要独立结局研究,而不是挑战式或病例式评估。 |
各行结合公司特定工作流暴露面和行业临床 AI 故障模式,并按 OpenEvidence 目前在床旁使用的情况加权。
[CR012, CR013, CR014, CR015, CR016, CR017]监管、质量与商业模式风险如何层层传导,影响医院采用、收入质量和估值。
[CR006, CR014, CR021, CR031, CR033, CR041]7.3 依赖、护城河压缩与平台竞争
OpenEvidence 的护城河真实存在,但依赖很重。Wiley/Cochrane 内容和更广泛的期刊 / 指南栈不是边缘输入; 它们是“答案扎根于临床医生可验证权威来源”这一主张的核心。这形成了双刃剑结构。授权若加深并保持差异化, OpenEvidence 看起来会比通用聊天机器人更可防守。若成本上升、排他性变弱或覆盖缺口更显眼,毛利率和信任可能 同时被侵蚀。基础设施和分发也一样。OpenEvidence 披露托管在 Google Cloud Platform 和 Vercel,而更大的竞争者 正直接进入同一工作流。OpenAI 现在向医院营销支持 HIPAA 的证据检索,Anthropic 提供可用于医疗的 CMS、NPI 和 PubMed 连接器,Google 推 MedLM 和 Vertex Search,Doximity 把临床 AI 打包进一个已覆盖大多数美国医生的医生网络, Microsoft 则把 Dragon Copilot 定位为企业外壳,可在同一环境式工作流中露出 OpenEvidence、UpToDate 和 Elsevier。 这部分是行业性风险,但公司层面的后果很清楚:如果临床问题在 EHR 或占主导的环境式助手内得到回答,独立查询份额和 定价权可能比收入增长暗示得更快压缩。[CR021, CR022, CR023, CR024, CR025, CR026]
| 依赖 | 交易对手 / 群体 | 角色 | 集中度 | 失效情景 | 严重性 | 缓释措施 | 残余风险敞口 |
|---|---|---|---|---|---|---|---|
| 授权证据语料库 | Wiley / Cochrane,加上 NEJM、JAMA、NCCN 和其他内容所有者 | 权威证据层和信任锚 | 高 | 失去授权、重新定价、独占性变弱或可见覆盖缺口,都会同时降低回答质量和护城河强度。 | 严重 | 分散出版方组合;保留透明引用;谈判多年续约。 | 高——核心产品质量和品牌承诺与授权内容紧密绑定。 |
| 云与应用托管 | Google Cloud Platform 和 Vercel | 核心基础设施、存储、计算、交付 | 中 | 宕机、成本上升、安全事件或架构错配会打击可靠性和利润率。 | 高 | 使用韧性架构、安全控制和多层监控。 | 中高——平台没有纵向整合,无法快速靠自托管脱困。 |
| 广告主需求 | 药企 / 器械赞助商和赞助项目买方 | 当前主要变现引擎 | 高 | 赞助商撤退、信任反噬,或医院推动无广告使用,都会在企业收入多元化落地前更快压缩收入。 | 严重 | 增长企业和生命科学收入;执行赞助商隔离;提高透明度。 | 高——当前商业模式质量仍取决于赞助商胃口。 |
| 企业工作流入口 | Microsoft Dragon Copilot、EHR 和环境式 AI 平台 | 控制临床医生在哪里提问、看到什么内容 | 高 | 独立 OpenEvidence 使用被集成式企业助手挤出。 | 高 | 以嵌入式内容伙伴或差异化独立参考层赢下位置。 | 高——工作流所有权通常流向集成最深的平台。 |
| 通用模型医疗厂商 | OpenAI、Anthropic、Google、Doximity、UpToDate 等平台 | 快速改进的替代品与补充品组合 | 高 | 竞争对手缩小质量差距、打包访问,或用更大的资产负债表补贴定价。 | 高 | 继续扩大医生特定数据护城河,并保住独特优质内容访问权。 | 高——通用模型能为更快迭代和更广企业销售提供资金。 |
各行按依赖削弱时对信任、使用量和利润率的预期影响排序。
[CR021, CR022, CR023, CR024, CR025, CR026]关键外部依赖,将 OpenEvidence 的产品承诺与信任、工作流入口和变现连接起来。
[CR021, CR024, CR025, CR026, CR027, CR029]7.4 人员、变现与地域扩张风险
执行风险压在三项相互牵连的依赖上:Daniel Nadler、广告资助变现和以美国为中心的访问。Nadler 显然不只是普通 创始人兼 CEO;融资、护城河表达和战略定位都围绕他展开,公开报道也持续把 OpenEvidence 放在他的 Kensho 履历, 以及“医生采用会形成独特数据护城河”的主张下叙述。这在今天是资产;一旦领导层连续性断裂,就是关键人物风险。 财务上,模型有前景但脆弱。CNBC 称 OpenEvidence 年化收入超过 $100 million,并有意使用应用内视频广告加速采用; STAT 则指出,公司现在被迫证明同一个广告模型能与正式医院关系共存。Open Pharma 和 Krafty Librarian 把担忧推到 最尖锐:如果赞助方资助的研究可见度、专科定向广告或不透明来源覆盖开始看起来像偏见,而不是免费访问,医疗系统可能 比收入结构多元化更快要求无广告层。地域进一步放大问题。条款把合规负担放在美国以外用户身上,Open Pharma 报道 OpenEvidence 因 AI Act 不确定性于 2026 年 4 月撤回了欧盟和英国访问。这让国际扩张成为真实约束,而不只是理论上的 未来门槛。[CR031, CR032, CR033, CR034, CR035, CR036]
| 角色 / 职能 | 依赖或缺口 | 可能性 | 严重性 | 缓释措施 | 尽调路径 |
|---|---|---|---|---|---|
| 创始人 CEO / 战略叙事 | Daniel Nadler 仍是融资、护城河阐述和产品雄心的主要外部面孔。 | 中 | 严重 | 董事会牵头继任规划,并强化第二梯队商业和产品领导力。 | 要求继任计划、董事会材料,以及融资、产品和企业销售的授权分工图。 |
| 技术 / 产品领导力 | Zachary Ziegler 和可见的小型领导梯队必须同时管理模型质量、工作流、安全和企业需求。 | 中 | 高 | 增补资深产品安全、监管和企业实施领导者。 | 审查组织架构、关键招聘、离职情况和产品安全决策权。 |
| 商业化转型 | 业务靠自下而上医生采用起家,但现在需要医院和医疗系统采购可信度。 | 高 | 高 | 面向企业买方打包治理、ROI、集成和无广告选项。 | 要求销售管线转化数据、销售周期长度和当前医院部署指标。 |
| 法务 / 合规运营负荷 | 管理层在扩张的同时,必须消化诉讼、隐私合规、AI 治理预期和伙伴谈判。 | 中高 | 高 | 在规模化前增加法务、合规和信任安全人手。 | 要求法务预算、合规人数和季度风险审查节奏。 |
人员风险按失败会中断护城河形成或医院采用的程度排序。
[CR011, CR039, CR040, CR042, CR043, CR045]7.5 缓释、监测与投资逻辑破裂触发点
OpenEvidence 并非无人管理。公开缓释包括仅限 HCP 准入、明确合同语言说明临床医生仍负责任、针对受保护 PHI 使用的 BAA、 HIPAA 和 SOC 2 声明、来源引用,以及授权优质内容而不是依赖开放互联网的刻意策略。公司也有有意义的战略可选性: 庞大的医生用户基础、早期企业拓展,以及足够资本继续投入。这些点很重要,也解释了为什么风险画像仍可投资,而不是已经破裂。 但每项缓释都有边界。引用不能保证临床医生在时间压力下会核验。HIPAA 姿态不能在广告技术、分析或合作伙伴工作流配置错误时 消除 OCR 风险。内容合作降低幻觉风险,却带来成本和续约依赖。创始人驱动的速度可能同时放大战略优势和集中度。承保时, 最清晰的投资逻辑破裂触发点都可监测:任何证据显示 OpenEvidence 的 UX 或合同让医院把它视为事实上的诊断引擎;任何公开报道的 临床错误或隐私事件暴露弱审计能力;核心内容授权流失或重新定价;持续无法把临床医生采用转化为非广告企业收入;或者 Daniel Nadler 离任且没有准备好的接班人。正确姿态不是假定失败,而是要求证明治理、变现质量和医院接受度正在追上采用。[CR019, CR031, CR033, CR038, CR044, CR045]
| 风险 | 可监测触发因素 | 阈值 / 终止事件 | 行动含义 |
|---|---|---|---|
| FDA / CDS 分类漂移 | 产品路线图、医院合同和 UI 语言 | 有证据显示医院或监管机构将 OpenEvidence 视为主要诊断或治疗引擎,而非辅助式 CDS | 立即重估监管路径、医院推广假设和责任准备金。 |
| 医疗事故 / 幻觉事件 | 公开事件报告、保险方争议、法律投诉或医院安全备忘录 | 可信的患者伤害案例与 OpenEvidence 输出相关,且事件审计能力薄弱 | 将风险评级上调一档;追加资本前要求安全数据和赔偿细节。 |
| HIPAA / 隐私泄露或 OCR 问题 | 泄露通知、OCR 往来函件或客户安全暂停 | 任何与追踪、录音或工作流数据相关的 PHI 泄露或 OCR 调查 | 视为投资逻辑受损,因为信任是企业采用的核心。 |
| 广告主冲突 / 收入脆弱性 | 头部赞助商集中、医院无广告需求,或企业收入结构切换变慢 | 到 2027 年收入仍主要由广告驱动,且医院明显抵触赞助商邻近 | 下调收入质量倍数,并要求非广告多元化证明。 |
| 内容授权受损 | 续约条款、合作伙伴公告或独家权丧失 | 一个或多个旗舰内容伙伴流失或大幅重新定价,削弱答案质量或利润率 | 重新评估护城河耐久性和毛利率假设。 |
| 创始人 / 领导层断档 | 高管流失或治理冲突 | Daniel Nadler 离职或明显退居二线,且没有可信接班人与治理方案 | 立即做董事会尽调;接班安排清晰前,按重大投资逻辑破裂处理。 |
| 国际化 / 监管停滞 | EU/UK 准入状态和合规路线图 | 没有可信路径重返 EU/UK 市场,也没有美国广告 / 医院之外等量的替代增长引擎 | 收紧 TAM 假设,把扩张可选性视为微不足道。 |
这些阈值刻意设计成可观察,而不是停留在理论层面,因此可以作为实时尽调监控项。
[CR031, CR033, CR037, CR038, CR044, CR045]08估值
8.1 融资重估,以及当前估值已经定价了什么
OpenEvidence 的估值历史已不再只由 2025 年 2 月 Series A 定义。公开报道现在显示,它从 2025 年 2 月的 $1 billion, 快速阶梯式升至 7 月 $3.5 billion、10 月 $6 billion、2026 年 1 月 $12 billion。即便按垂直 AI 标准,这也很罕见, 并实质改变了如何判断当前入场价。公司显然有真实牵引力:管理层和媒体来源都指向超过 40% 的医生日使用率、2025 年 12 月约 18 million 次咨询、超过 100 million 名美国人被使用该产品的医生触达,以及 2025 年收入突破 $100 million。但已披露的 变现证据仍落后于估值。Sacra 估计约 $150 million 年化收入,意味着 $12 billion 估值下收入倍数约 80x;而 2025 年 7 月 估值约 $50 million 年化收入,对应约 70x。关键结论是,最新一轮不只是更新 Series A 估值;它把 OpenEvidence 重新定价到一个 投资者已经在承保未来状态的类别,而不只是承保今天已披露的经营基础。[CV001, CV002, CV003, CV004, CV005, CV006]
| 建议 | 置信度 | 风险评级 | 估值立场 | 当前价格已隐含什么 | 决策含义 |
|---|---|---|---|---|---|
| 继续研究 | 中 | 高 | 昂贵 | 持续的乐观情景需要收入扩张远超当前公开披露,且商业化不再只靠广告。 | 只有尽调能核验经审计经济性、广告主耐久性和股权结构条款,或入场价格大幅下修时,才继续跟进。 |
建议仅基于截至 2026-05-25 的公开证据;由于收入质量、企业客户结构和优先股堆叠数据并未公开,明确避免 DCF 式精确化。
[CV017, CV041, CV044, CV045, CV047]决策链从真实医生采用出发,落到昂贵的当前估值标记和“继续研究”建议。
流程图反映本章判断:公司质量和估值支撑并不同步。
[CV017, CV037, CV038, CV041, CV047]8.2 可比公司组与变现纪律
最好的公开基准不是通用 SaaS 篮子,而是一组窄口径的医生工作流、临床软件和医疗 AI 公司。Doximity 是最接近的公开变现参照, 因为它把医生网络和广告经济性结合在一起;收入倍数约 5.8x、市值约 $3.64 billion,即便披露更充分,仍远低于 OpenEvidence 的 私募估值。Tempus AI 和 Veeva 证明,公开市场会为规模化医疗 AI 或临床软件业务支付 6-8x 收入倍数,且这些公司的收入基数大得多。 Phreesia 和 Health Catalyst 等低倍数工作流公司提醒投资者,一旦增长放缓或产品差异化变弱,医疗 IT 倍数会多快压缩。Sacra 对 高价医生广告 CPM 和约 $124 ARPU 的描述解释了广告模型为何有吸引力,但也凸显核心承保问题:公开证据尚未显示广告主集中度、 续约质量,或广告收入与任何企业合同之间的结构。在已披露基本面上,当前 $12 billion 估值明显跑在公开可比区间前面,因此需要 私人市场稀缺性和超增长假设,而这些尚未被公开证据充分去风险。[CV007, CV008, CV014, CV018, CV019, CV020]
| 可比对象 | 估值 / 状态 | 收入背景 | 参照意义 | 局限 |
|---|---|---|---|---|
| OpenEvidence(当前私营估值) | $12B 私营 Series D 轮 | 估计 2025 年收入年化约 $150M;官方披露收入门槛 >$100M | 本章当前投资测算的锚点 | 收入估计未经审计,私营估值定价机制可能夸大可比性 |
| Doximity | $3.64B 上市市值(~5.8x) | TTM 收入约 $0.63B;广告占比较高的医生网络 | 最接近的上市商业化参照,因为医生受众和广告经济性重叠 | 产品更成熟,通信 / 工作流产品组合更广 |
| Tempus AI | $8.29B 上市市值(~6.5x) | TTM 收入约 $1.27B | 用于衡量数据 / 软件溢价的上市医疗 AI 基准 | 诊断和基因组业务结构不同于医生搜索 |
| Veeva Systems | $26.13B 上市市值(~8.2x) | TTM 收入约 $3.19B | 衡量规模化信任和工作流价值的最佳临床软件基准 | 面向生命科学企业的企业 SaaS,不是免费的医生产品 |
| Phreesia | $0.55B 上市市值(~1.2x) | TTM 收入约 $0.46B | 工作流软件参照,用于约束下行倍数 | 患者接待和互动在经济性上不同于临床医生 AI |
| Health Catalyst | $0.095B 上市市值(~0.3x) | TTM 收入约 $0.31B | 说明公开市场如何严厉惩罚增长较慢或差异化较弱的健康 IT | 类似转型困境的画像,不是 AI 稀缺性资产 |
| Hippocratic AI | $3.5B 私营 Series C 轮 | 收入未公开披露 | 可参照的私营医疗 AI 稀缺性可比公司 | 终端市场不同,且没有公开收入披露 |
| UpToDate / Wolters Kluwer | 未披露独立估值 | 通过席位授权知识模型可参考约 $595M 收入 | 循证临床知识工具的有用商业化参照 | 没有独立交易倍数,母公司背景削弱可比性 |
上市公司样本以市值 / 收入作为可观察的实用代理,因为此处没有来源支撑企业价值口径调整;表格选取了直接可比的医生、临床软件和医疗 AI 参照,并非穷尽。
[CV017, CV020, CV021, CV024, CV027, CV029]在选定收入倍数下,用 OpenEvidence 估计的 2025 年化收入 $150M 推算隐含估值。
数值用 Sacra 估计的 2025 年化收入 ~$150M 做简单收入倍数敏感性分析,目的是显示当前估值标记嵌入了多少乐观预期。
[CV017, CV020, CV024, CV027, CV041]8.3 乐观、基准与悲观框架
乐观情景不难想象;只是要求很高。OpenEvidence 似乎已经拥有异常强的医生触达、不同寻常的出版商合作,以及仍愿意为医疗 AI 类别 领导者付费的融资市场。如果采用率保持在美国医生 40% 以上、收入再大致翻倍,并且公司在广告之上叠加耐久的企业或工作流变现, 当前 $12 billion 估值可能被证明站得住。基准情景没那么慷慨。在基准情景下,OpenEvidence 仍重要且快速增长,但变现比使用更快 正常化,公开市场式倍数更重要,投资者也会等经审计质量出现后才为进一步上调买单。悲观情景围绕一个简单错配:广告支持的临床搜索 可以是有价值的产品,但未必值 80x 收入倍数。竞争压力、广告主犹豫,或监管对智能体医疗辅助采取更严格解读,不会抹掉产品, 却可能大幅压缩估值。因此,投资逻辑与反向逻辑必须按价格敏感来框定,而不只是按公司质量敏感来框定。[CV009, CV013, CV014, CV017, CV032, CV034]
| 维度 | 投资逻辑 | 反向逻辑 | 哪些证据会改变判断 |
|---|---|---|---|
| 医生采用 | 超过 40% 的日活医生使用率和快速上升的问诊量,说明产品确实拉动工作流。 | 使用可以很广,但未必能按软件级经济性变现,也未必能在竞争下保持耐久。 | 展示多个季度按专科拆分的留存、重复使用和参与度队列。 |
| 内容护城河 | NEJM、JAMA、NCCN、Cochrane 等合作确实带来信任和分发优势。 | 出版方关系有价值,但如果竞争对手拿到类似访问权,或临床医生多平台并用,仍不足以支撑 $12B 估值。 | 用留存或转化数据证明授权内容带来的独家价值。 |
| 商业化 | 医生端广告溢价和免费访问可以绕开医院采购摩擦,快速放大。 | 广告依赖带来集中度、伦理和周期性风险,公开披露尚未量化。 | 披露广告主集中度、续约和非广告收入结构。 |
| 可比公司支撑 | 私营医疗 AI 稀缺性和公开市场 AI 热度,解释了其相对上市健康科技可比公司的部分溢价。 | 已披露倍数仍远高于公开可比公司区间,甚至高于许多有已知基本面的 AI 医疗同行。 | 展示足够强的收入增长和利润率,把隐含倍数压缩到更可辩护的前瞻水平。 |
| 监管姿态 | 当前 FDA 指引仍给临床医生支持工具留出空间,前提是产品保持透明且不作决定性判断。 | 如果产品转向更智能体化或处方化,可能进入更重监管或治理负担。 | 提供产品边界文件、人工复核设计和治理控制。 |
| 退出路径 | 强采用和品牌力可能支撑未来 IPO 或战略交易进程,前提是披露成熟。 | 目前还没有公开证据证明公司具备 IPO 级披露、耐久的企业合同或干净的优先股堆叠。 | 分享经审计财务、董事会条款和走向上市公司准备状态的可信路径。 |
反向逻辑刻意围绕估值敏感点,而不是泛泛列风险;每行都点出最能改变判断的一项披露或经营证明。
[CV009, CV013, CV014, CV017, CV032, CV034]| 情景 | 必须成立的条件 | 估值 / 回报逻辑 | 关键风险 | 概率信号 |
|---|---|---|---|---|
| 乐观 | 医生渗透率保持在 40% 以上,收入大约再次翻倍,广告 CPM 守住,企业端商业化开始有分量。 | $10B-$15B;如果 OpenEvidence 复利成长为定义品类的医生平台,而不是单一功能的临床搜索工具,当前估值可以成立。 | 执行仍取决于商业化质量、合作伙伴耐久性,以及没有重大治理冲击。 | 需要披露的收入运行率远高于当前公开证据,并且企业端证明更清晰。 |
| 基准 | 采用保持强劲,但商业化成熟更慢,披露仍不完整,估值向优质私营公司溢价倍数收敛,而不是极端稀缺性倍数。 | $5B-$8B;结局仍然很大,但低于当前估值,更接近优质私营公司相对公开可比公司的高溢价。 | 公开可比公司的估值引力、广告主质量参差,以及企业端商业化仅部分跑通。 | 与当前公开证据最一致。 |
| 悲观 | 使用增长放缓、广告需求转弱、竞争压力上升,或围绕智能体医疗辅助的监管收紧。 | $2B-$4B;产品仍然很强,但市场开始为真实经济性而不是可选性买单,估值会大幅压缩。 | 倍数压缩、信任受损,或下轮重置。 | 由收入质量弱、续约弱,或重大监管 / 合作伙伴挫折触发。 |
这些区间是判断性估值带,锚定已披露采用、估计收入、公开可比公司倍数和私营医疗 AI 稀缺性,而不是没有支撑的 DCF 精度。
[CV017, CV041, CV042, CV044, CV045, CV046]悲观、基准和乐观估值区间锚定公开证据,而不是数据室中的私有假设。
区间是判断性分档,锚定采用证据、估计收入、公开可比公司的估值纪律和私有 AI 稀缺性;它们不是贴现现金流模型输出。
[CV044, CV045, CV046]8.4 建议、退出准备度与最终尽调问题
有证据支撑的建议是继续研究,而不是买入。OpenEvidence 看起来是一家真实公司,也有真实采用,但当前私募价格已经假设公司会把使用 转化为耐久、高质量收入,并达到公开证据尚未记录的规模。务实看,今天的估值更像是对未来企业变现、未来广告主耐久性和未来监管韧性的 预支。这仍可能跑通,但让入场纪律变得关键。除非尽调能补上最大缺口——经审计收入和毛利率、广告主集中度和留存、已签署企业集成、 股权结构表优先权悬垂,以及临床质量治理——投资者应把公司视为跟踪或稍后承保的机会。因此,退出准备度是有条件的,而非已被证明。 如果 OpenEvidence 能把采用转化为披露的经常性经济性,并维持内容和监管利益相关方的信任,它可能具备 IPO 条件;但截至 2026-05-25, 公开证据还不足以在没有非公开尽调的情况下按当前估值承保这一结果。[CV034, CV035, CV036, CV043, CV044, CV045]
| 触发器 | 可观察阈值 / 事件 | 对投资逻辑的传导 | 行动含义 |
|---|---|---|---|
| 收入质量不达标 | 经审计收入或续约显著低于管理层对 2025 年末或 2026 年的评论所暗示水平。 | 打破当前采用正在转化为高质量经济性的判断。 | 停止按当前估值承销,重置到下行情景。 |
| 医生参与停滞 | 日活医生占比或问诊增长连续多个季度停止复合增长。 | 削弱 OpenEvidence 正成为默认临床医生工作流的核心主张。 | 把公司视为小众工具,而不是品类领导者。 |
| 广告或合作伙伴反弹 | 主要广告主、出版方或医学会关系因信任或利益冲突担忧而受损。 | 同时打击商业化耐久性和内容护城河。 | 假设增长更低、治理成本更高。 |
| 监管收紧 | FDA、州级或专业治理姿态收窄智能体临床辅助空间,或要求更重控制。 | 提高合规成本,并拖慢产品向更高价值工作流扩张。 | 在产品边界和合规姿态清晰前,推迟承销判断。 |
| 估值重置 | 下一轮融资或老股参考价显著低于 $12B。 | 确认当前私营估值领先于披露和市场支撑。 | 按公开可比公司和下行情景纪律重新估值。 |
这些是为投资纪律设计的可衡量止损标准,而不是一般经营风险。
[CV036, CV038, CV043, CV048, CV049]| 主题 | 缺失证据 | 为什么重要 | 负责人 / 尽调路径 |
|---|---|---|---|
| 经审计收入和毛利率 | 经审计 FY2025 和最新收入运行率、毛利率,以及广告主调整后的贡献利润率。 | 如果不知道商业化质量是否匹配使用质量,就无法干净地承销当前估值。 | 向 CFO 和主要投资人索取经审计财务和队列收入桥。 |
| 广告主集中度和续约 | 头部客户集中度、平均合同期、定价权和续约留存。 | 广告支持模式是高溢价结果和估值重置之间最大的未证实驱动。 | 索取前 10 大广告主结构、续约队列,以及赞助商与答案隔离政策。 |
| 企业端商业化深度 | 已签约的医疗系统、支付方或 EHR 合同,以及广告之外的管线。 | 可信的非广告层,是长期守住溢价倍数最干净的路径。 | 索取已执行合同、实施时间表和企业 ARR 桥。 |
| 临床质量治理 | 模型 QA 指标、幻觉升级流程、人工复核设计和事故日志。 | 一旦估值依赖更深的工作流集成,监管和信任耐久性更重要。 | 审查治理委员会材料、质量仪表盘和红队结果。 |
| 股权结构表和流动性压力 | 优先股堆叠、清算瀑布、董事会权利和老股交易历史。 | 即便私营估值很高,如果优先权压力损害普通股或后进入投资人的回报,仍可能缺乏吸引力。 | 索取完整股权结构表、融资条款,以及近期老股成交价。 |
| 留存和专科深度 | 按专科、医院类型和队列年份拆分的留存。 | 乐观情景需要持久习惯养成,而不是一次性试用。 | 索取用户队列分析和专科渗透趋势线。 |
这些问题按最能直接改变当前价格下估值判断的事项排序。
[CV043, CV047, CV048, CV049, CV050]投委会口径评分,衡量当前估值下承销 OpenEvidence 最关键的因素。
评分是判断性的,用来维持相对承销纪律,不是机械的组合排序。
[CV037, CV038, CV041, CV043, CV047]8.5 展品
免责声明
本报告是基于公开证据的尽调快照,不构成投资建议。重要的财务、法律、技术和合同事实仍未公开;作出任何投资决定前,应直接向管理层和一手文件核验。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | OpenEvidence is currently headquartered in Miami, Florida. | 高 | SO009, SO010 |
| CO002 | OpenEvidence was founded in 2022 by Daniel Nadler and Zachary Ziegler. | 高 | SO009, SO010 |
| CO003 | OpenEvidence launched its physician-facing platform in 2023. | 中 | SO007 |
| CO004 | OpenEvidence positions itself as an AI copilot and medical knowledge platform for clinicians making point-of-care decisions. | 高 | SO001, SO002, SO003 |
| CO005 | OpenEvidence says its answers are grounded in peer-reviewed medical literature and cite underlying sources. | 高 | SO003, SO008 |
| CO006 | Core OpenEvidence is free for verified U.S. clinicians. | 高 | SO007, SO009, SO011 |
| CO007 | OpenEvidence monetizes its core product through advertising, including pharmaceutical and medical-device promotions. | 高 | SO009, SO011, SO029 |
| CO008 | OpenEvidence Ask supports quick consults, differential diagnosis, treatment options, drug information, guideline summaries, prior authorizations, and patient education. | 中 | SO003 |
| CO009 | OpenEvidence Visits records encounters, generates clinical notes, and surfaces real-time decision support on web and mobile. | 高 | SO005, SO022 |
| CO010 | OpenEvidence Voice Mode provides hands-free spoken medical Q&A and preserves a written transcript with references. | 中 | SO006 |
| CO011 | OpenEvidence DeepConsult is positioned as a PhD-level research agent for complex clinical questions. | 高 | SO004, SO008 |
| CO012 | Daniel Nadler is OpenEvidence’s founder and CEO. | 高 | SO008, SO009, SO010 |
| CO013 | Zachary Ziegler is OpenEvidence’s co-founder and is described as a Harvard AI researcher or PhD student. | 高 | SO009, SO010 |
| CO014 | Before OpenEvidence, Daniel Nadler founded Kensho in 2013. | 高 | SO009, SO030 |
| CO015 | S&P Global agreed to acquire Kensho for approximately $550 million net of cash in 2018. | 中 | SO030 |
| CO016 | OpenEvidence’s about page lists a large medical-advisor network spanning institutions including Mayo Clinic, Cleveland Clinic, Mount Sinai, Duke, and HCA Healthcare. | 中 | SO002 |
| CO017 | OpenEvidence’s about page lists investors including Sequoia, Kleiner Perkins, Google Ventures, Nvidia, Thrive, Andreessen Horowitz, DST, Blackstone, and Mayo Clinic. | 中 | SO002 |
| CO018 | OpenEvidence closed a $75 million Series A led by Sequoia at a $1 billion valuation in February 2025. | 高 | SO007, SO009 |
| CO019 | The February 2025 Series A brought OpenEvidence’s total capital raised to over $100 million. | 中 | SO007 |
| CO020 | OpenEvidence announced a $210 million Series B at a $3.5 billion valuation in July 2025. | 中 | SO008 |
| CO021 | Google Ventures and Kleiner Perkins co-led the Series B, with Sequoia following on and Coatue, Conviction, and Thrive also investing. | 中 | SO008 |
| CO022 | OpenEvidence said the Series B left the company with more than $300 million raised since founding. | 中 | SO008 |
| CO023 | Fierce Healthcare reported OpenEvidence raised a $200 million Series C at a $6 billion valuation in October 2025. | 中 | SO013 |
| CO024 | Fierce Healthcare reported Google Ventures led the Series C and that Sequoia, Kleiner Perkins, Thrive, Coatue, BOND, Blackstone, and Craft also participated. | 中 | SO013 |
| CO025 | OpenEvidence announced a $250 million Series D at a $12 billion valuation in January 2026. | 高 | SO009, SO016, SO014 |
| CO026 | The Series D was co-led by Thrive Capital and DST Global. | 高 | SO009, SO016, SO014 |
| CO027 | Company press said the Series D brought OpenEvidence to roughly $700 million raised over the preceding 12 months. | 高 | SO016, SO014 |
| CO028 | CNBC’s May 2026 Disruptor profile listed OpenEvidence funding at $795.4 million. | 中 | SO010 |
| CO029 | Company and press sources consistently place OpenEvidence’s daily U.S. physician usage above 40% across more than 10,000 hospitals and medical centers. | 高 | SO008, SO009, SO012 |
| CO030 | OpenEvidence said it supported about 18 million U.S. clinical consultations in December 2025, up from about 3 million per month one year earlier. | 高 | SO016, SO012 |
| CO031 | OpenEvidence said more than 100 million Americans were treated in 2025 by a doctor using the platform. | 高 | SO016, SO008 |
| CO032 | OpenEvidence said it crossed one million clinical consultations in a single day on March 10, 2026. | 中 | SO017 |
| CO033 | CNBC reported OpenEvidence topped $100 million in annualized revenue in 2025. | 中 | SO009 |
| CO034 | CNBC reported Nadler said 95% of new OpenEvidence users hear about the product from another physician. | 中 | SO009 |
| CO035 | NBC News reported company claims that about 65% of U.S. doctors used OpenEvidence across almost 27 million clinical encounters in April 2026. | 中 | SO011 |
| CO036 | NBC News reported that MaineHealth asks its doctors not to enter protected health information into OpenEvidence despite the company’s HIPAA claims. | 中 | SO011 |
| CO037 | The NEJM Group agreement gives OpenEvidence access to content and multimedia from 1990 forward across NEJM, NEJM Evidence, NEJM AI, NEJM Catalyst, and NEJM Journal Watch. | 高 | SO007, SO015 |
| CO038 | The JAMA agreement covers full text and multimedia from JAMA, JAMA Network Open, and 11 JAMA specialty journals. | 中 | SO018 |
| CO039 | The Wiley partnership adds more than 400 journals and books plus the Cochrane Database of Systematic Reviews and Cochrane Clinical Answers to OpenEvidence. | 高 | SO019, SO020 |
| CO040 | OpenEvidence announced HIPAA compliance in April 2025 and its about page says the platform is SOC 2 Type II certified. | 高 | SO021, SO002 |
| CO041 | OpenEvidence launched Visits in August 2025 as real-time medical intelligence for the patient visit. | 高 | SO022, SO005 |
| CO042 | OpenEvidence announced a February 2026 collaboration with Sutter Health to bring evidence-based AI-powered insights into physician workflows. | 中 | SO031 |
| CO043 | OpenEvidence announced a May 2026 partnership with Cedars-Sinai to create patient-aware clinical intelligence with agentic clinical AI. | 中 | SO032 |
| CO044 | OpenEvidence announced an April 2026 NCCN collaboration to integrate canonical oncology treatment algorithms at the point of care. | 中 | SO023 |
| CO045 | OpenEvidence publicized an April 2025 Mayo Clinic study claiming comparable performance to physician clinical decision-making in common clinical scenarios. | 中 | SO024 |
| CO046 | NBC News reported clinician and researcher concerns about hallucinations, incomplete answers, limited patient-outcome studies, and erosion of clinician critical thinking from heavy OpenEvidence use. | 中 | SO011 |
| CO047 | A 2026 medRxiv preprint found pharmaceutical ads shifted advertised-drug selection by 12.7 percentage points across tested large-language-model scenarios. | 中 | SO028 |
| CO048 | Bloomberg Law reported OpenEvidence alleged in June 2025 that Doximity executives impersonated physicians and used prompt hacking to extract trade secrets. | 中 | SO025 |
| CO049 | CourtListener shows OpenEvidence filed a complaint against Doximity, Jey Balachandran, and Jake Konoske on June 20, 2025 in District of Massachusetts case 1:25-cv-11802. | 高 | SO025, SO026 |
| CO050 | HealthExec reported Doximity counterclaimed that OpenEvidence spread misinformation and tried to poach employees. | 中 | SO027 |
| CO051 | Gizmodo framed OpenEvidence as a free, ad-supported AI chatbot for doctors and amplified concern that ad-funded medical AI could shape decision-making. | 低 | SO029 |
| CO052 | OpenEvidence’s about page says the platform has supported more than 200 million AI-powered clinical consultations to date. | 中 | SO002 |
| CM001 | AHRQ defines clinical decision support as timely information, usually at the point of care, that helps inform decisions about a patient’s care. | 高 | SM001, SM002 |
| CM002 | AHRQ says CDS includes recommendations, databases, reminders, and alerts, and can lower costs, improve efficiency, and reduce patient inconvenience. | 中 | SM001 |
| CM003 | FDA guidance says certain CDS software functions are excluded from the device definition under the Cures Act non-device CDS criteria, while device software functions intended for patients or caregivers remain subject to digital-health policy. | 中 | SM003 |
| CM004 | The most relevant market boundary for OpenEvidence is clinician-facing point-of-care decision support and evidence retrieval rather than every health-AI workflow or every regulated software category. | 高 | SM001, SM003 |
| CM005 | Clinicians often answer questions through online databases, free internet search, and consultation with colleagues before adopting a new specialized tool. | 中 | SM028 |
| CM006 | More than 3 million clinicians and other healthcare professionals in more than 190 countries use UpToDate, and Wolters Kluwer says it integrates into mobile devices, EHRs, and systems. | 中 | SM018 |
| CM007 | EBSCO says DynaMed delivers daily evidence updates, one-click links to primary literature, EHR integration, and a mobile app for time-crunched clinicians. | 中 | SM019 |
| CM008 | AI-native adjacents such as Glass Health and Pathway/Doximity Ask are now part of the substitute set alongside legacy reference tools, widening the competitive boundary around point-of-care medical AI. | 中 | SM025, SM026 |
| CM009 | The Business Research Company estimates the global clinical decision support systems market will reach $3.96 billion in 2026. | 中 | SM011 |
| CM010 | Fortune Business Insights estimates the global clinical decision support systems market at $4.45 billion in 2026 and says North America held 35.77% of the 2025 market. | 中 | SM012 |
| CM011 | Fortune Business Insights’ 2026 global CDSS estimate is about $0.49 billion above The Business Research Company’s 2026 estimate, showing that top-down TAM depends materially on publisher taxonomy. | 中 | SM011, SM012 |
| CM012 | Grand View Research estimates the narrower artificial-intelligence-in-clinical-decision-support subsegment at $1.5 billion in 2026, with 17.1% CAGR to $4.5 billion by 2033. | 中 | SM013 |
| CM013 | Research and Markets structures CDSS as a multi-layered market segmented by component, model, delivery mode, application, and end user across 2020-2025 history and 2025-2030/2035 forecasts. | 低 | SM014 |
| CM014 | Fortune Business Insights says knowledge-based CDSS still held an 88.5% share, while non-knowledge-based CDSS is growing faster at 17.57% CAGR. | 中 | SM012 |
| CM015 | Cloud-based delivery and outpatient care are faster-growing CDSS segments than standalone deployments, while Grand View also says web/cloud AI-CDSS held an 82.9% revenue share in 2025. | 中 | SM012, SM013 |
| CM016 | U.S. physicians and surgeons held about 839,000 jobs in 2024. | 中 | SM015 |
| CM017 | U.S. physician assistants held about 162,700 jobs in 2024. | 中 | SM016 |
| CM018 | U.S. nurse practitioners held about 320,400 jobs in 2024. | 中 | SM017 |
| CM019 | The combined U.S. physician, physician-assistant, and nurse-practitioner seat base is about 1.322 million clinicians. | 中 | SM015, SM016, SM017 |
| CM020 | Epocrates publicly prices epocrates+ at $24.99 per month or $179.99 per year and offers online group purchasing for up to 30 licenses. | 中 | SM021, SM022 |
| CM021 | AMBOSS publicly lists clinician pricing at $29.99 per month or $259 per year, while its clinician site advertises a free trial and plans starting at $12.50 per month. | 中 | SM023, SM024 |
| CM022 | Applying public annual prices of roughly $180 to $259 to the 1.322 million U.S. physician, PA, and NP seat base yields an evidence-constrained advanced-clinician spend proxy of about $238 million to $342 million per year. | 中 | SM015, SM016, SM017, SM021, SM024 |
| CM023 | Applying the same public price band to physicians only yields a physician-first lower-bound subsegment of about $151 million to $217 million per year. | 中 | SM015, SM021, SM024 |
| CM024 | Bottom-up public-price ranges understate total institutional spend because enterprise implementation, validation, training, and integration services are usually quote-based rather than public. | 中 | SM007, SM008, SM018, SM020 |
| CM025 | Most physicians, PAs, and NPs practice in offices, hospitals, and clinics, so buyer logic is centered on provider organizations rather than consumers. | 中 | SM015, SM016, SM017 |
| CM026 | Public self-pay and small-group pricing from Epocrates and AMBOSS shows that direct and departmental purchase paths remain viable alongside enterprise sales. | 中 | SM021, SM022, SM023, SM024 |
| CM027 | UpToDate and DynaMed both compete on fast point-of-care answers, EHR integration, mobile access, and literature-backed recommendations rather than on model novelty alone. | 中 | SM018, SM019 |
| CM028 | Eighty-five percent of physicians want to be consulted or directly involved in decisions about adopting AI technologies. | 高 | SM004, SM005 |
| CM029 | HealthLeaders says physician AI adoption is happening faster than governance frameworks are evolving, forcing health systems to move from experimentation to operational oversight. | 中 | SM006 |
| CM030 | HIMSS says responsible deployment requires structured governance, continuous monitoring, and revalidation of AI tools in the field. | 中 | SM007, SM008 |
| CM031 | HIMSS26 coverage says scaled deployment depends on measurable ROI, transparency, and the ability for clinicians to trace and verify sources of key insights. | 中 | SM009, SM010 |
| CM032 | More than 80% of physicians now use AI professionally, and average AI use cases have risen to 2.3. | 高 | SM004, SM005 |
| CM033 | Thirty-nine percent of physicians use AI to summarize medical research and standards of care. | 高 | SM004, SM005 |
| CM034 | Seventy percent of physicians see AI as a tool to automate tasks that contribute to burnout, and 76% say AI can help with patient care. | 高 | SM004, SM005 |
| CM035 | Eighty-eight percent of physicians report at least some concern about AI-related skill loss. | 高 | SM004, SM005 |
| CM036 | Physicians cite data privacy protections (86%) and robust safety and efficacy validation (88%) as critical for broader AI adoption. | 高 | SM004, SM005 |
| CM037 | Clear liability frameworks rank highest among the regulatory actions physicians say would increase trust in and adoption of AI tools. | 高 | SM004, SM005 |
| CM038 | Primary care physicians used the EHR on a median 39% of PTO days, and 39.5% of PTO EHR time was spent on inbox-related tasks. | 中 | SM029 |
| CM039 | Clinical questions are frequent, time is the biggest barrier to following up on them, and more than half of point-of-care questions may never be pursued. | 中 | SM028 |
| CM040 | Alert fatigue reduces CDS responsiveness: physicians in the highest recent-alert quartile were far less likely to respond to a new alert than those in the lowest quartile (adjusted OR 0.38). | 中 | SM027 |
| CM041 | OpenEvidence positions itself as a clinician-facing medical knowledge platform with official content partnerships spanning NEJM, JAMA, NCCN, Cochrane, and Wiley. | 中 | SM030 |
| CM042 | Incumbents are adding AI onto established reference workflows rather than abandoning them: EBSCO launched Dyna AI Mode within DynaMed and DynaMedex with evidence visibility and clinician control. | 中 | SM019, SM020 |
| CM043 | Because OpenEvidence explicitly markets to physicians as a medical-knowledge platform, a physician-first lower-bound spend lens is more decision-useful than an all-clinician or all-CDSS headline TAM when framing SOM. | 中 | SM015, SM030 |
| CP001 | OpenEvidence presents itself as an official medical knowledge platform with AI partnerships spanning NEJM, JAMA, NCCN, and Cochrane content. | 高 | SP001, SP002 |
| CP002 | OpenEvidence says more than 100 million Americans were treated by a clinician using the product this year and that the platform has supported over 200 million AI-powered clinical consultations. | 中 | SP002 |
| CP003 | OpenEvidence says the product is free to doctors and ad-supported. | 中 | SP004 |
| CP004 | OpenEvidence says it is HIPAA compliant and SOC 2 Type II certified. | 高 | SP002, SP003 |
| CP005 | OpenEvidence announced a January 2026 Series D that valued the company at $12 billion. | 中 | SP004 |
| CP006 | OpenEvidence says total capital raised reached roughly $700 million over the prior 12 months. | 中 | SP004 |
| CP007 | OpenEvidence says it reached $100 million in annual revenue in less than a year after building out its commercial team. | 中 | SP004 |
| CP008 | Glass Health publicly positions itself as an ambient scribing and clinical decision support product. | 中 | SP005 |
| CP009 | Pathway's homepage says the product served hundreds of thousands of clinicians and that its capabilities now live in Doximity Ask after Pathway joined Doximity in 2025. | 中 | SP006 |
| CP010 | Doximity says its Clinical AI Suite bundles Ask, Scribe, and Dialer on a platform where more than 85% of U.S. physicians already practice. | 中 | SP007 |
| CP011 | Doximity Ask is free for verified U.S. clinicians listed in the FAQ, including physicians, nurse practitioners, physician assistants, pharmacists, podiatrists, CRNAs, and medical students. | 中 | SP008 |
| CP012 | Doximity Ask says it is HIPAA compliant and can answer referenced, evidence-based clinical questions while also generating chart notes, patient materials, and translations. | 中 | SP008 |
| CP013 | Doximity acquired Pathway for $26 million in cash and up to $37 million in additional equity grants. | 高 | SP009, SP010, SP011 |
| CP014 | Pathway had hundreds of thousands of registered users and thousands paying $300 per year for its premium product before being folded into Doximity Ask. | 高 | SP009, SP010 |
| CP015 | Fierce Healthcare reported that Doximity was beta-testing the integrated Pathway product with thousands of physician users and that analysts expected enterprise-grade monetization over time. | 中 | SP010 |
| CP016 | Wolters Kluwer says UpToDate offers individual, group, and enterprise solutions and is trusted by more than 3 million health professionals worldwide. | 中 | SP012 |
| CP017 | Wolters Kluwer's subscription help page says UpToDate has annual, 30-day recurring, 90-day recurring, and group options, with country-specific pricing surfaced through its store. | 中 | SP013 |
| CP018 | Wolters Kluwer says UpToDate Expert AI provides context-rich answers with single-click access to assumptions, source content, and step-by-step rationale and is being deployed through enterprise workflows. | 高 | SP012, SP014 |
| CP019 | Wolters Kluwer says UpToDate Expert AI's clinical intelligence is driven by the expertise of 7,600 medical experts. | 高 | SP012, SP014 |
| CP020 | Wolters Kluwer reported 2024 revenue of €5.9 billion and approximately 21,600 employees worldwide. | 中 | SP014 |
| CP021 | EBSCO says DynaMedex combines DynaMed disease evidence and Micromedex drug evidence and was recognized as 2026 Best in KLAS for point-of-care disease reference. | 中 | SP015 |
| CP022 | EBSCO says Dyna AI for DynaMedex is integrating expanded Micromedex dosing and medication safety content into the platform. | 中 | SP015 |
| CP023 | Epocrates+ says it includes disease content, lab information, herbs and supplements, ICD-10, interaction checking, and more than 150 guideline synopses. | 中 | SP016 |
| CP024 | Epocrates+ lists pricing of $24.99 per month or $179.99 per year and says medical students can receive complimentary access. | 中 | SP016 |
| CP025 | Micromedex says it offers 2,500+ drug monographs, 700+ clinical calculators, RED BOOK supplier and pricing data, and is trusted in over 80 countries. | 中 | SP017 |
| CP026 | ClinicalKey says ClinicalKey AI delivers responsible-AI answers sourced from trusted evidence-based content and is designed for EHR integration and continuing education workflows. | 中 | SP018 |
| CP027 | OpenAI for Healthcare says ChatGPT for Healthcare provides transparent citations to peer-reviewed studies, public-health guidance, and clinical guidelines, along with BAAs, audit logs, data residency options, customer-managed encryption keys, and no training on customer content. | 高 | SP019, SP021 |
| CP028 | OpenAI's published API pricing starts at $5 input and $30 output per 1 million tokens for GPT-5.5, with lower-cost tiers also listed. | 中 | SP020 |
| CP029 | OpenAI says thousands of organizations have already configured its API for HIPAA-compliant use and names major hospital partners rolling out ChatGPT for Healthcare. | 中 | SP019 |
| CP030 | Anthropic says Claude for Healthcare is HIPAA-ready and can connect to CMS coverage data, ICD-10, NPI, PubMed, and FHIR-oriented workflows. | 高 | SP022, SP023 |
| CP031 | Anthropic says its healthcare and life-sciences capabilities are available across Claude subscription tiers and that users' health data is not used to train models. | 中 | SP023 |
| CP032 | Anthropic's pricing page lists Team seats at $20 per month billed annually or $25 billed monthly and describes a HIPAA-ready enterprise offering alongside API pricing such as Opus 4.7 at $5 input and $25 output per MTok. | 中 | SP024 |
| CP033 | FDA's January 2026 CDS guidance says software remains outside device regulation only when it supports rather than replaces clinician judgment and when clinicians can independently review the basis for recommendations. | 中 | SP025 |
| CP034 | Frontiers argued in 2026 that LLM hallucinations can produce plausible but incorrect medical information, creating patient-safety, trust, and responsibility concerns. | 中 | SP026 |
| CP035 | A 2025 MDPI study evaluating 12 RAG variants on 250 de-identified clinical vignettes reported that self-reflective RAG lowered hallucinations to 5.8% but still emphasized secure on-premises deployment, provenance tagging, and audit trails. | 中 | SP027 |
| CP036 | OpenEvidence's moat appears to rely more on licensed medical content, clinician mindshare, and published trust controls than on paid-seat or enterprise lock-in. | 中 | SP001, SP002, SP003, SP004 |
| CP037 | Doximity's moat is primarily distribution and workflow bundling because it can put clinical reference inside a physician network that already reaches more than 80%-85% of U.S. physicians. | 中 | SP007, SP009, SP010 |
| CP038 | UpToDate, DynaMedex, ClinicalKey, and Micromedex retain incumbent power through curated editorial depth, enterprise procurement, and existing workflow embedding rather than through public price transparency. | 中 | SP012, SP014, SP015, SP017, SP018 |
| CP039 | Pricing transparency is highest for OpenEvidence, Doximity Ask, Epocrates, OpenAI, and Anthropic, while incumbent enterprise suites mostly channel buyers toward product bundles, country-specific stores, or negotiated contracts. | 中 | SP004, SP008, SP013, SP016, SP020, SP024 |
| CP040 | Switching costs are highest where buyers already depend on curated enterprise references and EHR-linked workflows, and lower where clinicians can multi-home across free AI assistants. | 中 | SP008, SP012, SP014, SP015, SP018, SP019, SP023 |
| CP041 | Doximity Ask's own FAQ warns that AI outputs may occasionally include inaccuracies or hallucinations and should always be verified by clinicians. | 中 | SP008 |
| CP042 | General-model entrants reduce internal-build friction because OpenAI and Anthropic both pair healthcare controls with published API or seat pricing and workflow tooling. | 中 | SP019, SP020, SP023, SP024 |
| CP043 | Pathway's move from a $300-per-year premium app into free Doximity Ask shows that standalone clinical reference products can be commoditized by a larger workflow platform. | 中 | SP006, SP009, SP010 |
| CP044 | OpenEvidence and UpToDate Expert AI now compete on citation-backed reasoning, but UpToDate packages that capability inside an entrenched enterprise product while OpenEvidence competes with a free ad-supported destination. | 中 | SP004, SP012, SP014 |
| CP045 | Drug-heavy substitutes such as Micromedex and Epocrates remain relevant because medication safety, formularies, and pricing workflows are narrower and more operationally embedded than generic clinical Q&A. | 中 | SP016, SP017 |
| CP046 | Regulatory and buyer defensibility increasingly favor AI products that expose sources, rationale, and auditability over opaque black-box answers. | 中 | SP014, SP019, SP022, SP025, SP027 |
| CI001 | OpenEvidence is publicly described as free to doctors and ad-supported rather than subscription-priced for the core clinician product. | 中 | SI008, SI011, SI013 |
| CI002 | Company-linked January 2026 coverage says OpenEvidence topped $100 million in annual revenue during 2025. | 中 | SI008, SI011, SI013 |
| CI003 | Sacra estimates OpenEvidence reached $150 million of annualized revenue in 2025, up from $7.9 million in 2024. | 中 | SI016 |
| CI004 | Sacra estimates gross margin at roughly 90% in 2025. | 低 | SI016 |
| CI005 | Sacra estimates OpenEvidence can monetize physician attention at roughly $70-$1,000+ CPM and about $124 of ARPU. | 中 | SI016, SI017 |
| CI006 | Sacra argues the free product and limited initial dependence on patient-record integration helped OpenEvidence bypass the roughly 18-month hospital procurement cycle. | 中 | SI017 |
| CI007 | OpenEvidence said it supported about 18 million clinical consultations in December 2025 versus roughly 3 million per month a year earlier. | 中 | SI008, SI012, SI013 |
| CI008 | Public company and news materials say the platform is used daily by more than 40% of U.S. physicians and touches more than 10,000 hospitals and medical centers. | 中 | SI008, SI011, SI013, SI024 |
| CI009 | Mid-2025 funding coverage said OpenEvidence was adding more than 65,000 new verified U.S. clinicians per month and handling more than 8.5 million monthly consultations. | 中 | SI009, SI024, SI025 |
| CI010 | Mount Sinai announced the first enterprise-scale OpenEvidence deployment across physicians, nurses, and pharmacists within Epic in March 2026. | 中 | SI007 |
| CI011 | The Veeva partnership positions OpenEvidence to monetize life-sciences use cases such as clinical-trial matching, drug discovery insight, and medicine adoption support. | 中 | SI010 |
| CI012 | Sacra characterizes OpenEvidence as a B2B2C freemium model that starts with free clinician access and can later expand into enterprise subscriptions. | 中 | SI016 |
| CI013 | Each DeepConsult run requires more than 100 times the compute and cost of a standard OpenEvidence search even though the feature remains free for verified clinicians. | 中 | SI009, SI024, SI025 |
| CI014 | Management said recent financing would be used mainly for R&D, model training, compute, and expanded content partnerships. | 中 | SI008, SI013, SI018 |
| CI015 | OpenEvidence's content moat depends on licensed relationships with NEJM, JAMA, NCCN, Wiley/Cochrane, and other trusted medical sources. | 中 | SI001, SI002, SI021 |
| CI016 | The Visits workflow says encounter data handling is configurable and that OpenEvidence does not train AI models on protected health information. | 中 | SI004 |
| CI017 | The user guide shows 2026 releases in coding, note styling, messaging, and faxing, indicating expansion from search into broader clinical workflow tools. | 中 | SI003 |
| CI018 | No retained public source discloses pricing for Mount Sinai, Sutter, or a future non-ad enterprise tier. | 低 | SI006, SI007, SI015, SI016 |
| CI019 | OpenEvidence's July 2025 Series B raised $210 million at a $3.5 billion valuation. | 中 | SI009, SI024, SI025 |
| CI020 | OpenEvidence's October 2025 Series C raised about $200 million at roughly a $6 billion valuation. | 中 | SI012, SI016 |
| CI021 | OpenEvidence's January 2026 Series D raised $250 million at a $12 billion valuation. | 中 | SI008, SI011, SI012, SI013, SI018 |
| CI022 | Official and independent January 2026 coverage put cumulative disclosed funding at roughly $700 million. | 中 | SI008, SI011, SI013, SI016 |
| CI023 | Cooley framed the Series D as growth financing for research and compute rather than as a refinancing driven by disclosed balance-sheet stress. | 中 | SI018 |
| CI024 | SEC EDGAR search returns one recent OpenEvidence-related Form D result: HII OpenEvidence-01, a pooled investment fund filed in April 2026. | 中 | SI019, SI020 |
| CI025 | The Form D reports about $5.761 million sold and appears to document a pooled investment vehicle rather than a direct operating-company financing by OpenEvidence itself. | 中 | SI019, SI020 |
| CI026 | CNBC quoted management saying OpenEvidence is trying to balance growth with eventual profitability and is not planning to burn billions over the next year, but no quantified burn target was disclosed. | 中 | SI011 |
| CI027 | Sacra says a separate non-ad-supported enterprise version is in development for large systems requiring greater customization. | 中 | SI016 |
| CI028 | STAT reports OpenEvidence is seeking more formal health-system relationships while facing questions about whether the ad-based model can keep driving growth. | 中 | SI015 |
| CI029 | MedCity says the platform monetizes through pharma ads and notes privacy-policy-based targeting using user engagement, topic, and device data. | 中 | SI014 |
| CI030 | Open Pharma raises concern that AI-generated summaries may change which medical evidence becomes discoverable, credible, and influential. | 中 | SI022 |
| CI031 | Krafty Librarian criticizes OpenEvidence for opaque article selection and ranking and for limiting outside evaluation by information professionals. | 中 | SI023 |
| CI032 | Open Pharma says OpenEvidence withdrew access for EU and UK domains on 27 April 2026 because of uncertainty under the EU AI Act. | 中 | SI022 |
| CI033 | Public use-of-funds disclosures imply an opex-heavy model centered on compute, R&D, licensing, and commercial expansion rather than on hard capex or project finance. | 中 | SI008, SI013, SI018 |
| CI034 | The strongest public traction and revenue figures remain company-claimed rather than audited in public financial statements. | 中 | SI008, SI011, SI013, SI016 |
| CI035 | No retained public source discloses OpenEvidence's cash on hand, monthly burn, or runway. | 低 | SI011, SI013, SI018 |
| CI036 | No retained public source breaks revenue into advertising, enterprise contracts, or life-sciences revenue or discloses advertiser concentration and renewal rates. | 低 | SI014, SI016, SI017 |
| CI037 | No retained public source discloses debt facilities or project-finance obligations, and the only filing evidence found relates to a secondary-market pooled investment vehicle. | 低 | SI019, SI020 |
| CI038 | Public underwriting remains blocked by missing cash and burn disclosure, missing enterprise pricing, unknown revenue mix, and the absence of audited public statements. | 中 | SI014, SI015, SI016, SI018 |
| CI039 | R&D World reports Wiley licensed Cochrane and more than 400 journals and books to OpenEvidence for point-of-care physician use. | 中 | SI021 |
| CI040 | OpenEvidence's about page says the company has supported more than 200 million AI-powered clinical consultations to date and is HIPAA compliant and SOC 2 Type II certified. | 中 | SI002 |
| CE001 | OpenEvidence publicly brands itself as an official medical knowledge platform with major medical content partners. | 中 | SE001 |
| CE002 | The OpenEvidence user guide exposes a broad clinician workflow surface that spans consult, research, documentation, and communications functions. | 中 | SE002 |
| CE003 | Voice Mode is a hands-free interface on web and mobile that reads evidence-backed answers aloud. | 中 | SE003 |
| CE004 | Deep Consult is positioned as an advanced AI agent that generates long-form research reports for complex questions. | 中 | SE004 |
| CE005 | OpenEvidence launched native iOS and Android apps and said they provide the same cited answers as the web product. | 中 | SE005 |
| CE006 | Google Play describes OpenEvidence as point-of-care clinical decision support for verified healthcare professionals with answers sourced, cited, and grounded in peer-reviewed medical literature. | 中 | SE006 |
| CE007 | The NEJM agreement gives OpenEvidence access to NEJM family content and multimedia from 1990 forward to inform answers on the platform. | 中 | SE007 |
| CE008 | The JAMA agreement gives OpenEvidence access to full text and multimedia from JAMA, JAMA Network Open, and 11 specialty journals. | 中 | SE008, SE009 |
| CE009 | The NCCN collaboration adds NCCN Guidelines and JNCCN content to OpenEvidence via natural-language oncology search. | 中 | SE010, SE011 |
| CE010 | NCCN says content surfaced on OpenEvidence follows NCCN QA review and links back to source documents for additional information. | 中 | SE010 |
| CE011 | Wiley licenses Cochrane systematic reviews, Cochrane Clinical Answers, and more than 400 Wiley journals and books into OpenEvidence. | 中 | SE012 |
| CE012 | Google Play says OpenEvidence now sources from more than 300 medical journals as well as the FDA and CDC. | 中 | SE006 |
| CE013 | Wiley describes OpenEvidence’s operating principle as specialized models trained on peer-reviewed literature rather than the open internet, with answers grounded in drillable sources. | 中 | SE012 |
| CE014 | Mount Sinai’s March 2026 rollout made OpenEvidence available directly within the electronic health record across physicians, registered nurses, and pharmacists. | 中 | SE015, SE019 |
| CE015 | Mount Sinai’s Epic workflow lets clinicians ask medical questions in natural language and receive answers grounded in peer-reviewed literature and clinical guidelines. | 中 | SE015, SE019 |
| CE016 | Cedars-Sinai’s 2026 partnership integrates patient context from Epic directly within OpenEvidence. | 中 | SE016, SE020 |
| CE017 | Cedars says patient-aware answers can account for prior procedures, comorbidities, medications, allergies, and longitudinal health data. | 中 | SE016 |
| CE018 | DotFlows lets clinicians create or browse reusable natural-language prompts that customize OpenEvidence responses. | 中 | SE017, SE002 |
| CE019 | OpenEvidence’s public 2026 workflow extension includes automatic CPT suggestions, E/M level recommendations with MDM rationale, and ICD-10 codes generated when a visit is complete. | 中 | SE002, SE020 |
| CE020 | The 2026 user guide surfaces Voices, Messages, and Fax features, including patient texting with consent controls and send-receive fax inside the app. | 中 | SE002 |
| CE021 | Fierce Healthcare says OpenEvidence fields more than 1 million clinical questions per day and has expanded from search into visit transcription, doctor dialer, and coding workflows. | 中 | SE020 |
| CE022 | PRNewswire said OpenEvidence reached one million physician-AI clinical consultations in a single 24-hour period on March 10, 2026. | 中 | SE021 |
| CE023 | OpenEvidence announced HIPAA compliance and secure PHI upload in April 2025. | 中 | SE014 |
| CE024 | OpenEvidence says covered entities that input PHI do so under a Business Associate Agreement. | 中 | SE014 |
| CE025 | OpenEvidence says conversations are private by default and users control access through Share controls. | 中 | SE014 |
| CE026 | Voice Mode keeps a written transcript with references alongside spoken answers, extending the same verification model into audio interaction. | 中 | SE003, SE020 |
| CE027 | OpenEvidence restricts product access to verified healthcare professionals rather than the general public. | 中 | SE006 |
| CE028 | OpenEvidence’s public hiring footprint still looks concentrated: a May 2026 engineering job described a 30-person engineering team with in-person roles in San Francisco or Miami, while Built In lists Cambridge as HQ. | 中 | SE022, SE023 |
| CE029 | Medical Economics reported 40,000 new verified doctors register each month and 75% of physicians use OpenEvidence at the point of care. | 中 | SE018 |
| CE030 | Medical Economics said OpenEvidence planned to expand its AI scientist team and develop medical-domain-specific large language models. | 中 | SE018 |
| CE031 | Publicly surfaced evidence quality is strongest on licensed-content breadth and workflow integration, not on disclosed infrastructure architecture or benchmarked latency. | 低 | SE012, SE015, SE020 |
| CE032 | OpenEvidence’s coverage depth depends materially on continuing publisher and society licenses because each new agreement explicitly expands what content can inform answers. | 中 | SE007, SE010, SE012 |
| CE033 | Post-hoc explainability tools such as SHAP and LIME often fail to provide causal explanations for hallucinations, and even RAG systems remain vulnerable when retrieval fails or retrieved content is misinterpreted. | 中 | SE025 |
| CE034 | In a 2025 medical-LLM study, 91.8% of surveyed clinicians had encountered medical hallucinations and 84.7% considered them capable of causing patient harm. | 中 | SE024 |
| CE035 | The same study found 64–72% of residual medical hallucinations stemmed from reasoning failures rather than pure knowledge gaps. | 中 | SE024 |
| CE036 | Public uptime SLAs, historical incident reporting, and query-latency benchmarks remain open questions in the public materials reviewed for this chapter. | 低 | |
| CE037 | OpenEvidence’s cited-answer model improves clinician verification, but the public sources reviewed here do not show an external audit of citation faithfulness in complex or patient-aware workflows. | 低 | SE012, SE016, SE025 |
| CE038 | OpenEvidence’s practical differentiation is workflow-native evidence retrieval inside Epic rather than forcing clinicians into separate Boolean or browser search workflows. | 中 | SE015, SE019, SE020 |
| CE039 | Public documentation suggests OpenEvidence is evolving from a single search interface into a broader clinician workbench spanning consults, research, documentation, coding, and communications. | 中 | SE002, SE020 |
| CE040 | Spoken answers in Voice Mode retain the same underlying references in the conversation, so the company treats voice as an interface extension rather than a different evidence standard. | 中 | SE003, SE020 |
| CU001 | OpenEvidence restricts access to verified healthcare professionals rather than to consumers. | 中 | SU001, SU005, SU018 |
| CU002 | OpenEvidence's primary public user persona is the individual clinician, especially physicians. | 中 | SU001, SU005 |
| CU003 | Official OpenEvidence surfaces say the platform is used daily, on average, by over 40% of U.S. physicians. | 中 | SU001, SU002, SU008 |
| CU004 | Official OpenEvidence surfaces say usage spans more than 10,000 hospitals and medical centers. | 中 | SU001, SU002, SU008 |
| CU005 | A July 2025 company press release said OpenEvidence was adding more than 65,000 new verified U.S. clinician registrations per month. | 中 | SU002 |
| CU006 | The same July 2025 company press release said OpenEvidence supported more than 8.5 million monthly clinical consultations. | 中 | SU002 |
| CU007 | Company releases and Sacra said OpenEvidence reached 1 million clinical consultations in a single 24-hour period on March 10, 2026. | 中 | SU003, SU007, SU024 |
| CU008 | NBC reported that OpenEvidence was used by about 65% of U.S. doctors across almost 27 million clinical encounters in April 2026. | 中 | SU005, SU006 |
| CU009 | Anupam Jena told NBC that about 60% of OpenEvidence searches are about making clinical decisions. | 中 | SU005, SU006 |
| CU010 | Sanford Health CMO Jeremy Cauwels told NBC that OpenEvidence is easy to adopt because it is free and phone-friendly. | 中 | SU005, SU006 |
| CU011 | Daniel Nadler told NBC that core OpenEvidence will always be free for users. | 中 | SU005 |
| CU012 | NBC, Sacra, and the App Store listing say OpenEvidence monetizes through pharmaceutical and medical-device advertising. | 中 | SU005, SU007, SU018 |
| CU013 | Sacra estimated OpenEvidence reached $150 million of annualized revenue in 2025. | 中 | SU007 |
| CU014 | Sacra estimated OpenEvidence's 2025 gross margin at about 90%. | 中 | SU007 |
| CU015 | Sacra said advertiser CPMs can range from $70 to $1,000+. | 中 | SU007 |
| CU016 | Sacra said clinical consultation volume reached about 20 million per month in January 2026. | 中 | SU007 |
| CU017 | Sacra said the free access model helped OpenEvidence bypass traditional hospital procurement cycles. | 中 | SU007, SU025 |
| CU018 | STAT reported that OpenEvidence is now seeking formal health-system relationships after building a massive doctor user base. | 中 | SU025 |
| CU019 | Cedars-Sinai announced enterprise access to OpenEvidence for physicians, nurses, pharmacists, and therapists. | 高 | SU004, SU013, SU014 |
| CU020 | Cedars-Sinai said OpenEvidence links medical literature to patient-specific EHR context. | 高 | SU013, SU014 |
| CU021 | Cedars-Sinai said patient data used in these contextual queries will not be stored by OpenEvidence. | 高 | SU013, SU014 |
| CU022 | Mount Sinai announced enterprise deployment of OpenEvidence across its seven hospitals. | 高 | SU015, SU016, SU017 |
| CU023 | Mount Sinai said nurses and pharmacists, not just physicians, will have access to OpenEvidence in the Epic workflow. | 高 | SU015, SU017 |
| CU024 | Becker's described the Mount Sinai agreement as OpenEvidence's first enterprise deal with a health system. | 中 | SU016, SU017 |
| CU025 | Healthcare IT News and Sacra both referenced Sutter Health integrating OpenEvidence into Epic workflows in 2026. | 中 | SU015, SU007 |
| CU026 | ACOG said its collaboration integrates ACOG guidance directly into OpenEvidence. | 中 | SU010 |
| CU027 | ACOG said women's health is one of the most frequently searched topics on OpenEvidence. | 中 | SU010 |
| CU028 | Wiley said its partnership adds the Cochrane Database of Systematic Reviews and more than 400 journals to OpenEvidence. | 高 | SU011, SU012 |
| CU029 | OpenEvidence markets NEJM and JAMA as official AI partners. | 中 | SU001, SU002, SU009 |
| CU030 | An official testimonial from Dr. John Lee says OpenEvidence helped him find information he could not find through Google or PubMed alone. | 中 | SU001, SU018, SU020 |
| CU031 | An official testimonial from Dr. Ram Dandillaya says OpenEvidence is more up-to-date than UpToDate for specific patient fact patterns. | 中 | SU001, SU018, SU020 |
| CU032 | An official testimonial from Dr. Antonio Jorge Forte says OpenEvidence can be foundational technology for clinical decision tools. | 中 | SU001, SU018, SU020 |
| CU033 | Paul Sax told NBC that OpenEvidence often borders on miraculous. | 中 | SU005, SU006 |
| CU034 | The App Store listing shows a 4.9 out of 5 rating from 10K ratings. | 中 | SU018, SU019 |
| CU035 | AppBrain reports a 4.80 out of 5 rating from 3,706 reviews. | 中 | SU021 |
| CU036 | AppBrain reports more than 100,000 Google Play downloads and approximately 420,000 lifetime downloads. | 中 | SU021 |
| CU037 | The App Store listing says the app contains advertising. | 中 | SU018 |
| CU038 | Trustpilot's archived page shows a 1.6 out of 5 rating from 23 reviews. | 中 | SU022 |
| CU039 | Multiple archived Trustpilot reviews allege inaccurate or harmful ME/CFS guidance. | 中 | SU022 |
| CU040 | NBC reported that some clinicians worry about hallucinations, incomplete answers, and erosion of clinical judgment. | 中 | SU005, SU006 |
| CU041 | NBC said some health systems remain cautious about PHI handling, citing MaineHealth guidance not to enter PHI into OpenEvidence. | 中 | SU005, SU006 |
| CU042 | NBC said some clinicians click through citations only when an answer looks surprising. | 中 | SU005, SU006 |
| CU043 | NBC reported that an unpeer-reviewed December study found OpenEvidence answered more complex medical questions accurately less than 45% of the time. | 低 | SU005, SU006 |
| CU044 | AusDoc quoted an expert warning that limiting evidence largely to JAMA and NEJM sources can raise AI bias questions. | 中 | SU023 |
| CU045 | The reviewed public sources do not disclose NRR, GRR, churn, renewal rates, or contract lengths for OpenEvidence customers. | 中 | SU001, SU005, SU007, SU008, SU013, SU015 |
| CU046 | The reviewed public sources do not disclose top-advertiser concentration or the enterprise share of revenue. | 中 | SU005, SU007, SU025 |
| CU047 | Public customer proof is strongest on named enterprise deployments and named clinician testimonials rather than on spend, renewal, or patient-outcome evidence. | 中 | SU013, SU015, SU018, SU022 |
| CU048 | Journal and medical-society deals are partner proof rather than proof that those organizations are paying customers of the core product. | 高 | SU010, SU011, SU012 |
| CU049 | OpenEvidence's public expansion pattern is land individual doctors first and then sell into health systems. | 中 | SU007, SU015, SU025 |
| CU050 | Public enterprise deployments expand the user base beyond physicians to multidisciplinary care teams. | 高 | SU013, SU015, SU017 |
| CU051 | Content partnerships with ACOG, Wiley, NEJM, JAMA, NCCN, and Cochrane deepen specialty coverage but do not prove monetization. | 中 | SU010, SU011, SU001 |
| CU052 | Free bottom-up adoption can create real usage without proving paid durability or retention. | 中 | SU007, SU025, SU005 |
| CR001 | FDA’s January 2026 clinical decision support guidance says certain CDS software functions are excluded from device status only if they satisfy all four non-device CDS criteria in section 520(o)(1)(E) of the FD&C Act. | 高 | SR001, SR002 |
| CR002 | One of those criteria requires that a healthcare professional be able to independently review the basis for the recommendation so the clinician does not rely primarily on the software to make a diagnosis or treatment decision. | 高 | SR001, SR002 |
| CR003 | OpenEvidence’s December 2025 terms describe the platform as SaaS that supports healthcare professionals, medical researchers, and clinical teams with clinical decision support tools and medical content. | 高 | SR011, SR010 |
| CR004 | The same terms say the service is educational and informational only and is not intended to serve as a diagnostic service, recommend a particular therapy, or substitute for a qualified clinician’s judgment. | 中 | SR011 |
| CR005 | OpenEvidence reserves the right to limit availability by person, geographic area, or jurisdiction, and users outside the United States are responsible for local-law compliance. | 中 | SR011 |
| CR006 | HHS OCR says disclosures of PHI to tracking technology vendors for marketing without HIPAA-compliant authorization are impermissible, and BAAs are required when those vendors function as business associates. | 高 | SR006, SR012 |
| CR007 | OpenEvidence’s privacy policy says it may collect or match NPI or medical-license data and use that information to tailor content and advertisements to a user’s interests and needs. | 中 | SR010 |
| CR008 | The privacy policy and terms explicitly describe sponsored programs, advertising, market-research opportunities, and audience-extension advertising in the United States as part of the platform’s commercial design. | 高 | SR010, SR011 |
| CR009 | OpenEvidence’s security page says the company supports HIPAA workflows, maintains SOC 2 Type II certification, encrypts data in transit and at rest, runs annual penetration tests, and uses standard BAAs for PHI handling. | 高 | SR012, SR011 |
| CR010 | OpenEvidence publicly announced HIPAA compliance and secure PHI uploads, confirming the product is moving beyond generic search into regulated patient-data workflows. | 中 | SR013, SR012 |
| CR011 | OpenEvidence sued Doximity in June 2025 alleging physician impersonation, prompt stealing, scraping, trade-secret misappropriation, unfair competition, and defamation; the CourtListener docket confirms the complaint and subsequent case activity. | 高 | SR041, SR040 |
| CR012 | Voice Mode turns OpenEvidence into a hands-free, real-time spoken interface with written transcripts and references, enlarging the operational surface for speech-recognition, context, and transcript-review errors. | 中 | SR015, SR031 |
| CR013 | OpenEvidence’s terms disclose that Visits records conversations and processes them on OpenEvidence systems, while placing consent, de-identification, and authorization obligations on the user. | 高 | SR011, SR012 |
| CR014 | GLACIS argues liability risk rises when clinicians and hospitals cannot reconstruct which audio, transcript, model version, prompt path, and guardrail state produced a contested output. | 中 | SR030, SR033 |
| CR015 | Patient Safety & Quality Healthcare summarizes a JAMA Network Open study finding that AI-generated discharge summaries contained incomplete or misleading information in about 18% of cases. | 中 | SR031 |
| CR016 | Stanford’s 2026 clinical AI review says many physician-level or “superhuman” claims depend on narrow benchmarks that do not reflect uncertainty, incomplete information, or real workflow complexity. | 高 | SR029, SR031 |
| CR017 | The same review says nearly half of more than 500 medical AI studies used exam-style questions and only about five percent used real patient data. | 中 | SR029 |
| CR018 | Stanford also highlights over-reliance as a separate safety issue: clinicians in some studies followed incorrect AI recommendations even when the errors were detectable. | 高 | SR029, SR031 |
| CR019 | HIMSS argues healthcare AI should be governed with the same rigor as any clinical intervention, not through ad-hoc experimentation. | 高 | SR033, SR029 |
| CR020 | Microsoft’s Dragon Copilot materials include an explicit medical-device disclaimer and say the product is not a substitute for professional medical advice, diagnosis, treatment, or judgment. | 中 | SR042, SR011 |
| CR021 | Wiley’s 2026 partnership gives OpenEvidence access to over 400 journals and books plus Cochrane Database of Systematic Reviews and Cochrane Clinical Answers. | 高 | SR022, SR023 |
| CR022 | Cochrane says the partnership is intended to help ensure AI outputs are based on the best possible evidence, underscoring how central licensed content is to OpenEvidence’s quality promise. | 高 | SR023, SR022 |
| CR023 | Open Pharma describes OpenEvidence as drawing on NEJM, JAMA, NCCN, Cochrane, and Wiley, but questions whether users can know that summaries are representative of the full evidence base when not every journal is covered. | 中 | SR027, SR028 |
| CR024 | OpenEvidence’s security page says the service is primarily hosted on Google Cloud Platform and Vercel, creating infrastructure dependence outside the company’s direct control. | 中 | SR012 |
| CR025 | OpenAI for Healthcare now offers hospitals HIPAA-supporting BAAs, evidence retrieval with transparent citations, institutional-policy integration, and enterprise controls. | 高 | SR034, SR024 |
| CR026 | Anthropic’s Claude for Healthcare offers HIPAA-ready products plus connectors to CMS coverage data, ICD-10, NPI, and PubMed. | 高 | SR035, SR024 |
| CR027 | Google’s MedLM and Vertex AI Search for Healthcare give providers medically tuned models, multimodal clinical search, and enterprise deployment paths for clinician questions and documentation workflows. | 高 | SR038, SR039 |
| CR028 | Doximity says more than 85% of U.S. physicians already use its platform and that its Clinical AI Suite combines Ask, Scribe, and Dialer with free full-PDF journal access. | 中 | SR036 |
| CR029 | Wolters Kluwer launched UpToDate Expert AI in 2025 with enterprise deployment, step-by-step rationale, source traceability, and 7,600 medical experts behind its content. | 高 | SR037, SR042 |
| CR030 | Microsoft says Dragon Copilot will surface Elsevier, OpenEvidence, and UpToDate inside the same enterprise workflow shell, implying standalone query share can be displaced by larger workflow platforms. | 高 | SR042, SR043 |
| CR031 | CNBC says OpenEvidence exceeded $100 million in annualized revenue in 2025 and intentionally relies on in-app video advertising rather than subscription pricing as its primary monetization engine. | 高 | SR024, SR026 |
| CR032 | CNBC and Fierce both report that more than 40% of U.S. physicians use OpenEvidence, showing strong bottoms-up adoption but also tying growth to clinician habit rather than locked enterprise contracts. | 高 | SR024, SR025 |
| CR033 | STAT says OpenEvidence now has to prove that its ad-based model can coexist with formal hospital relationships as the company seeks system-level deals. | 高 | SR026, SR024 |
| CR034 | Open Pharma and Krafty Librarian both frame sponsor adjacency and source-opacity as trust risks, especially if specialty-targeted advertising or incomplete source coverage shapes what clinicians see. | 中 | SR027, SR028 |
| CR035 | Krafty Librarian argues that external information professionals cannot independently inspect OpenEvidence’s source coverage or search precision because access is gated to verified U.S. clinicians with an NPI. | 中 | SR028 |
| CR036 | Bloomberg Law says OpenEvidence access is exclusive to healthcare professionals using a 10-digit CMS identifier, and the complaint alleges Doximity employees impersonated physicians to get in. | 高 | SR041, SR040 |
| CR037 | Open Pharma reports that OpenEvidence withdrew EU and UK access on 27 April 2026 because of mounting uncertainty under the EU AI Act, showing that international constraints are already concrete rather than hypothetical. | 中 | SR027, SR008 |
| CR038 | The European Commission says high-risk AI systems face obligations around risk management, data quality, logging, documentation, human oversight, cybersecurity, and accuracy, with main transparency rules taking effect in August 2026 and embedded-product rules pushed further out. | 高 | SR008, SR001 |
| CR039 | Public coverage still centers Daniel Nadler as the founder-CEO, principal moat narrator, and public face of the company’s strategy, making leadership continuity a material execution variable. | 高 | SR024, SR025 |
| CR040 | Nadler’s own moat argument depends on physician focus, premium content, and hundreds of millions of consultations from verified physicians, reinforcing both a real data asset and meaningful founder concentration. | 中 | SR024, SR025 |
| CR041 | OpenEvidence’s terms and privacy policy allow extensive commercial use of nonpersonal or usage data, advertising, and market research, which can become a governance and trust problem if hospitals view sponsor adjacency as incompatible with bedside decision support. | 中 | SR011, SR010, SR027 |
| CR042 | Mount Sinai’s Dragon Copilot rollout after a multi-vendor evaluation shows large health systems are willing to standardize on rival clinical-AI stacks when governance, workflow fit, and enterprise integration are stronger. | 高 | SR043, SR042 |
| CR043 | Stanford says OpenEvidence’s rise shows doctors are bypassing traditional IT gatekeepers to use AI in clinical care, accelerating adoption but also increasing the chance that governance trails usage. | 高 | SR029, SR033 |
| CR044 | The highest residual exposures for OpenEvidence as of May 2026 are FDA/CDS classification drift, malpractice-liability from hallucination or over-reliance, privacy/HIPAA risk around PHI and ad-tech, advertiser-conflict and monetization fragility, content-licensing dependence, founder concentration, general-model competition, and constrained international expansion. | 中 | SR001, SR006, SR022, SR024, SR027, SR029 |
| CR045 | The most visible mitigations are HCP-only gating, contractual clinician-responsibility language, BAAs and HIPAA/SOC2 claims, source citations, diversified publisher licenses, and early enterprise outreach, but none removes the need for human review and hospital governance. | 中 | SR011, SR012, SR022, SR026 |
| CR046 | OpenEvidence has publicized a Mayo Clinic-branded benchmark claim that it performs comparably to physician clinical decision-making in common scenarios, but that kind of company-claimed challenge result does not substitute for independent real-world outcome evidence. | 中 | SR021, SR029 |
| CV001 | OpenEvidence's February 2025 Series A raised $75 million at a $1 billion valuation. | 高 | SV001, SV002 |
| CV002 | OpenEvidence's July 2025 Series B raised $210 million at a $3.5 billion valuation. | 高 | SV003, SV004, SV025 |
| CV003 | OpenEvidence's October 2025 Series C raised $200 million at a $6 billion valuation. | 中 | SV006, SV025 |
| CV004 | OpenEvidence's January 2026 Series D raised $250 million at a $12 billion valuation. | 高 | SV005, SV006, SV007, SV025 |
| CV005 | OpenEvidence raised roughly $700 million over the prior 12 months through January 2026. | 高 | SV005, SV006, SV007, SV025 |
| CV006 | OpenEvidence reported surpassing $100 million in annual revenue by late 2025. | 高 | SV005, SV007, SV008, SV026 |
| CV007 | Sacra estimates OpenEvidence reached about $150 million of annualized 2025 revenue after growing from about $7.9 million in 2024 with roughly 90% gross margins. | 中 | SV009 |
| CV008 | OpenEvidence is free to physicians and monetizes primarily through advertising. | 高 | SV002, SV009, SV024, SV026 |
| CV009 | The company says more than 40% of U.S. physicians use OpenEvidence daily across more than 10,000 hospitals and medical centers. | 中 | SV006, SV007, SV026, SV027 |
| CV010 | OpenEvidence supported about 18 million clinical consultations in December 2025 versus about 3 million per month a year earlier. | 高 | SV005, SV006, SV026 |
| CV011 | OpenEvidence says it has supported more than 200 million cumulative AI-powered clinical consultations. | 中 | SV024, SV027 |
| CV012 | OpenEvidence says clinicians using the platform will treat or treated more than 100 million Americans in 2025 or 2026. | 中 | SV005, SV008, SV024, SV026 |
| CV013 | OpenEvidence cites official AI partnerships or content agreements with NEJM, JAMA, NCCN, Cochrane, Wiley, and other medical publishers. | 中 | SV003, SV024 |
| CV014 | Sacra reports that OpenEvidence's ad model can command CPMs of roughly $70 to $1,000+ and about $124 ARPU. | 中 | SV009 |
| CV015 | OpenEvidence's valuation increased from $1 billion in February 2025 to $12 billion in January 2026, a 12x step-up in under a year. | 中 | SV002, SV005, SV007, SV025 |
| CV016 | Sacra says the July 2025 $3.5 billion mark implied roughly a 70x multiple on about $50 million of annualized revenue. | 中 | SV009 |
| CV017 | Using Sacra's $150 million annualized 2025 revenue estimate, the January 2026 $12 billion valuation implies roughly an 80x revenue multiple. | 中 | SV005, SV009 |
| CV018 | Doximity's market cap was about $3.64 billion in May 2026. | 中 | SV013 |
| CV019 | Doximity's trailing revenue was about $0.63 billion in 2025. | 中 | SV014 |
| CV020 | Doximity traded at roughly 5.8x market cap to trailing revenue in May 2026. | 中 | SV013, SV014 |
| CV021 | Sacra describes Doximity as generating about $570 million of trailing revenue with roughly 80% from advertising and about $228 ARPU. | 中 | SV009 |
| CV022 | Tempus AI's market cap was about $8.29 billion in May 2026. | 中 | SV015 |
| CV023 | Tempus AI's trailing revenue was about $1.27 billion in 2025. | 中 | SV016 |
| CV024 | Tempus AI traded at roughly 6.5x market cap to trailing revenue in May 2026. | 中 | SV015, SV016 |
| CV025 | Veeva Systems' market cap was about $26.13 billion in May 2026. | 中 | SV017 |
| CV026 | Veeva Systems generated about $3.19 billion of trailing revenue in 2026 TTM. | 中 | SV018 |
| CV027 | Veeva traded at roughly 8.2x market cap to trailing revenue in May 2026. | 中 | SV017, SV018 |
| CV028 | Health Catalyst's market cap was about $95 million in May 2026 against roughly $310 million of TTM revenue. | 中 | SV019, SV020 |
| CV029 | Health Catalyst traded at roughly 0.3x market cap to trailing revenue in May 2026. | 中 | SV019, SV020 |
| CV030 | Phreesia's market cap was about $0.55 billion in May 2026 against roughly $0.46 billion of TTM revenue. | 中 | SV021, SV022 |
| CV031 | Phreesia traded at roughly 1.2x market cap to trailing revenue in May 2026. | 中 | SV021, SV022 |
| CV032 | Hippocratic AI raised $126 million at a $3.5 billion valuation in late 2025, showing that private medical-AI scarcity still commands a premium but below OpenEvidence's $12 billion mark. | 中 | SV012 |
| CV033 | UpToDate remains a directly relevant clinical knowledge reference because it monetizes evidence-based physician workflow software through seat licenses rather than ads. | 中 | SV009, SV023 |
| CV034 | FDA's January 2026 CDS guidance says certain clinician-facing decision-support software can remain outside device regulation if clinicians can independently review the basis and the software is not the sole basis for decisions. | 中 | SV010 |
| CV035 | AMA principles for clinical AI call for transparency, oversight, and documentation in deployment. | 中 | SV011 |
| CV036 | OpenEvidence's move toward multi-agent "medical superintelligence" could increase oversight risk if product behavior becomes more prescriptive than the transparent support model contemplated by FDA's non-device CDS boundary. | 中 | SV005, SV010, SV011 |
| CV037 | OpenEvidence's content moat and physician distribution are real strengths because the company combines broad clinician reach with exclusive publisher relationships and specialist-trained medical models. | 中 | SV005, SV024, SV027 |
| CV038 | The ad-supported model lowers adoption friction but leaves long-term monetization quality dependent on advertiser demand, content-partner durability, and clinician trust. | 中 | SV002, SV009, SV024, SV026 |
| CV039 | OpenEvidence's current $12 billion mark is more than 3x Doximity's public market cap despite far lower disclosed revenue. | 中 | SV007, SV013, SV014 |
| CV040 | OpenEvidence's $12 billion valuation is about 46% of Veeva's market cap despite Veeva generating more than $3 billion of revenue. | 中 | SV005, SV017, SV018 |
| CV041 | The comparable set suggests today's price already assumes sustained hypergrowth and premium monetization rather than merely continued product relevance. | 中 | SV009, SV013, SV014, SV015, SV016, SV017, SV018, SV021, SV022 |
| CV042 | The clearest bull-case supports are unusually fast physician adoption, a differentiated content moat, and a still-open private funding market for healthcare AI. | 中 | SV005, SV006, SV012, SV024 |
| CV043 | The clearest anti-thesis is that public evidence on revenue durability, advertiser concentration, enterprise contract depth, and cap-table overhang is still thin relative to the valuation step-up. | 中 | SV009, SV010, SV011 |
| CV044 | A reasonable public-evidence base case is that OpenEvidence deserves a large but not $12 billion valuation until monetization and governance disclosure catch up with adoption. | 中 | SV009, SV013, SV014, SV015, SV016, SV017, SV018, SV021, SV022 |
| CV045 | The current $12 billion price looks supportable only in a bull case where revenue roughly doubles again, physician penetration remains above 40%, and monetization expands beyond today's ad-led model. | 中 | SV005, SV006, SV007, SV009, SV024 |
| CV046 | A bear case would combine lower ad demand, slower clinician engagement, competitive share loss, or tighter oversight and could compress value toward low-single-digit billions. | 中 | SV009, SV010, SV011, SV021, SV022 |
| CV047 | The evidence supports a research-more recommendation rather than a buy because company quality is strong but the entry price already embeds unusually optimistic assumptions. | 中 | SV009, SV013, SV014, SV017, SV018, SV025 |
| CV048 | A key thesis-break trigger is any disclosure showing revenue scale or advertiser retention materially below what a $12 billion mark requires. | 中 | SV009, SV026 |
| CV049 | Another thesis-break trigger is any regulatory or partner change that narrows the product's ability to deliver citation-transparent physician support while staying outside heavier device oversight. | 中 | SV010, SV011, SV024 |
| CV050 | Final diligence should focus on audited revenue and gross margin, advertiser concentration and renewals, enterprise contract pipeline, cap-table preferences, and clinical quality governance. | 中 | SV009, SV010, SV011, SV026 |
| CV051 | Doximity, Veeva, and Health Catalyst are SEC-reporting public issuers, so their May 2026 market-cap data are continuously repriced public comparables rather than private marks. | 中 | SV013, SV017, SV019, SV028, SV029, SV030 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | OpenEvidence | OpenEvidence | OpenEvidence is the leading medical information platform. |
| SO002 | OpenEvidence | About | OpenEvidence | To date, we have supported over 200 million AI-powered clinical consultations from U.S. doctors and other frontline clinicians. |
| SO003 | OpenEvidence | Ask Overview | User Guide | OpenEvidence | Every answer cites its sources, grounding answers in the guidelines and medical literature and supporting exploration of the evidence. |
| SO004 | OpenEvidence | Deep Consult | User Guide | OpenEvidence | An advanced AI agent that generates PhD-level research reports to help you ramp up on a body of knowledge or answer the most complex questions. |
| SO005 | OpenEvidence | Visits Overview | User Guide | OpenEvidence | Record encounters, generate clinical notes, and surface real-time decision support — on web and mobile. |
| SO006 | OpenEvidence | Voice Mode | User Guide | OpenEvidence | Hands-free, real-time medical conversations — speak your clinical question and hear evidence-backed answers read aloud. |
| SO007 | PR Newswire | OpenEvidence Achieves $1 Billion Valuation in Sequoia-led Round and Announces Content Partnership with the New England Journal of Medicine | OpenEvidence, the fastest-growing platform for doctors in history, has closed a Series A with Sequoia Capital at a $1 billion valuation. |
| SO008 | PR Newswire | OpenEvidence, the Fastest-Growing Application for Physicians in History, Announces $210 Million Round at $3.5 Billion Valuation | OpenEvidence, the most widely used medical search and AI application among verified U.S. clinicians, today announced a $210 million Series B round at a $3.5 Billion valuation. |
| SO009 | CNBC | OpenEvidence, 'ChatGPT for doctors,' doubles valuation to $12 billion | OpenEvidence, based in Miami, Florida, closed a $250 million financing, led by Thrive Capital and DST, the company told CNBC. |
| SO010 | CNBC | 32. OpenEvidence | The company, founded just five years ago, estimates that 40% of U.S. doctors use its tool daily and says it counts 10,000 medical centers as clients. |
| SO011 | NBC News | Your doctor is probably using AI, even if they haven’t told you about it. | Yet with OpenEvidence’s skyrocketing popularity, some experts worry about potential hallucinations or incomplete answers, a lack of rigorous scientific studies on the tool’s patient impact, and the potential for doctors’ critical thinking and evaluation skills to erode with increased OpenEvidence use and dependence. |
| SO012 | Fierce Healthcare | JPM26: OpenEvidence co-founder talks the future of healthcare AI | It is now actively used daily, on average, by more than 40% of physicians in the U.S., spanning more than 10,000 hospitals and medical centers nationwide, according to the company. |
| SO013 | Fierce Healthcare | HLTH25: OpenEvidence valuation hits $6B with $200M series C | The three-year-old company’s valuation hit $6 billion post-series C raise, Daniel Nadler, Ph.D., one of OpenEvidence’s founders confirmed to Fierce Healthcare on Monday. |
| SO014 | MobiHealthNews | OpenEvidence scores $250M, doubles valuation to $12B | OpenEvidence, an AI-enabled medical research aggregate platform for doctors, has closed a $250 million Series D funding round, bringing its total raise to nearly $700 million over the past 12 months and doubling its valuation to $12 billion. |
| SO015 | Medical Economics | Clinical AI platform to be trained on 30 years of NEJM archives | Under the multi-year agreement, OpenEvidence will integrate published content and multimedia from NEJM and its affiliated publications ... dating back to 1990. |
| SO016 | Business Wire | OpenEvidence Raises $250 Million to Build Medical Superintelligence for Doctors | OpenEvidence, which is free to doctors and ad-supported, became one of the fastest companies in history of any kind to reach $100m in annual revenue. |
| SO017 | PR Newswire | OpenEvidence Achieves Historic Milestone: 1 Million Clinical Consultations Between Verified Doctors and an Artificial Intelligence System in a Single Day | On March 10, 2026, for the first time in history, one million clinical consultations were conducted between NPI verified physicians and an AI system, OpenEvidence, within a single 24-hour period. |
| SO018 | PR Newswire | OpenEvidence and the JAMA Network sign strategic content agreement | Under this agreement, all published content from JAMA, JAMA Network Open, and the 11 JAMA specialty journals ... including full text and multimedia, will inform answers delivered on the OpenEvidence platform. |
| SO019 | Cochrane | Cochrane evidence to inform OpenEvidence users | A new partnership between Cochrane’s publishing partner Wiley and the medical AI platform OpenEvidence will see Cochrane evidence helping to inform the platform’s users, including more than 40% of US physicians. |
| SO020 | Wiley | Wiley and OpenEvidence Partner to Deliver Trusted Research to Physicians at the Point of Care | Wiley will license a comprehensive portfolio of scientific and medical content to OpenEvidence, including the Cochrane Database of Systematic Reviews ... and over 400 Wiley journals and books. |
| SO021 | OpenEvidence | OpenEvidence is now HIPAA compliant | OpenEvidence is now HIPAA compliant: Health care professionals using OpenEvidence can securely upload PHI. |
| SO022 | OpenEvidence | OpenEvidence Launches “Visits”: Real-Time Medical Intelligence for the Patient Visit | OpenEvidence Launches “Visits”: Real-Time Medical Intelligence for the Patient Visit. |
| SO023 | OpenEvidence | OpenEvidence Collaborates With NCCN to Integrate Canonical Oncology Treatment Algorithms at the Point-of-Care | OpenEvidence Collaborates With NCCN to Integrate Canonical Oncology Treatment Algorithms at the Point-of-Care. |
| SO024 | OpenEvidence | Mayo Clinic study: OpenEvidence performs comparably to physician CDM in common clinical scenarios | Mayo Clinic study: OpenEvidence performs comparably to physician CDM in common clinical scenarios. |
| SO025 | Bloomberg Law | Medical AI Firm Says Competitor Hacked Prompts to Steal Secrets | Doximity Inc. executives impersonated physicians and conducted a months-long cyberattack using malicious inputs to extract trade secrets from competitor OpenEvidence Inc.’s artificial intelligence model. |
| SO026 | CourtListener | OpenEvidence Inc. v. Doximity, Inc. | COMPLAINT against Jake Konoske, Jey Balachandran, Doximity, Inc. ... filed by OpenEvidence Inc. (Entered: 06/20/2025) |
| SO027 | HealthExec | Doximity and OpenEvidence sue each other in spat over medical AI trade secrets | Two of the top producers of a “ChatGPT for doctors” are suing each other, each claiming the other is guilty of back-and-forth civil violations that include corporate espionage and defamation. |
| SO028 | medRxiv | Ad-verse Effects: Pharmaceutical Advertising Shifts Drug Recommendations by Consumer-Facing AI | When two drugs were both guideline-appropriate, advertising shifted selection of the advertised drug by +12.7 percentage points. |
| SO029 | Gizmodo | Your Doctor Is Most Likely Consulting This Free AI Chatbot, Report Says | How would you like it if ... your doctor consulted a free, ad-supported AI chatbot? That’s not actually a hypothetical. |
| SO030 | S&P Global | S&P Global to Acquire Kensho; Bolsters Core Capabilities in Artificial Intelligence, Natural Language Processing and Data Analytics | S&P Global will acquire Kensho for approximately $550m, net of cash acquired, for a mix of cash and stock. |
| SO031 | OpenEvidence | Sutter Health Collaborates with OpenEvidence to Bring Evidence-Based, AI-Powered Insights into Physician Workflows | Sutter Health Collaborates with OpenEvidence to Bring Evidence-Based, AI-Powered Insights into Physician Workflows. |
| SO032 | OpenEvidence | OpenEvidence Partners with Cedars-Sinai to Create Patient-Aware Clinical Intelligence With Agentic Clinical AI | OpenEvidence Partners with Cedars-Sinai to Create Patient-Aware Clinical Intelligence With Agentic Clinical AI. |
| SM001 | Agency for Healthcare Research and Quality | Clinical Decision Support | |
| SM002 | Office of the National Coordinator for Health Information Technology | Clinical Decision Support | |
| SM003 | U.S. Food and Drug Administration | Clinical Decision Support Software - Guidance | |
| SM004 | American Medical Association | More than 80% of physicians use AI professionally: AMA survey | |
| SM005 | American Medical Association | AMA Augmented Intelligence Research | AMA | |
| SM006 | HealthLeaders | Physician AI Adoption Surges, Forcing Health System Leaders to Shift From Experimentation to Governance | |
| SM007 | HIMSS | HIMSS Releases Guidance for Responsible AI Governance and Deployment in Healthcare | |
| SM008 | HIMSS | Operationalizing AI: A Strategic Framework for Safe Deployment in Healthcare | |
| SM009 | HIMSS | Five AI Signals from HIMSS26 That Will Shape the Next Era of Healthcare | |
| SM010 | HIMSS | AI-Supported Decision-Making in Clinical Settings: Transforming Healthcare Delivery | |
| SM011 | The Business Research Company | The Business Research Company - Market Research & Business Intelligence | |
| SM012 | Fortune Business Insights | Clinical Decision Support Systems Market Size, Share [2034] | |
| SM013 | Grand View Research | Artificial Intelligence In Clinical Decision Support Market Report, 2033 | |
| SM014 | Research and Markets | Clinical Decision Support Systems Market Report 2026 | |
| SM015 | U.S. Bureau of Labor Statistics | Physicians and Surgeons | |
| SM016 | U.S. Bureau of Labor Statistics | Physician Assistants | |
| SM017 | U.S. Bureau of Labor Statistics | Nurse Anesthetists, Nurse Midwives, and Nurse Practitioners | |
| SM018 | Wolters Kluwer | Why Use UpToDate | Wolters Kluwer | |
| SM019 | EBSCO | DynaMed | Point of Care | Clinical Decision Support | Evidence Based | EBSCO | |
| SM020 | EBSCO | EBSCO Clinical Decisions Launches Dyna AI Mode - News | EBSCO | |
| SM021 | Epocrates | Upgrade to epocrates+ | epocrates | |
| SM022 | Epocrates | epocrates+ Group | |
| SM023 | AMBOSS | AMBOSS: Empowering doctors to provide the best possible care | |
| SM024 | AMBOSS | AMBOSS Pricing | |
| SM025 | Glass Health | Glass Health | Ambient Scribing & Clinical Decision Support | |
| SM026 | Pathway | Pathway | |
| SM027 | PubMed | Distinct components of alert fatigue in physicians' responses to a noninterruptive clinical decision support alert - PubMed | |
| SM028 | Journal of the Medical Library Association / PMC | Unanswered clinical questions: a survey of specialists and primary care providers | |
| SM029 | JAMA Network Open | Electronic Health Record Use During Paid Time Off Among Primary Care Physicians | |
| SM030 | OpenEvidence | OpenEvidence | |
| SP001 | OpenEvidence | OpenEvidence | America's Official Medical Knowledge Platform. |
| SP002 | OpenEvidence | About | OpenEvidence | To date, we have supported over 200 million AI-powered clinical consultations from U.S. doctors and other frontline clinicians. |
| SP003 | OpenEvidence | Security and Compliance | OpenEvidence | We fully comply with the U.S. Health Insurance Portability and Accountability Act (HIPAA) ... OpenEvidence has achieved SOC 2 Type II certification. |
| SP004 | BusinessWire | OpenEvidence Raises $250 Million to Build Medical Superintelligence for Doctors | OpenEvidence, which is free to doctors and ad-supported, became one of the fastest companies in history of any kind to reach $100m in annual revenue. |
| SP005 | Glass Health | Glass Health | Ambient Scribing & Clinical Decision Support | Glass Health | Ambient Scribing & Clinical Decision Support |
| SP006 | Pathway | Pathway | Pathway joined Doximity in 2025, and its capabilities now live on in Doximity Ask, a free AI-powered clinical reference tool. |
| SP007 | Doximity | Introducing the Doximity Clinical AI Suite | AI-powered tools built to support physicians across every layer of their daily workflow, all on the platform where more than 85% of U.S. physicians already practice. |
| SP008 | Doximity | Doximity Ask FAQs | AI outputs may occasionally include inaccuracies (hallucinations) ... Always verify clinical outputs. |
| SP009 | BusinessWire | Doximity Acquires Pathway, a Leader in AI Clinical Reference | Hundreds of thousands of users have registered for Pathway, and thousands pay $300 per year for our premium product. |
| SP010 | Fierce Healthcare | Doximity acquires Pathway Medical for $63M to bolster AI tools for doctors | At present, we believe the company will continue to provide the full AI toolset for free to providers but expect that over the longer term ... it could move to sell enterprise-grade solutions to health systems. |
| SP011 | MobiHealthNews | Doximity acquires Pathway Medical for $63M | Doximity, a networking platform for healthcare professionals, announced it acquired Montreal-based Pathway Medical ... in a deal worth $63 million. |
| SP012 | Wolters Kluwer | UpToDate: Trusted, evidence-based solutions for modern healthcare | Explore personal and small group subscription options ... trusted by over 3 million health professionals worldwide. |
| SP013 | Wolters Kluwer | UpToDate Subscription Costs | UpToDate offers individual subscription options ... annual options (1, 2 and 3 year), as well as 30/90 Day Recurring Billing ... and group options. |
| SP014 | BusinessWire | Wolters Kluwer’s New UpToDate Expert AI Provides Clinicians and Health Systems with the Fast, Reliable GenAI Clinical Decision Support They Need | Users receive context-rich responses ... with single-click access to the assumptions driving the answers, the source of the answer in UpToDate, and the step-by-step rationale used to generate it. |
| SP015 | EBSCO | DynaMedex® Recognized as 2026 Best in KLAS for Clinical Decision Support | DynaMedex combines two Best in KLAS solutions, DynaMed ... and Micromedex ... into a single, integrated solution. |
| SP016 | epocrates | Upgrade to epocrates+ | epocrates | Monthly $24.99 /month ... Annual $179.99 /year. |
| SP017 | Merative | Micromedex drug database | Merative | Micromedex is an award-winning clinical decision support solution ... trusted in over 80 countries around the world. |
| SP018 | Elsevier | ClinicalKey | Elsevier | Find answers faster with ClinicalKey AI ... delivering relevant insights at the point of care ... sourced from trusted, evidence-based content. |
| SP019 | OpenAI | Introducing OpenAI for Healthcare | Evidence retrieval with transparent citations ... Data control and support for HIPAA compliance ... Content shared with ChatGPT for Healthcare is not used to train models. |
| SP020 | OpenAI | OpenAI API Pricing | GPT-5.5 ... Input: $5.00 / 1M tokens ... Output: $30.00 / 1M tokens. |
| SP021 | OpenAI | Security and privacy at OpenAI | OpenAI supports our customers’ compliance ... including ... HIPAA ... and offers a ... Business Associate Agreement for customers. |
| SP022 | Claude by Anthropic | Healthcare | Claude by Anthropic | Claude is built on HIPAA-ready infrastructure with safety guardrails designed for healthcare. Answers can be traced back to the source, so you can verify before you act. |
| SP023 | Anthropic | Advancing Claude in healthcare and the life sciences | Since HIPAA-compliant organizations can now use Claude for Enterprise, they can also access ... PubMed ... Finally, we've added ... FHIR development. |
| SP024 | Claude by Anthropic | Plans & Pricing | Claude by Anthropic | Team ... $20 Per seat / month if billed annually. $25 if billed monthly ... HIPAA-ready offering available. |
| SP025 | U.S. Food and Drug Administration | Clinical Decision Support Software - Guidance | This guidance clarifies FDA’s thinking on the types of clinical decision support (CDS) software functions that are excluded from the definition of device. |
| SP026 | Frontiers in Digital Health | Through the looking glass: ethical considerations regarding LLM-induced hallucinations to medical questions | |
| SP027 | Electronics (MDPI) | Evaluating Retrieval-Augmented Generation Variants for Clinical Decision Support: Hallucination Mitigation and Secure On-Premises Deployment | Self-reflective RAG, on the other hand, lowered hallucinations to 5.8% ... incorporating encryption, provenance tagging, and audit trails. |
| SI001 | OpenEvidence | OpenEvidence | An Official AI Partner of The New England Journal of Medicine. |
| SI002 | OpenEvidence | About | OpenEvidence | This year, more than 100 million Americans will be treated by a clinician using OpenEvidence. |
| SI003 | OpenEvidence | Get Started | User Guide | OpenEvidence | Automatic CPT code suggestions, E/M level recommendations with MDM rationale written into your note, and ICD-10 codes — all generated the moment a visit is complete. |
| SI004 | OpenEvidence | Visits Overview | User Guide | OpenEvidence | We do not train AI models on protected health information. |
| SI005 | OpenEvidence | Deep Consult | User Guide | OpenEvidence | An advanced AI agent that generates PhD-level research reports to help you ramp up on a body of knowledge or answer the most complex questions. |
| SI006 | OpenEvidence | Sutter Health Collaborates with OpenEvidence to Bring Evidence-Based, AI-Powered Insights into Physician Workflows | Sutter Health and OpenEvidence are teaming up to deliver the latest medical data and information to doctors working within electronic health records. |
| SI007 | Mount Sinai Health System | Mount Sinai Health System Collaborates with OpenEvidence to Provide Evidence-Based Knowledge Within Electronic Medical Record | The collaboration marks the first enterprise-scale OpenEvidence deployment to extend access across the full clinical care team—including physicians, registered nurses, and pharmacists. |
| SI008 | Business Wire | OpenEvidence Raises $250 Million to Build Medical Superintelligence for Doctors | OpenEvidence, which is free to doctors and ad-supported, became one of the fastest companies in history of any kind to reach $100m in annual revenue. |
| SI009 | PR Newswire | OpenEvidence, the Fastest-Growing Application for Physicians in History, Announces $210 Million Round at $3.5 Billion Valuation | Each DeepConsult run requires over 100 times the compute and cost of a standard OpenEvidence search. |
| SI010 | PR Newswire | OpenEvidence and Veeva Announce Open Vista Partnership | Open Vista will use AI to increase patient access to clinical trials, accelerate drug discovery through better understanding of unmet needs, and improve understanding and adoption of existing approved medicines. |
| SI011 | CNBC | OpenEvidence, the 'ChatGPT for doctors,' doubles valuation to $12 billion | CEO Daniel Nadler told CNBC that OpenEvidence is used by 40% of physicians in the U.S. and topped $100 million in annual revenue last year. |
| SI012 | TechCrunch | OpenEvidence hits $12B valuation, with new round led by Thrive, DST | The company says that the free, ad-supported platform served 18 million clinical consultations from verified healthcare professionals in the U.S. in December alone. |
| SI013 | Fierce Healthcare | OpenEvidence clinches $250M, doubles valuation to $12B | The lion's share is training new models, training the next generation of digital intelligence, and compute costs. |
| SI014 | MedCity News | Thunderstruck By OpenEvidence’s $12B Valuation? Don’t Be. | According to its privacy policy, OpenEvidence collects information about how clinicians use the platform, including engagement with particular topics and device data. |
| SI015 | STAT | OpenEvidence makes its pitch to hospitals: 'We're not crazy ...' | The company is facing competitive pressures and questions about whether its ad-based business model can continue to propel the company forward. |
| SI016 | Sacra | OpenEvidence revenue, valuation & funding | Sacra estimates OpenEvidence hit $150 million in annualized revenue in 2025, up 1,803% from $7.9 million in 2024, with reported 90% gross margins. |
| SI017 | Sacra | OpenEvidence at $50M/year growing 30% MoM | Where UpToDate charges hospitals $500/seat, OpenEvidence gives its chatbot away for free, using pharma and med-device advertisements to monetize its audience. |
| SI018 | Cooley LLP | OpenEvidence Raises $250 Million Series D | The new funding will be used to invest heavily in the research and development and compute costs associated with the multi-AI agentic architecture of OpenEvidence. |
| SI019 | U.S. Securities and Exchange Commission | HII OpenEvidence-01 a Series of HII OpenEvidence LLC — Form D Filing | HII OpenEvidence-01 a Series of HII OpenEvidence LLC. |
| SI020 | U.S. Securities and Exchange Commission (EDGAR search) | SEC EDGAR search results for OpenEvidence Form D filings | The search returns one hit: HII OpenEvidence-01 a Series of HII OpenEvidence LLC. |
| SI021 | R&D World | Wiley moves deeper into clinical AI with OpenEvidence content licensing pact | The platform is free for verified physicians, supported by pharmaceutical advertising, and claims more than 18 million clinical consultations per month. |
| SI022 | Open Pharma | OpenEvidence at ISMPP Annual: what we heard, what we’re asking and what comes next | Without access to every journal publishing biomedical research, how can we be confident the generated summaries are truly representative of the full evidence base? |
| SI023 | The Krafty Librarian | OpenEvidence: Smart Medicine or Smart Marketing? | Opaque curation – OpenEvidence is not clear about its article selection and ranking. |
| SI024 | HLTH | OpenEvidence Raises $210M, Launches Free AI Agent for Physicians | Despite requiring over 100 times the compute and cost of a standard search, OpenEvidence is offering it entirely free to all verified U.S. clinicians. |
| SI025 | HIT Consultant | OpenEvidence Secures $210M, Launches Free AI Agent for Clinicians | OpenEvidence is offering DeepConsult entirely free to all verified U.S. clinicians, regardless of their institution or workplace. |
| SE001 | OpenEvidence | OpenEvidence | America's Official Medical Knowledge Platform. An Official AI Partner of The New England Journal of Medicine. |
| SE002 | OpenEvidence | Get Started | User Guide | OpenEvidence | |
| SE003 | OpenEvidence | Voice Mode | User Guide | OpenEvidence | Voice Mode is available on both web and the mobile app. Every voice conversation has a written transcript with references alongside it. |
| SE004 | OpenEvidence | Deep Consult | User Guide | OpenEvidence | An advanced AI agent that generates PhD-level research reports to help you ramp up on a body of knowledge or answer the most complex questions. |
| SE005 | OpenEvidence | OpenEvidence is now available for iOS and Android | OpenEvidence | The OpenEvidence app is now available for download. The app provides the same high-quality, cited answers as on the web, now available as a native application. |
| SE006 | Google Play | OpenEvidence - Apps on Google Play | OpenEvidence is used daily, on average, by over 40% of physicians in the United States, spanning more than 10,000 hospitals and medical centers nationwide. |
| SE007 | OpenEvidence | OpenEvidence and NEJM Group, publisher of the New England Journal of Medicine, sign content agreement: Clinicians using OpenEvidence benefit from content sourced from NEJM Group journals | OpenEvidence | All published content and multimedia from 1990 forward from NEJM, NEJM Evidence, NEJM AI, NEJM Catalyst, and NEJM Journal Watch will be provided to OpenEvidence to inform answers delivered on the OpenEvidence platform. |
| SE008 | JAMA Network | OpenEvidence and the JAMA Network sign strategic content agreement - For The Media - JAMA Network | All published content from JAMA, JAMA Network Open, and the 11 JAMA specialty journals — including full text and multimedia — will inform answers delivered on the OpenEvidence platform. |
| SE009 | OpenEvidence | OpenEvidence and the JAMA Network sign strategic content agreement | OpenEvidence | |
| SE010 | NCCN | NewsDetails | All NCCN content surfaced on OpenEvidence strictly adheres to NCCN’s established QA review and links to the source document for additional information. |
| SE011 | OpenEvidence | NCCN and OpenEvidence Collaborate to Bring Clinical Oncology Guidelines to Medical AI | OpenEvidence | OpenEvidence will integrate content from the NCCN Clinical Practice Guidelines in Oncology and related content from JNCCN into its clinician-facing AI platform. |
| SE012 | Wiley | Wiley and OpenEvidence Partner to Deliver Trusted Research to Physicians at the Point of Care | OpenEvidence was built on a principle its founders call gold in, gold out: specialized models trained on peer-reviewed literature, not the open internet, with every answer grounded in sources a physician can drill into and verify. |
| SE013 | OpenEvidence | Wiley and OpenEvidence Partner to Deliver Trusted Research to Physicians at the Point of Care | OpenEvidence | |
| SE014 | OpenEvidence | OpenEvidence is now HIPAA compliant: Health care professionals using OpenEvidence can securely upload PHI | OpenEvidence | To protect sensitive information, user conversations on OpenEvidence are private by default. Users have full control over access to their conversations by using the Share button. |
| SE015 | OpenEvidence | Mount Sinai Health System Collaborates with OpenEvidence to Provide Evidence-Based Knowledge Within Electronic Medical Record | OpenEvidence | The collaboration marks the first enterprise-scale OpenEvidence deployment to extend access across the full clinical care team — including physicians, registered nurses, and pharmacists. |
| SE016 | OpenEvidence | OpenEvidence Partners with Cedars-Sinai to Create Patient-Aware Clinical Intelligence With Agentic Clinical AI | OpenEvidence | By embedding OpenEvidence within the clinician workflow, physicians can ask complex clinical questions in natural language and receive answers dynamically tailored to the patient in front of them. |
| SE017 | OpenEvidence | OpenEvidence Introduces DotFlows: Flexible Natural Language Customization for Every Clinician | OpenEvidence | Dotflows are reusable natural language prompts that customize how OpenEvidence responds. |
| SE018 | Medical Economics | Clinical AI platform to be trained on 30 years of NEJM archives | According to the company, 40,000 new verified doctors register each month. The platform is widely used during clinical hours, with 75% of physicians leveraging it at the point of care to inform decision-making. |
| SE019 | HIT Consultant | Mount Sinai Integrates OpenEvidence AI Platform into Epic EHR for Enterprise-Wide Clinical Decision Support | Embedding this capability directly inside Epic enables Mount Sinai to solve the last mile problem of clinical AI. |
| SE020 | Fierce Healthcare | OpenEvidence launches hands-free voice AI feature, expands hospital footprint with Cedars-Sinai tie-up | There are now 860,000 licensed-verified U.S. clinicians, including nurses, nurse practitioners and physician assistants, using OpenEvidence, with more than 650,000 licensed-verified U.S. physicians. |
| SE021 | PRNewswire | OpenEvidence Achieves Historic Milestone: 1 Million Clinical Consultations between Verified Doctors and an Artificial Intelligence System in a Single Day | On March 10, 2026, for the first time in history, one million clinical consultations were conducted between NPI verified physicians and an AI system, OpenEvidence, within a single 24-hour period. |
| SE022 | Built In | OpenEvidence Careers, Perks + Culture | |
| SE023 | Harvard FAS Mignone Center for Career Success | OpenEvidence – Software Engineer, Fullsatck | We are a $12B company with a 30-person engineering team. All full-time roles on our engineering team are in-person 5 days a week in SF or Miami. |
| SE024 | arXiv | Medical Hallucinations in Foundation Models and Their Impact on Healthcare | A global survey of clinicians validated real-world impact: 91.8% had encountered medical hallucinations, and 84.7% considered them capable of causing patient harm. |
| SE025 | Open Research Europe | Comparison of explainability methods for hallucination analysis in LLMs | Even RAG systems remain vulnerable to hallucinations. These can occur when retrieval fails or when the model misinterprets retrieved content. |
| SU001 | OpenEvidence | OpenEvidence | OpenEvidence is used daily, on average, by over 40% of physicians in the United States, spanning more than 10,000 hospitals and medical centers nationwide. |
| SU002 | PR Newswire | OpenEvidence, the Fastest-Growing Application for Physicians in History, Announces $210 Million Round at $3.5 Billion Valuation | |
| SU003 | PR Newswire | OpenEvidence Achieves Historic Milestone: 1 Million Clinical Consultations between Verified Doctors and an Artificial Intelligence System in a Single Day | On March 10, 2026, for the first time in history, one million clinical consultations were conducted between NPI verified physicians and an AI system, OpenEvidence, within a single 24-hour period. |
| SU004 | OpenEvidence | OpenEvidence Partners with Cedars-Sinai to Create Patient-Aware Clinical Intelligence With Agentic Clinical AI | |
| SU005 | NBC News | Most U.S. doctors are quietly using this AI tool. Few patients know about it. | Yet with OpenEvidence's skyrocketing popularity, some experts worry about potential hallucinations or incomplete answers, a lack of rigorous scientific studies on the tool's patient impact, and the potential for doctors' critical thinking and evaluation skills to erode with increased OpenEvidence use and dependence. |
| SU006 | NBC New York | Most U.S. doctors are quietly using this AI tool. Few patients know about it | |
| SU007 | Sacra | OpenEvidence revenue, valuation & funding | The company monetizes through pharmaceutical and medical device advertisements, achieving CPMs that can range from $70 to $1,000+, compared to $5-15 for typical social media platforms. |
| SU008 | Fierce Healthcare | OpenEvidence clinches $250M series D as AI platform sees explosive growth with doctors | |
| SU009 | Fierce Healthcare | JPM26: OpenEvidence makes the case for AI-powered 'medical super-intelligence' | |
| SU010 | American College of Obstetricians and Gynecologists | ACOG and OpenEvidence Announce Strategic Collaboration to Advance Ob-Gyn Health Care | Women’s health is one of the most frequently searched topics on OpenEvidence. |
| SU011 | Wiley | Wiley and OpenEvidence Partner to Deliver Trusted Research to Physicians at the Point of Care | |
| SU012 | Business Wire | Wiley and OpenEvidence Partner to Deliver Trusted Research to Physicians at the Point of Care | |
| SU013 | Cedars-Sinai | Cedars-Sinai and OpenEvidence Partner on AI Health Solution | Patient information from the electronic health record will be used only to support care decisions for individual patients and will not be stored by OpenEvidence or used for any other purpose. |
| SU014 | HIT Consultant | Cedars-Sinai Deploys OpenEvidence Enterprise Platform to Drive Precision Clinical Decision-Making | |
| SU015 | Healthcare IT News | Mount Sinai to integrate OpenEvidence AI enterprise-wide | |
| SU016 | Becker's Hospital Review | Mount Sinai inks 1st enterprise deal with OpenEvidence | |
| SU017 | HLTH | Mount Sinai and OpenEvidence Integrate AI Clinical Decision Support into Epic EHR | |
| SU018 | Apple App Store | OpenEvidence App - App Store | 4.9 out of 5 — 10K Ratings. |
| SU019 | Apple App Store | OpenEvidence - Ratings & Reviews - App Store | |
| SU020 | Google Play | OpenEvidence - Apps on Google Play | |
| SU021 | AppBrain | OpenEvidence: Free Medical - Stats, Ratings & Downloads | AppBrain | OpenEvidence is rated 4.80 out of 5 stars, based on 3.7 thousand ratings. |
| SU022 | Trustpilot (archived) | Openevidence is rated 'Bad' with 1.6 / 5 on Trustpilot | OpenEvidence references outdated information and advises doctors to recommend dangerous 'treatments' to their patients. |
| SU023 | AusDoc | JAMA and NEJM back $1.5 billion AI ‘that cannot hallucinate’ to answer doctors’ medical questions | However, an expert says basing answers purely on the two journal groups raises new questions about AI bias. |
| SU024 | Newswise | OpenEvidence Achieves Historic Milestone: 1 Million Clinical Consultations between Verified Doctors and an Artificial Intelligence System in a Single Day | |
| SU025 | STAT | OpenEvidence makes its pitch to hospitals. 'We’re not crazy monsters' | The company is facing competitive pressures and questions about whether its ad-based business model can continue to propel the company forward. |
| SR001 | Food and Drug Administration | Clinical Decision Support Software - Guidance | Section 3060(a) of the 21st Century Cures Act ... exclude[s] certain decision support software from the definition of device ... and this guidance clarifies the types of CDS software functions that are excluded. |
| SR002 | Food and Drug Administration | Clinical Decision Support Software - Guidance for Industry and Food and Drug Administration Staff | Intended for the purpose of enabling an HCP to independently review the basis for the recommendations ... so that it is not the intent that the HCP rely primarily on any of such recommendations to make a clinical diagnosis or treatment decision. |
| SR003 | Food and Drug Administration | Town Hall – Clinical Decision Support Software, Final Guidance | |
| SR006 | U.S. Department of Health and Human Services | Use of Online Tracking Technologies by HIPAA Covered Entities and Business Associates | Disclosures of PHI to tracking technology vendors for marketing purposes, without individuals’ HIPAA-compliant authorizations, would constitute impermissible disclosures. |
| SR008 | European Commission | AI Act | High-risk AI systems are subject to strict obligations before they can be put on the market: adequate risk assessment and mitigation systems ... logging ... documentation ... human oversight ... robustness, cybersecurity and accuracy. |
| SR010 | OpenEvidence | Privacy Policy | OpenEvidence | We may collect NPI numbers from third parties and match those numbers to Personal Information in our system to provide you with content and advertisements tailored to your individual interests and needs. |
| SR011 | OpenEvidence | Terms of Use | OpenEvidence | The information and tools that we make available through the Services are provided for educational and informational purposes only ... in no way intended to serve as a diagnostic service ... or otherwise substitute for the clinical judgment of a qualified healthcare professional. |
| SR012 | OpenEvidence | Security and Compliance | OpenEvidence | OpenEvidence has achieved SOC 2 Type II certification ... and is primarily hosted on Google Cloud Platform and Vercel. |
| SR013 | OpenEvidence | OpenEvidence is now HIPAA compliant: Health care professionals using OpenEvidence can securely upload PHI | OpenEvidence | OpenEvidence is now HIPAA compliant: Health care professionals using OpenEvidence can securely upload PHI. |
| SR015 | OpenEvidence | Voice Mode | User Guide | OpenEvidence | Voice Mode enables hands-free, real-time medical conversations ... Every voice conversation has a written transcript with references alongside it. |
| SR021 | OpenEvidence | Mayo Clinic study: OpenEvidence performs comparably to physician CDM in common clinical scenarios | OpenEvidence | Mayo Clinic study: OpenEvidence performs comparably to physician CDM in common clinical scenarios. |
| SR022 | Wiley | Wiley and OpenEvidence Partner to Deliver Trusted Research to Physicians at the Point of Care | Wiley will license a comprehensive portfolio of scientific and medical content to OpenEvidence ... including the Cochrane Database of Systematic Reviews and Cochrane Clinical Answers ... and over 400 medical journals and books. |
| SR023 | Cochrane | Cochrane evidence to inform OpenEvidence users | AI is rapidly changing the ways that people use and access evidence, and we hope that this partnership helps to ensure that outputs from AI-powered tools are based on the best possible evidence. |
| SR024 | CNBC | OpenEvidence, the 'ChatGPT for doctors,' doubles valuation to $12 billion | OpenEvidence said it topped $100 million in annualized revenue last year ... OpenEvidence was one of the first AI startups to rely on advertising for revenue, which Nadler said allows faster adoption and wider use versus a paid subscription model. |
| SR025 | Fierce Healthcare | JPM26: OpenEvidence makes the case for AI-powered 'medical super-intelligence' | OpenEvidence is now actively used daily, on average, by more than 40% of physicians in the U.S. ... Nadler said the goal is medical super-intelligence. |
| SR026 | STAT | OpenEvidence makes its pitch to hospitals. 'We’re not crazy monsters’ | The company is facing competitive pressures and questions about whether its ad-based business model can continue to propel the company forward. |
| SR027 | Open Pharma | OpenEvidence at ISMPP Annual: what we heard, what we’re asking and what comes next | OpenEvidence provides evidence-informed answers ... and is free to access, supported by advertising and venture capital backing ... without access to every journal ... how can we be confident the generated summaries are truly representative of the full evidence base? |
| SR028 | Krafty Librarian | Smart Medicine or Smart Marketing? – Krafty Librarian | Until more voices from the information side of healthcare are included and kick the AI’s tires, it’s hard to fully know if OpenEvidence is smart medicine ... or just smart marketing. |
| SR029 | Stanford Medicine | Clinical AI Has Boomed. A New Stanford-Harvard State of Clinical AI Report Shows What Holds Up in Practice. | Many claims of physician-level or “superhuman” performance rely on narrow benchmarks or controlled evaluations that do not reflect the uncertainty, incomplete information, and workflow complexity of everyday care. |
| SR030 | GLACIS | When AI Hallucinations Become Malpractice Risk | Without evidence-grade documentation, these questions become much harder to answer. In litigation or internal investigation, that missing context can become a serious risk factor. |
| SR031 | Patient Safety & Quality Healthcare | AI in Healthcare: Addressing the Reality of Hallucinations - Patient Safety & Quality Healthcare | A JAMA Network Open paper in 2023 found AI-generated discharge summaries contained incomplete or misleading information about 18% of cases. |
| SR033 | HIMSS | Operationalizing AI: A Strategic Framework for Safe Deployment in Healthcare | AI should be held to the same level of rigor as any clinical intervention. |
| SR034 | OpenAI | Introducing OpenAI for Healthcare | OpenAI for Healthcare ... includes evidence retrieval with transparent citations ... and a Business Associate Agreement (BAA) with OpenAI to support HIPAA-compliant use. |
| SR035 | Anthropic | Advancing Claude in healthcare and the life sciences | We’re introducing Claude for Healthcare ... allowing healthcare providers, payers, and health tech companies to use Claude for medical purposes through HIPAA-ready products. |
| SR036 | Doximity | Introducing the Doximity Clinical AI Suite | More than 85% of U.S. physicians already practice on Doximity ... Ask is the only AI platform that includes free, full PDF access to top medical journals. |
| SR037 | Wolters Kluwer | Wolters Kluwer’s New UpToDate Expert AI Provides Clinicians and Health Systems with the Fast, Reliable GenAI Clinical Decision Support They Need | UpToDate Expert AI is the next evolution of UpToDate ... used at thousands of hospitals around the world ... with single-click access to the assumptions driving the answers and the source of the answer. |
| SR038 | Google Cloud | Introducing MedLM for the healthcare industry | Google Cloud Blog | MedLM is now available to Google Cloud customers in the United States ... informed by healthcare customer needs, such as answering a healthcare provider’s medical questions and drafting summaries. |
| SR039 | Google Cloud | Google Cloud Enhances Vertex AI Search for Healthcare with Multimodal AI | Vertex AI Search for healthcare's new, multimodal search capabilities can help organizations give doctors, nurses, and others a more comprehensive view of patient health. |
| SR040 | CourtListener | OpenEvidence Inc. v. Doximity, Inc., 1:25-cv-11802 - CourtListener.com | COMPLAINT ... filed by OpenEvidence Inc. against Doximity, Inc. on Jun 20, 2025. |
| SR041 | Bloomberg Law | Medical AI Firm Says Competitor Hacked Prompts to Steal Secrets | Doximity Inc. executives impersonated physicians and conducted a months-long cyberattack using malicious inputs to extract trade secrets from competitor OpenEvidence Inc.'s artificial intelligence model. |
| SR042 | Microsoft | Microsoft extends AI advancements in Dragon Copilot to nurses and partners to enhance patient care | Through partnerships with Elsevier, OpenEvidence and Wolters Kluwer UpToDate, Dragon Copilot will provide access to curated clinical content directly within the workflow. |
| SR043 | Mount Sinai Health System | Mount Sinai Health System to Roll Out Microsoft Dragon Copilot | Mount Sinai's adoption of Dragon Copilot, after a multi-vendor evaluation, represents a transformative step forward ... The rollout has begun with select departments with plans to expand system-wide in 2026. |
| SV001 | PR Newswire | OpenEvidence Achieves $1 Billion Valuation in Sequoia-led Round and Announces Content Partnership with the New England Journal of Medicine | OpenEvidence Achieves $1 Billion Valuation in Sequoia-led Round. |
| SV002 | CNBC | AI health-care startup OpenEvidence raises funding from Sequoia at $1 billion valuation | OpenEvidence offers its chatbot for free and makes money off of advertising. |
| SV003 | OpenEvidence | OpenEvidence, the Fastest-Growing Application for Physicians in History, Announces $210 Million Round at $3.5 Billion Valuation | |
| SV004 | PR Newswire | OpenEvidence, the Fastest-Growing Application for Physicians in History, Announces $210 Million Round at $3.5 Billion Valuation | |
| SV005 | Business Wire | OpenEvidence Raises $250 Million to Build Medical Superintelligence for Doctors | OpenEvidence ... closed a Series D round of funding, valuing the company at $12 billion. |
| SV006 | Fierce Healthcare | OpenEvidence clinches $250M series D as AI platform sees explosive growth with doctors | |
| SV007 | CNBC | OpenEvidence, the "ChatGPT for doctors," doubles valuation to $12 billion | OpenEvidence raised money at a valuation of $12 billion in a round led by Thrive and DST. |
| SV008 | Becker's Hospital Review | At $12B, OpenEvidence becomes most valuable healthcare AI company | Founded in 2022, the startup has surpassed $100 million in annual revenue. |
| SV009 | Sacra | OpenEvidence revenue, valuation & funding | Sacra estimates OpenEvidence hit $150 million in annualized revenue in 2025 ... with reported 90% gross margins. |
| SV010 | U.S. Food and Drug Administration | Clinical Decision Support Software - Guidance | This guidance clarifies FDA’s thinking on the types of clinical decision support software functions that are excluded from the definition of device. |
| SV011 | American Medical Association | AMA issues new principles for AI development, deployment & use | |
| SV012 | Business Wire | Hippocratic AI Raises $126 Million in Series C at $3.5 Billion Valuation | |
| SV013 | CompaniesMarketCap | Doximity (DOCS) - Market capitalization | |
| SV014 | CompaniesMarketCap | Doximity (DOCS) - Revenue | |
| SV015 | CompaniesMarketCap | Tempus AI (TEM) - Market capitalization | |
| SV016 | CompaniesMarketCap | Tempus AI (TEM) - Revenue | |
| SV017 | CompaniesMarketCap | Veeva Systems (VEEV) - Market capitalization | |
| SV018 | CompaniesMarketCap | Veeva Systems (VEEV) - Revenue | |
| SV019 | CompaniesMarketCap | Health Catalyst (HCAT) - Market capitalization | |
| SV020 | CompaniesMarketCap | Health Catalyst (HCAT) - Revenue | |
| SV021 | CompaniesMarketCap | Phreesia (PHR) - Market capitalization | |
| SV022 | CompaniesMarketCap | Phreesia (PHR) - Revenue | |
| SV023 | Wolters Kluwer | UpToDate: Trusted, evidence-based solutions for modern healthcare | |
| SV024 | OpenEvidence | About | OpenEvidence | To date, we have supported over 200 million AI-powered clinical consultations from U.S. doctors and other frontline clinicians. |
| SV025 | MobiHealthNews | OpenEvidence scores $250M, doubles valuation to $12B | In October, OpenEvidence scored $200 million in a Series C round, boosting its valuation to $6 billion. |
| SV026 | Healthcare Chief | OpenEvidence Secures $250M Series D Amid Rapid Physician Adoption | |
| SV027 | HLTH | OpenEvidence Doubles Valuation to $12bn as Physician Adoption Accelerates | |
| SV028 | SEC EDGAR | Doximity SEC company search results | |
| SV029 | SEC EDGAR | Veeva Systems SEC company search results | |
| SV030 | SEC EDGAR | Health Catalyst SEC company search results |