OpenEvidence is a Florida-based medical search and answer service for physicians and other licensed clinicians, led by founder and CEO Daniel Nadler. Verified clinicians type a clinical question and receive a written answer with references to published literature, free of charge to verified users, with advertising covering costs according to the company. Most of what is publicly said about its accuracy and reach comes from the company itself; independent studies so far are few, small, and mixed.
Key takeaways
- The service is built for clinicians, and its terms of use say it is not intended for patients and does not replace a clinician's judgment.
- Answers draw on licensed content that, per the company, includes NEJM Group journals, the JAMA Network, and NCCN oncology guidelines.
- Canadian physicians can verify with a MINC number under a partnership announced on September 11, 2026.
- A 2026 audit of 4,979 citations in npj Health Systems found no fabricated reference, but it did not test whether the references support each answer.
- Usage figures such as "more than 50% of physicians in Canada" are company statements that this page could not confirm independently.
What the service does and who may use it
OpenEvidence describes itself as a medical information platform built for clinicians. A user enters a question or case details and gets a synthesized answer with citations that can be opened and checked. The company launched the product in 2023, according to its February 2025 release, and added features since: a deeper research mode called DeepConsult in July 2025 and OpenEvidence Visits in August 2025, which transcribes a patient encounter and drafts notes with evidence attached. In September 2026 it announced a model family named Osler, Sackett, and Snow that trade speed for search depth, along with a fourth model, Darwin, available only by application as a research preview.
The network terms of use, last updated September 17, 2026, state that the services are for physicians and other health professionals acting in a professional capacity, not for patients or other consumers. They describe the information as educational and informational, say the platform is not a diagnostic service, and say it does not recommend a particular product or therapy or substitute for clinical judgment. The terms bar use from the European Union, the United Kingdom, and the European Economic Area. Accounts are verified, and the company says access is free to verified U.S. clinicians and, since September 2026, to Canadian physicians through a separate route.
Where the content comes from and how it is paid for
Answers depend on what the company is allowed to read. On February 19, 2025, OpenEvidence and NEJM Group announced a multi-year agreement giving the platform published content from 1990 onward from NEJM, NEJM Evidence, NEJM AI, NEJM Catalyst, and NEJM Journal Watch. A July 15, 2025 company release lists the American Medical Association, NEJM, JAMA, and the eleven JAMA specialty journals as content partners, and a November 5, 2025 release with the National Comprehensive Cancer Network describes a license for NCCN oncology guidelines. The company's site also names Nature and Cochrane. The agreements show what is licensed, not how often any source is used in a given answer.
The company says advertising pays for the free service. Its advertising policy, dated August 12, 2024, says a limited number of controlled ads are shown, that ads are marked and kept separate from answers, and that advertisers cannot influence answer content. That is a statement of policy by the company. No outside audit of ad separation was found.
Access for Canadian physicians
On September 11, 2026, OpenEvidence and MINC#NIMC announced that Canadian physicians can verify by entering their Medical Identification Number for Canada at sign-up or linking it to an existing account, with access staying free. MINC#NIMC is a not-for-profit set up by the Federation of Medical Regulatory Authorities of Canada and the Medical Council of Canada. The announcement says more than 95 percent of Canadian physicians hold a MINC and that the platform is already used by over 50 percent of physicians in Canada. The second figure is the company's, and no independent count was located. Whether a given province, hospital, or regulator permits use in practice is a separate question for the clinician's employer or college.
What independent studies have measured
Four items are summarized below. They test different things, and none measures patient outcomes.
| Study | What was tested | Result | Main limits |
|---|---|---|---|
| Low et al., Digital Health, June 2025 | 50 clinical questions answered by five systems, graded by nine physicians | 24 percent of OpenEvidence answers rated relevant and evidence-based, against 58 percent for ChatRWD and 2 to 10 percent for general chatbots | Funded by the maker of ChatRWD; questions limited to those answerable by cohort studies; raters not blinded |
| Jagarapu et al., medRxiv preprint, December 2025 | 100 subspecialty multiple-choice questions from MedXpertQA | 31 percent average accuracy for quick search, 39.5 percent for Deep Consult, not significantly different | Preprint, not peer reviewed; versions from mid-2025; small pilot |
| Khan et al., Journal of Orthopaedic Experience & Innovation, 2026 | 111 text-only questions from the 2014 orthopaedic in-training exam | 84.7 percent correct, equal to Gemini 2.5 Pro and not significantly different from ChatGPT-5 | One exam year; models may have seen the public exam |
| Lotan et al., npj Health Systems, September 2, 2026 | All 4,979 citations returned for 150 prompts across five specialties | No reference confirmed as fabricated; three abstracts had attribution errors | Checked that references exist, not that they support the claims |
The Low study lists a conflict worth knowing: OpenEvidence supplied platform access and technical support during manuscript preparation, and the funder develops a competing system. The spread between 24 percent on real-world cohort questions and 85 percent on an exam is a reminder that accuracy depends on the question type.
Company benchmark claims
The September 3, 2026 announcement says Darwin scored 72.8 percent on MedXpertQA, 82.7 percent on HealthBench Professional, 87.2 percent on NOHARM, and 100 percent on MedQA. These are company-reported results for a model in restricted preview, published by the company, and the independent studies above tested earlier versions, so the figures cannot be compared. Multiple-choice benchmarks also measure recall of exam-style facts, which differs from deciding what to do for one patient.
Funding and ownership, as far as it is confirmed
A July 15, 2025 release reports a USD 210 million Series B at a USD 3.5 billion valuation, co-led by Google Ventures and Kleiner Perkins. A January 22, 2026 release describes a USD 250 million Series D at a USD 12 billion valuation, co-led by Thrive Capital and DST Global, bringing total funding to nearly USD 700 million. Press reports in late September 2026 described a further round at USD 15 billion based on unnamed sources; no confirming company announcement was found, so the figure is not treated as established.
Questions a clinician or clinic can ask
- Does the answer cite sources that can be opened, and do they say what the answer claims?
- Is the question one that exam-style testing reflects, or a rare, multi-factor case?
- Which version or mode produced the answer, and how long ago did the underlying evidence change?
- Where does patient information go when case details are entered, and which privacy rules apply?
- Who is accountable for the final decision? The terms place it with the clinician.
This article gives general information about a software product and published studies. It is not medical advice, does not recommend or discourage the use of any tool, and does not support decisions about diagnosis or treatment. Patients with health questions should speak with a physician or pharmacist.
FAQ
Is OpenEvidence free?
Company materials say it is free for verified clinicians, including Canadian physicians who verify with a MINC. The company says advertising pays for the service.
Can patients use it?
The terms of use say the services are not intended for patients or other consumers. The company announced a "Patient Take-Homes" feature on August 19, 2026, but its design and audience were not reviewed here.
Does it make up references?
One 2026 peer-reviewed audit found no fabricated reference among 4,979 citations. That finding concerns whether references exist, not whether they support the statements made.
Has it been shown to improve patient care?
No study of patient outcomes was located. Published work so far tests answer quality, exam performance, and citation accuracy.
Related reading
- How vector search powers retrieval-augmented AI
- What the evidence says about AI and supplements
- Pew's measurements of AI summaries in search
- How AI systems solve problems
More in AI Tools and Industry.
Last verified: October 2026
Sources
- OpenEvidence and NEJM Group sign content agreement, OpenEvidence, February 19, 2025
- OpenEvidence announces USD 210 million round at USD 3.5 billion valuation, OpenEvidence, July 15, 2025
- OpenEvidence Launches Visits, OpenEvidence, August 21, 2025
- NCCN and OpenEvidence collaborate, NCCN and OpenEvidence, November 5, 2025
- OpenEvidence raises USD 250 million Series D, coverage on Cooley LLP, January 22, 2026
- Introducing the OpenEvidence Model Family, OpenEvidence, September 3, 2026
- OpenEvidence and MINC#NIMC partner, OpenEvidence, September 11, 2026
- Network Terms of Use, last updated September 17, 2026, and Advertising Policy, last updated August 12, 2024, OpenEvidence Inc.
- The Medical Identification Number for Canada Corporation, Medical Council of Canada
- Low YS et al., Answering real-world clinical questions using large language model, retrieval-augmented generation, and agentic systems, Digital Health, 2025
- Jagarapu J et al., The accuracy and repeatability of OpenEvidence on complex medical subspecialty scenarios: a pilot study, medRxiv preprint, December 4, 2025
- Khan AA et al., Evidence on Demand: OpenEvidence versus Established LLMs on the Orthopaedic In-Training Examination, Journal of Orthopaedic Experience & Innovation, 2026
- Lotan T et al., Reference quality of OpenEvidence across five medical specialties, npj Health Systems, September 2, 2026