Back to Blog
Your Doctor Probably Asks an AI for Answers. What 271 Million OpenEvidence Questions Show

Your Doctor Probably Asks an AI for Answers. What 271 Million OpenEvidence Questions Show


U.S. clinicians asked OpenEvidence 271 million questions in two and a half years. A study published October 15 in JAMA counted 952,008 users, including 556,200 physicians, between January 2024 and June 2026.[1]

Your doctor may already be using artificial intelligence (AI) to look up medical evidence. The study cannot tell you whether your particular doctor does, or whether an AI answer influenced your last appointment.

It does give patients a reason to ask how their doctor checks the information behind a recommendation.

Our earlier OpenEvidence valuation article covered the business. Here, the question is how a tool used at this scale fits into your care.

What OpenEvidence Does for Your Doctor

OpenEvidence summarizes medical research in response to clinicians' questions. It generates answers with citations to medical journals and professional guidelines. The study describes free access for healthcare professionals verified through a National Provider Identifier, the identification number used for U.S. healthcare providers.[1]

Its terms describe an educational and informational service for professionals. They explicitly preserve the clinician's independent judgment and say the service is not intended for patients or consumers.[2]

For you, that means an AI answer should be one input your doctor evaluates. You should still be able to discuss the reasoning behind a prescription, test, or referral.

This serves a different purpose from an AI scribe that records a visit and drafts the note. Asking which tool is involved helps you understand what it is doing.

Who Built OpenEvidence

Daniel Nadler and Zachary Ziegler founded OpenEvidence in November 2021, according to the company's court filing, which also identifies Nadler as CEO.[3] Ziegler is chief technology officer. He previously studied machine learning as a Harvard doctoral student, and his speaker biography describes OpenEvidence's launch through Mayo Clinic Platform Accelerate.[4]

Nadler previously founded Kensho, which developed AI and analytics for financial institutions. S&P Global announced an agreement to acquire that company in March 2018.[5] OpenEvidence brought the founders' work into medical information, where a generated summary can become part of a clinician's research process.

What the Usage Numbers Show

The study documents repeated use, with a substantial share of physician questions coming from doctors early in their careers.[1][6]

Finding

What it tells patients

Monthly active users reached 703,265 in June 2026

Hundreds of thousands of clinicians were using the tool in a single month.[6]

About 90% of clinicians active in one month returned the next in 2026

Use often continued beyond an initial trial.[6]

Physicians within 10 years of medical school graduation submitted 61.8% of physician questions, while representing 24.3% of the physician workforce

Early-career physicians contributed disproportionately to question volume.[6]

Treatment and management accounted for 33.1% of classified physician questions; diagnosis accounted for 13.9%

Doctors frequently sought help with care decisions beyond identifying a condition.[6]

The retention figure describes returning activity. It does not mean 90% of doctors agreed with the answers or that 90% of patients benefited.

Career stage also needs careful interpretation. These figures cannot establish whether a younger or older doctor makes better decisions with AI.

What 271 Million Questions Cannot Establish

The question count is not a patient count. Follow-up questions counted separately. Researchers also could not distinguish questions asked during appointments from preparation or general learning.[1]

The content analysis used AI to classify a sample of 38.8 million questions, with classifications checked against clinicians' labels. It covered one platform's main question interface.[1] Those boundaries matter when interpreting percentages as a picture of how doctors use AI generally.

JAMA's accompanying editor's note identifies a further gap: researchers did not evaluate answer quality, clinicians' assessments of usefulness or accuracy, or effects on patient care.[7]

That leaves practical questions unanswered. Did a doctor change a treatment after reading a response? Did the change help? Was an important exception missed? Frequent use cannot answer those questions.

The potential benefit is easier access to relevant evidence. A doctor could use a summary to identify research worth reviewing, then discuss how it fits your situation. JAMA's editors also caution that a citation alone does not establish that a generated answer interprets the evidence correctly or applies to an individual patient.[7]

For patients, the useful distinction is between finding information and deciding what to do with it. Your doctor's explanation should connect the evidence to your history, preferences, and treatment options.

The Company's Role in the Research

OpenEvidence supplied the data and contributed to analysis. Three authors disclosed company employment during the study, including its chief medical officer.[1][6]

That involvement gave researchers access to detailed usage records. It also makes the company's role relevant when weighing the findings.

JAMA Network disclosed that it licenses journal content to OpenEvidence. The editor's note says the study underwent independent editor assignment, peer review, and editorial decision-making.[7] Patients can recognize both the commercial relationships and the journal's stated review process without treating publication as proof of clinical benefit.

Bloomberg Law separately reported that Doximity settled OpenEvidence's lawsuit alleging impersonation of doctors to extract AI trade secrets. The joint notice did not disclose settlement details.[8] That dispute concerned access and proprietary technology; it provides no measure of whether patients received accurate medical advice.

FAQ

Does the study show that OpenEvidence improves care?

No. It describes use and question patterns. It did not measure whether answers improved clinical decisions or patient outcomes.[7]

Should I choose a doctor based on whether they use AI?

Ask how the doctor evaluates information and explains decisions. When choosing a concierge or direct primary care (DPC) doctor, include that conversation alongside your questions about access, communication, and follow-up.

What to Ask Your Doctor About AI in Your Care

You do not need to examine every search your doctor makes. Start with the decision you are discussing, then ask:

  1. Did you use AI to help answer a question about my care? Ask which tool and what question it helped investigate.

  2. How did you check the answer? Ask whether the doctor reviewed the underlying research or guideline supporting the recommendation.

  3. How does that evidence apply to me? Invite an explanation of the benefits, risks, alternatives, and uncertainties that matter in your situation.

  4. What information about me went into the tool? Ask whether identifiable records were entered and what the practice's privacy policy allows. Using a research assistant does not by itself tell you what information was shared.

  5. What happens if the answer conflicts with your judgment? Ask how the doctor resolves disagreement and whether another opinion would help.

You can find physician-led concierge and DPC practices on NextMD and bring these questions to an introductory conversation. A useful place to start is: “Did AI contribute to this recommendation, and how did you decide it was right for me?”

Sources

  1. Patel, V. R., Finlayson, S. G., Liu-Hass, M., et al. (2026, October 15). Clinician Use of an Artificial Intelligence–Enabled Clinical Information Platform. JAMA. doi:10.1001/jama.2026.21868. Read the original OpenEvidence usage study.

  2. OpenEvidence. (Accessed October 17, 2026). Terms of Use. Read the professional-use and clinical-judgment terms.

  3. OpenEvidence, Inc. (2025, June 20). Complaint in OpenEvidence, Inc. v. Doximity, Inc. et al., paragraphs 34 and 49, PDF pages 15 and 21. Read the company's account of its founding.

  4. Symposium on Artificial Intelligence for Learning Health Systems. (2025). Zachary Ziegler. Read the cofounder's background and company origin.

  5. S&P Global. (2018, March 6). S&P Global to Acquire Kensho; Bolsters Core Capabilities in Artificial Intelligence, Natural Language Processing and Data Analytics. Read the Kensho acquisition announcement.

  6. Bruce, G. (2026, October 16). OpenEvidence drew 271M queries from 952,000 clinicians in 2.5 years. Becker's Hospital Review. Read Becker's report on the JAMA findings.

  7. Khera, R., Bibbins-Domingo, K., and Ross, J. S. (2026, October 15). Engaging With Medical Evidence in the Age of AI. JAMA. doi:10.1001/jama.2026.21894. Read the editor's note on evidence, verification, and patient outcomes.

  8. Jahner, K. (2026, October 15; corrected October 16). Doximity Settles OpenEvidence AI Prompt-Hack Dispute (Correct). Bloomberg Law. Read the settlement report.


Related Posts