Nurse Practitioners and PAs Outquery Doctors on OpenEvidence, JAMA Finds

Part of Nurse.org’s Nursing AI Watch, our ongoing investigation into how artificial intelligence is reshaping nursing practice.

Nurse practitioners and physician assistants are the most intensive users of OpenEvidence, the AI clinical reference tool often described as “ChatGPT for clinicians,” according to a study published October 15, 2026 in JAMA. The journal published an accompanying editor’s note, “Engaging With Medical Evidence in the Age of AI,” signaling that JAMA leadership views the findings as a significant moment for how clinical knowledge is accessed.

Researchers at Mass General Brigham and Harvard Medical School analyzed query logs from 952,008 US clinicians with verified National Provider Identifiers between January 2024 and June 2026. Over that 30-month window, those clinicians submitted 271 million queries. Nurse practitioners made up 171,034 of the users, more than double the physician assistant count of 93,241, making NPs the single largest nonphysician group on the platform.

Want to see more Nurse.org articles in your Google results? Add us as a preferred source.

NPs and PAs averaged 44.5 queries per active user-month, compared with 37.1 for physicians. The pattern held across every clinician category the researchers examined.

Query intensity by clinician type (queries per active user-month)

Source: Patel et al., JAMA, October 15, 2026

Monthly active users on the platform rose from 355 in January 2024 to 703,265 by June 2026. Average monthly queries per active user climbed from 18 to 39.5. Month-over-month retention improved from roughly 70% to 90%. The authors noted that these counts measure queries submitted through the platform’s main question interface only, which makes the totals a lower bound on actual use.

OpenEvidence growth, January 2024 to June 2026

Source: Patel et al., JAMA, October 15, 2026

Physicians made up 556,200 of the verified users, with NPs, PAs, pharmacists, and other clinicians accounting for the rest. Early-career physicians, those within a decade of medical school graduation, represented 24.3% of the physician workforce in the dataset but generated 61.8% of physician queries.

The generational pattern was not limited to physicians. Across all clinician types combined, those with fewer than 10 years in practice averaged 43.4 queries per user-month, compared with 29.9 for those with 20 or more years. The pooled figure does not establish identical rates within every profession, but it points to a broader experience-related pattern in use.

To classify what clinicians were asking about, the researchers used an AI model to label a stratified random sample of 38.8 million queries by purpose and clinical topic. Treatment and management accounted for 33.1% of physician queries. Diagnosis, by contrast, accounted for 13.9%. Medication information came in at 14.1%, essentially tied with diagnosis. The authors noted that most published evaluations of clinical AI tools focus on diagnostic accuracy and knowledge recall, which leaves treatment selection, medication guidance, and evidence synthesis, the categories that appear to dominate real-world use, largely unmeasured.

Clinicians also used the tool to look well beyond their own specialties. Out-of-specialty queries ranged from 35% among oncologists and hematologists to 88% among emergency physicians.

The study did not break down NP and PA queries by specialty or practice setting, so the mechanics behind the higher per-user rate remain an open question. Several possibilities fit the pattern. NPs and PAs often work across broader clinical territory than specialist physicians, which can push them to look up guidance more frequently. A free, cited reference tool fills that gap.

>>Listen to The Latest Nurse News Podcast

For NPs and PAs who are already using OpenEvidence, the study is a useful external data point. The usage pattern looks like sustained use rather than curiosity, with high retention and rising per-user query volume over time. The study did not observe whether queries happened during patient encounters, advance preparation, or education.

But if NPs and PAs are the heaviest users of a general clinical AI tool, nursing-specific guidance on how to use these tools well, how to document their use, and how to teach safe querying belongs in orientation and continuing education, not after the fact.

It also sharpens a question we have been tracking across Nursing AI Watch: nurses are increasingly central to how clinical AI actually gets used, but their position in how these tools are built, validated, and governed is not catching up. Our 2026 State of Nursing Survey found that only 19% of nurses say nurses have meaningful input into how AI tools are selected and deployed at their workplace. Our earlier reporting on the ChatGPT-Epic integration showed one place where that gap surfaces: OpenAI publicly described physician-led evaluations but did not identify nurse evaluators, raising questions about nursing representation in the validation process.

The JAMA study does not evaluate whether OpenEvidence’s answers are accurate, safe, or associated with better patient outcomes. Those questions remain open. 

One caveat worth noting: three of the study’s seven authors were affiliated with OpenEvidence during the research, including the company’s chief medical officer. OpenEvidence provided the query data and contributed to the analysis. JAMA Network also disclosed that it licenses journal content to OpenEvidence, while noting that the paper underwent independent peer review and editorial decision-making. The study provides an unusually detailed look at platform use, but is not a fully independent evaluation of the tool.

What it does show is that hundreds of thousands of verified US clinicians were using OpenEvidence each month by June 2026, and nurse practitioners, alongside physician assistants, are among the platform’s most intensive users.

For a primer on OpenEvidence itself, see Nurse.org’s earlier coverage.

Related Nursing AI Watch Analysis

🤔 Are you using OpenEvidence or another clinical AI tool at the bedside? Tell us what’s working, and what isn’t, in the comments below.

Nursing Industry Research