OpenEvidence moves from individual clinician use to systemwide Epic support at NYU Langone
OpenEvidence is moving deeper into hospital infrastructure with a systemwide integration at NYU Langone Health, placing its physician-facing clinical AI directly inside Epic workflows across one of New York’s largest academic systems. The deployment spans hospitals, ambulatory sites and Family Health Centers, and reaches physicians training at both the NYU Grossman School of Medicine and the NYU Grossman Long Island School of Medicine.
The scale is hard to miss. NYU Langone logs more than 12 million outpatient visits each year, along with 421,000 emergency department visits, seven inpatient locations and more than 400 outpatient locations. More than 6,000 physicians and faculty, 2,000-plus residents and fellows and 54,000 employees work inside that network. Embedding an evidence engine into that setting is not a cosmetic add-on; it places a query tool directly where clinicians make decisions, document care and move from suspicion to treatment.
Clinical AI is no longer sitting outside the chart.
At the center of the partnership is OpenEvidence’s AI-powered medical search engine and generative chatbot built exclusively for doctors. The company describes the platform as a way to summarize and simplify evidence-based medical information, while its “patient-aware clinical intelligence” integrates details from the patient record with medical literature. Within NYU Langone’s Epic environment, physicians will be able to ask complex clinical questions in natural language and receive answers grounded in peer-reviewed literature without leaving the EHR.
That workflow matters because clinical decision-making is often compressed into seconds. A physician weighing an antimicrobial choice, an oncology regimen, a diagnostic workup or a management pathway does not need a separate browser tab and a search session that interrupts the encounter. A tool embedded in the chart can, in theory, reduce friction between the question and the evidence, especially when the patient-specific context is already available in the same workspace. The biological logic is indirect but practical: better retrieval of evidence should improve the probability that the next action matches current standards, particularly in fast-moving specialties where treatment details shift quickly.
Nader Mherabi, executive vice president and vice dean and chief digital and information officer at NYU Langone Health, framed the partnership as an effort to make the latest technology available to clinicians “in a safe and practical environment at scale.” That phrasing fits a system that spans outpatient, inpatient and training environments, where the same decision-support tool may be used by attending physicians, residents and fellows under different time pressures and levels of experience.
Adoption data show a tool already embedded in daily physician work
OpenEvidence is not entering the market as an untested concept. The company says roughly two-thirds of U.S. physicians use its platform, and as of October it had 1.2 million medical license-verified U.S. clinicians, including physicians and nurses, using the system. OpenEvidence executives said verified U.S. clinicians consulted the platform 45 million times in October alone.
A recent JAMA study helps explain why health systems are treating the platform as more than a novelty. Nearly 1 million U.S. clinicians submitted 271 million AI-enabled clinical queries through OpenEvidence from January 2024 to June 2026. More than half of the users in that analysis, 556,200, were physicians. Those physicians submitted an average of 37 queries per month of use, and early-career physicians as well as clinicians in oncology, hematology, internal medicine and family medicine accounted for a higher-than-average share of queries.
The query mix is clinically revealing. Most requests were tied to clinical decision-making, which accounted for 51.9% of queries, with treatment and management forming the largest slice of that category. Diagnosis represented 13.9% of queries. In other words, the platform is not being used as a generic medical encyclopedia; it is being used at the bedside and in the chart for decisions that affect therapy, workup and next steps.
Treatment questions dominate the workflow.
The growth curve was steep. Monthly active OpenEvidence users rose from 355 in January 2024 to 703,265 in June 2026. Queries per active user climbed from 18 per month to 39.5, and month-to-month retention reached roughly 90% in 2026. By June, 92.5% of active users were returning users rather than brief adopters. Those figures suggest repeated use, not sporadic experimentation, and they help explain why health systems are increasingly willing to integrate the tool into official clinical infrastructure.
New York academic medicine is becoming a proving ground for clinical AI
NYU Langone is not OpenEvidence’s only major New York foothold. The company already has an enterprise rollout across NewYork-Presbyterian’s 10 hospitals and nearly 350 care sites, a systemwide deployment at Mount Sinai and a precision oncology partnership with Memorial Sloan Kettering Cancer Center. It also has partnerships with New York City medical schools around training and research.
The clustering matters because academic centers create a demanding environment for clinical AI. Tertiary and quaternary care physicians see complex cases, subspecialty questions and high-acuity patients where evidence can be difficult to retrieve quickly and where overreliance on memory is risky. A system that surfaces peer-reviewed literature in real time has obvious appeal in those settings, especially when clinicians are practicing across multiple specialties and care venues.
Daniel Nadler, founder and CEO of OpenEvidence, tied the NYU Langone partnership to the scale of New York medicine itself, arguing that the city is where American medicine proves itself at scale. In practical terms, the institutional value lies in embedding the tool where it can influence a very large volume of encounters, including more than 12 million outpatient visits annually.
The unanswered question is not whether physicians will use the tool. They already are. The harder issue is how consistently AI-generated literature support will affect diagnosis, prescribing, escalation decisions and specialist consultation in real-world care. The JAMA analysis established the size and shape of usage, but not downstream patient outcomes. In a system as large as NYU Langone, those outcomes will depend on how well the platform fits into clinical judgment, training culture and the cadence of Epic-based care.