OpenEvidence: What Healthcare Providers Need to Know

If you have been in a clinical setting recently, you have probably heard a colleague mention OpenEvidence. Perhaps you have seen a resident pull it up on their phone during rounds, or a colleague reference it when discussing a complicated case. In just a few years, this AI-powered platform has become one of the most talked-about tools in American healthcare. But what exactly is it, who can use it, and how reliable is it?

This post breaks down what healthcare providers need to know about OpenEvidence, organized around the essential understandings that matter most for clinical professionals considering whether and how to integrate it into their practice.

Essential Understanding: OpenEvidence was born from the same AI expertise that transformed Wall Street, now redirected toward solving the crisis of medical information overload.

A Brief History of OpenEvidence

OpenEvidence was founded in late 2021 by Daniel Nadler, PhD, and Zachary Ziegler. Nadler was not new to the AI world. While still a PhD student at Harvard, he co-founded Kensho Technologies, an artificial intelligence company that built machine learning systems for financial analysis. In 2018, S&P Global acquired Kensho for $550 million, validating the idea that specialized AI could make sense of massive, complex datasets.

The leap from finance to medicine was not random. Nadler recognized a direct parallel: just as Wall Street professionals were drowning in financial data, physicians were struggling to keep pace with an ever-expanding body of medical literature. Medical knowledge is estimated to double in volume every 73 days. No clinician, no matter how dedicated, can stay current by reading journals alone.

The personal dimension mattered too. Nadler's grandfather died due to a medical error, and Ziegler's brother-in-law went through treatment for leukemia. These experiences fueled the mission to build a tool that could put the right research in front of the right clinician at the right moment.

OpenEvidence participated in the Mayo Clinic Platform Accelerate program to refine its vision, and in July 2023 announced that its AI model had become the first to score above 90% on the United States Medical Licensing Examination. In 2025, the model reached a perfect 100% on that same exam.

The growth trajectory has been extraordinary. In February 2025, the company raised $75 million in a Series A round led by Sequoia Capital at a $1 billion valuation. By July 2025, a $210 million Series B followed at $3.5 billion, co-led by Google Ventures and Kleiner Perkins. By October 2025, a Series C round valued the company at $6 billion. Investors include Sequoia, Google Ventures, Kleiner Perkins, Nvidia, Coatue, Mayo Clinic, and others. In 2025, Nadler was named to the TIME100 Health list of the 100 Most Influential People in global health.

Essential Understanding: OpenEvidence functions as an AI-powered clinical copilot that translates natural language questions into evidence-based answers drawn exclusively from peer-reviewed medical literature.

What OpenEvidence Is and How Clinicians Use It

At its core, OpenEvidence is an AI-powered medical search engine and clinical decision support platform. Unlike general-purpose AI tools like ChatGPT, OpenEvidence is purpose-built for medicine and trained exclusively on peer-reviewed medical literature without connection to the public internet. The company describes its technology as an ensemble of specialized models rather than a single large model, using a retrieval-augmented generation (RAG) architecture.

The platform draws from over 300 medical journals, including formal content partnerships with the New England Journal of Medicine (full text and multimedia from 1990 onward), JAMA and its 11 specialty journals, as well as collaborations with organizations such as the NCCN, ACC, ADA, ACEP, AAFP, AAOS, and others. It also incorporates content from the FDA and CDC.

How Clinicians Actually Use It

The experience is straightforward. A clinician types a natural language question, much like texting a colleague, and receives an evidence-based answer within seconds, complete with citations to specific studies and guidelines. For example, a primary care physician might ask about the latest recommended therapy for severe asthma not controlled by inhalers and receive a summary of recent biologic treatments from current trials. A cardiologist might query the platform during a consultation to compare two medications' efficacy using head-to-head trial data.

Beyond Search: Additional Features

OpenEvidence has expanded well beyond basic search. The platform now offers a Visits feature that can listen to or transcribe patient encounters and draft structured clinical notes (such as SOAP notes). It includes a HIPAA-secure Dialer that lets clinicians call patients from personal devices while displaying the clinic's office number. A DeepConsult feature introduced in mid-2025 uses reasoning models to synthesize findings across multiple studies for more complex questions. The platform also generates patient handouts, creates risk score calculations, drafts prior authorization letters, and offers free CME credits (AMA PRA Category 1) for verified NPI users.

Essential Understanding: Access to OpenEvidence is gated by professional credential verification, and a National Provider Identifier (NPI) number is the primary key to full platform access.

Who Can Use OpenEvidence and How to Qualify

OpenEvidence is not available to the general public. This is a deliberate design choice. The platform contains professional-level clinical information that requires clinical judgment to interpret, and the company has intentionally restricted access to prevent patients from using it for self-diagnosis.

NPI-Verified Professionals (Full Access, Free)

The primary pathway to full, unlimited access is through a valid National Provider Identifier (NPI) number. During the sign-up process, users enter their NPI, which the platform uses to verify their professional status. This includes physicians (MDs and DOs), nurse practitioners, physician assistants, registered nurses, pharmacists, dentists, physical therapists, and other licensed professionals who hold an NPI. For these verified users, the platform is completely free.

Medical Students (Expanded Access)

Access was originally limited exclusively to NPI holders, but OpenEvidence has since expanded to include medical students. Students typically verify their status through institutional credentials or proof of enrollment rather than an NPI number.

Non-NPI Users (Limited Access)

Users without an NPI can create an account but are restricted to approximately two searches per 24-hour period. This allows learners and others to sample the platform while maintaining the professional gating that OpenEvidence considers a core safety feature.

International Users

The platform is primarily optimized for the U.S. healthcare system. International clinicians may be able to access OpenEvidence, but full verification and free access features are currently designed around U.S. credentials. Canadian users, for example, may face limitations without a U.S. NPI.

How to Get an NPI Number

For healthcare professionals who do not yet have an NPI, the process is straightforward. The National Plan and Provider Enumeration System (NPPES), managed by the Centers for Medicare and Medicaid Services (CMS), handles NPI registration at nppes.cms.hhs.gov. There is no cost to apply. Once issued, the NPI serves as a unique, permanent identifier for the provider, and it can be used to register for OpenEvidence immediately.

Essential Understanding: OpenEvidence is a decision-support tool with significant legal, privacy, and practical boundaries that every clinician must understand before relying on it.

Restrictions and Important Considerations for Healthcare Providers

It Is Not a Medical Device (Yet)

OpenEvidence currently positions itself as an information and clinical decision support tool, not as a diagnostic or treatment device. Its terms of use explicitly state that the platform makes no warranties regarding accuracy, completeness, or suitability, and that the information should not replace the clinician's own professional judgment. If regulators determine that OpenEvidence directly influences clinical decision-making rather than simply providing information, it could face FDA oversight as a medical device. For now, the legal responsibility for clinical decisions remains entirely with the provider.

Privacy and Data Considerations

OpenEvidence is HIPAA compliant and SOC 2 Type II certified. However, clinicians should be aware that their usage data is not entirely private. The platform's privacy policy states that it may share limited identifiers (such as hashed email addresses, device identifiers, IP addresses, and NPI numbers) along with interest categories inferred from on-platform activity with advertising and identity-resolution partners. Importantly, the company states that it does not share the text of questions or conversations for advertising purposes. Clinicians should ensure they never include Protected Health Information (PHI) in their queries, as the platform is not designed to serve as a patient record system.

Revenue Model: Advertising, Not Subscriptions

OpenEvidence is free to verified clinicians because it generates revenue through pharmaceutical advertising and market research programs. Clinicians should be aware that, like many free platforms, they are interacting with a product that monetizes their attention and professional data. The platform may contact users about market research opportunities and may share aggregated usage information with third parties.

Not a Replacement for Clinical Judgment

Multiple sources, including OpenEvidence's own documentation, emphasize that the platform is one source among many and should not be the sole basis for clinical decisions. It is crucial to check what the original references say and to continue comparing outputs with trusted sources. All AI systems may produce errors, and without knowing the full scope of which sources are consulted or excluded, clinicians should cross-reference important findings.

Essential Understanding: OpenEvidence performs well on standardized medical knowledge tests and routine clinical questions, but independent research reveals important accuracy limitations in complex subspecialty scenarios that every clinician should weigh.

Accuracy, Evidence, and What the Research Says

The Headline Number: 100% on the USMLE

OpenEvidence became the first AI system to achieve a perfect score on the United States Medical Licensing Examination. While impressive, it is important to understand what this does and does not mean. The USMLE is a standardized, multiple-choice exam testing broad medical knowledge. Scoring 100% demonstrates strong knowledge retrieval and reasoning on structured questions. It does not, by itself, prove that the system can handle the nuanced, ambiguous, and multifactorial nature of real-world clinical decision-making.

What Independent Studies Show

A peer-reviewed study published in the Journal of Primary Care & Community Health evaluated OpenEvidence across five common chronic conditions (hypertension, hyperlipidemia, diabetes mellitus type 2, depression, and obesity). Four independent physicians rated the platform's responses. The results were encouraging on several dimensions: clarity scored 3.55 out of 4, relevance scored 3.75, and evidence-based support scored 3.35. Physician satisfaction was high at 3.60. However, the impact on actual clinical decision-making was limited, scoring only 1.95 out of 4. The authors concluded that OpenEvidence primarily reinforced existing physician plans rather than modifying them.

A preprint study on medRxiv tested OpenEvidence on complex medical subspecialty board-level questions (from the MedXpertQA dataset) and found that performance on these more challenging scenarios was considerably less impressive than the USMLE results would suggest. The researchers noted that scoring well on standardized student-level questions may show significant accuracy limitations when faced with the complexity of real-world clinical decision-making.

Other comparative studies have produced mixed results. One study comparing OpenEvidence with ChatGPT-4o for structural heart disease questions found that ChatGPT-4o produced more reliable answers as judged by subject matter experts. Another study comparing multiple AI platforms on 50 open-ended clinical scenarios scored OpenEvidence at 24% accuracy, behind at least one competitor.

What Physicians Are Saying

Physician sentiment is genuinely split. In a Sermo poll, 54% of doctors described themselves as cautiously open to AI in decision-making, while 21% were somewhere between concerned and skeptical. Notably, 60% of respondents had only vaguely heard of OpenEvidence, despite the company's reported adoption numbers.

Among enthusiastic users, praise centers on speed and convenience. One oncologist called it an incredible lifeline for daily practitioners, while a neurologist reported using it frequently throughout the day. The late Nobel laureate Daniel Kahneman praised the platform's approach to making medicine more evidence-based.

Critics raise valid concerns. One internal medicine physician stated plainly that when the platform is wrong, it is annoyingly confident in its wrong answer. Others have flagged inaccuracies, outdated information, slow response times for advanced features, and limited control over output formatting. Researchers have pointed to a lack of transparency around how articles are curated or excluded, the timeliness of information, and the platform's clinical relevance in specialized areas.

The Bottom Line on Accuracy

OpenEvidence is a powerful tool for rapid evidence retrieval, particularly for common clinical questions where the medical literature is robust. It is not infallible. Like any AI system, it can produce errors (sometimes called hallucinations), and it may be less reliable in highly specialized or edge-case scenarios. The platform is best understood as a well-read colleague who can quickly pull relevant research, not as an oracle that replaces clinical reasoning.

Final Thoughts

Essential Understanding: The most effective use of OpenEvidence is as one component within a broader, critically evaluated clinical decision-making process, not as a standalone authority.

OpenEvidence represents something genuinely new in healthcare: an AI tool built specifically for clinicians, trained on the literature clinicians trust, and adopted at a pace that suggests it is meeting a real and urgent need. For healthcare providers with an NPI, there is no financial barrier to trying it. The platform's strength lies in its speed, its grounding in peer-reviewed sources, and its expanding feature set.

But the technology is still young, and healthy skepticism remains appropriate. Cross-reference important findings. Be aware of the privacy and advertising dimensions. Understand that a perfect USMLE score does not equal perfect clinical judgment. Use it as a tool that augments your expertise, not one that replaces it.

For clinicians navigating the rapidly expanding world of healthcare AI, OpenEvidence is worth knowing about, worth trying, and worth evaluating critically. That combination of openness and rigor is exactly the mindset that evidence-based practice demands.