OpenEvidence: the AI medical tool most American doctors already use without you knowing
Artificial intelligence is already inside American medical offices, and most patients have no idea.
Over the past two years, a tool called OpenEvidence was quietly adopted by a massive share of doctors in the United States, transforming the way they look up information, make clinical decisions, and even prepare for medical licensing exams. It is essentially a chatbot tailor-made for healthcare professionals, and the speed at which it spread across the country caught a lot of people off guard.
The numbers are hard to ignore: this past April, the platform was used in nearly 27 million clinical encounters, representing about 65% of active physicians in the country 🤯
To put that in perspective, we are talking about roughly 650,000 healthcare professionals in the United States and another 1.2 million internationally who use OpenEvidence to answer questions about real patients, review treatment options, and consult the latest medical literature.
But what exactly is this tool, why did it grow so fast, what are the risks involved, and what does it mean for the future of medicine? That is exactly what we are going to break down here 👇
What is OpenEvidence and how does it work
OpenEvidence is an artificial intelligence platform developed specifically for healthcare professionals. Unlike generic tools like ChatGPT or Google, it was built with a very clear focus: delivering fast, accurate, evidence-based answers to doctors who need information at the exact moment of patient care.
The platform’s homepage introduces itself as America’s Official Medical Knowledge Platform and features a search bar where doctors can type questions like alternatives if metformin causes diarrhea or what are the latest advances in gene therapy for Duchenne muscular dystrophy.
The way it works is relatively simple on the surface, but technically sophisticated under the hood. A doctor types in a clinical question and the system returns a structured response, complete with references to peer-reviewed articles and medical guidelines that back up the answer. The language model behind the tool was trained and fine-tuned with high-quality medical literature, which significantly reduces the risk of hallucinations — that classic AI problem where the model confidently makes up information.
The technical edge OpenEvidence has comes from the licensing partnerships the company secured with the most prestigious medical journals in the world, such as the New England Journal of Medicine and the Journal of the American Medical Association. On top of that, the platform also has agreements with specialized organizations like the National Comprehensive Cancer Network and the American Diabetes Association, ensuring access to the most up-to-date treatment guidelines.
As Daniel Nadler, the company’s CEO, explained in an interview with NBC News: the company thinks of AI as a search glue. They have access to the full text and figures from all their partners, and they do not need the AI to invent answers from scratch.
To use the platform, a healthcare professional needs to register with their unique identification number issued by the U.S. government. Once registered, they can ask unlimited questions — for free. The service is funded by advertising, including from pharmaceutical and medical device companies, although several doctors interviewed by NBC News commented that the ads are either discreet or practically nonexistent in the user experience.
Why doctors adopted it so quickly
The speed at which OpenEvidence was adopted among American doctors is striking even for those who have followed the tech industry for a long time. In a sector historically conservative when it comes to changing tools and workflows, seeing 65% of an entire country’s active physicians using the same platform in under two years is, at the very least, surprising.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Nadler himself did not hide his pride when talking about it: according to him, the company did the hardest thing in the history of the American healthcare system — convincing the majority of doctors to voluntarily adopt a single technology platform.
Dr. Jeremy Cauwels, a hospitalist and chief medical officer at the Sanford Health system, based in Sioux Falls, South Dakota, summed up the phenomenon well. According to him, OpenEvidence is one of those remarkably easy tools to adopt: it is available for free, it works great on a phone, and it can answer questions faster than any other method.
The first reason for this massive adoption is the information overload healthcare professionals face every day. Medicine is one of the fields that produces the most new research, and it is humanly impossible for any doctor to stay current in every specialty at once. When a question comes up about an uncommon medication, a rare drug interaction, or a recently updated protocol, the professional previously had to turn to databases like PubMed, UpToDate, or even Google Scholar, which ate up valuable time during a visit. OpenEvidence compressed that process down to seconds.
Another decisive factor was the trust the tool built over time. Doctors are professionals who make decisions based on evidence, and any tool that wants to break into their world needs to speak the same language. OpenEvidence understood this from the start and built the platform with a layer of transparency that generic tools simply do not offer. Each response comes with the references that support it, which lets the doctor verify the source with one click and decide whether that information applies to their specific patient’s case.
There is also a very practical component that cannot be overlooked: the interface is clean, straightforward, and does not require any special training. In an environment where healthcare professionals already deal with heavy hospital systems, unintuitive electronic health records, and a mountain of administrative platforms, having an artificial intelligence tool that just works with no learning curve was a huge competitive advantage.
Real-world use cases in everyday clinical practice
NBC News spoke with more than two dozen doctors, hospital administrators, medical students, and healthcare researchers, from Hawaii to Maine, to understand how OpenEvidence is being used in practice. Every person interviewed said they either used the tool regularly or knew someone who did.
A medical resident at a hospital in New Hampshire, for example, said that when he saw a patient’s potassium level plummet, he checked OpenEvidence to confirm whether it was a normal side effect of a medication or a new emergency. After pulling from peer-reviewed medical publications, the tool confirmed it was a common side effect and presented several options for restoring normal potassium levels.
On the other side of the country, a doctor with the Indian Health Service on the Pine Ridge reservation in South Dakota, a rural and remote area, was not convinced that a patient’s spine was fractured after looking at spots on an X-ray. He vaguely remembered from medical school that a different type of imaging might be needed for a definitive diagnosis. When he consulted OpenEvidence, he got confirmation that a CT scan was preferred for confirming that type of fracture, along with links to scientific articles with more details.
Dr. Anupam Jena, an internal medicine physician at Massachusetts General Hospital in Boston and a health policy professor at Harvard, shared his own frustrating experience trying to look up — with traditional tools — how to adjust antibiotics for a patient who had his spleen removed years earlier. He tried multiple search approaches and simply could not find the answer. On OpenEvidence, the answer came up quickly, with a reference to a 2014 New England Journal of Medicine article he could not find anywhere on Google.
A kidney specialist, who asked for anonymity because the tool had not been explicitly approved by his hospital, said OpenEvidence regularly saved him 30 minutes of fruitless searches through older systems, including UpToDate.
The crucial role outside a doctor’s specialty
An interesting pattern that emerged from the data and interviews is that OpenEvidence proves particularly valuable when doctors need to handle questions outside their main specialty.
Dr. Jena, who is currently analyzing 90 million queries submitted to OpenEvidence since 2024 as part of a new research project, explained that about 60% of all searches on the platform are about how to make clinical decisions. Doctors ask things like: for this specific patient, with this profile and this condition, maybe with other comorbidities, what is the right treatment?
He gave a practical example: if you are a surgeon, you know how to handle everything related to surgery. But if you see a patient and notice that their blood pressure or heart rate is a little high, you might not be sure whether you can stop a medication that keeps those values under control. It is exactly in those situations that doctors are turning to OpenEvidence to answer questions that are part of their clinical practice but not specific to the area they were trained in.
The real impact on clinical decisions and how it compares to UpToDate
When we talk about clinical decisions, we are talking about choices that directly affect the lives of real people. Which medication to prescribe, which test to order, how to interpret an atypical result, when to refer to a specialist. These are decisions doctors make dozens of times a day, under time pressure, with incomplete information, and often with the feeling that they might be missing something.
Dr. Paul Sax, an infectious disease specialist at Brigham and Women’s Hospital in Boston, said OpenEvidence frequently borders on the miraculous, with its AI-powered search delivering personalized answers that other tools once considered the gold standard simply cannot match right now.
For years, Sax explained, clinicians turned to a medical reference site called UpToDate for treatment recommendations and clinical decision support. However, UpToDate consists of long, peer-reviewed summaries of the latest research, which are difficult to search for doctors with targeted questions about specific scenarios. According to Sax, OpenEvidence’s search feature is far more flexible, and the process of finding answers is frictionless. That is the power of large language models: you do not have to search by specific terms, you search with the actual question.
UpToDate, for its part, is racing to roll out its own AI tool, called Expert AI. A company spokesperson, Suzanne Moran, said about 2,000 hospitals and health systems had signed up for Expert AI by April 30.
In April, OpenAI also launched its own medical-focused version, called ChatGPT for Clinicians. However, the service does not currently license the same level of cutting-edge medical information accessible to OpenEvidence users.
Legitimate concerns about accuracy and safety
Despite the enthusiasm, it is not all sunshine and roses. With OpenEvidence’s explosive popularity, some experts are raising legitimate concerns about potential hallucinations or incomplete answers, the lack of rigorous scientific studies on the tool’s impact on patient outcomes, and the risk of eroding doctors’ critical thinking and evaluation skills as use of and reliance on the platform grows.
Several doctors interviewed noted that OpenEvidence sometimes gets things wrong or overstates its responses, particularly with rare conditions or edge cases. Some pointed out that the tool occasionally draws overly strong conclusions from medical studies with small sample sizes, although others observed that even the errors tend to err on the side of caution.
Dr. John Rozehnal, an emergency medicine physician in New York, shared that OpenEvidence incorrectly suggested that injecting a certain medication could damage a patient’s liver, when in reality the risk of liver damage was very low and far more likely to be caused by the patient’s excessive alcohol consumption. Weeks later, Rozehnal told NBC News that OpenEvidence had improved its response and now correctly reflected the role of the patient’s alcohol use.
While OpenEvidence highlights that it scored 100% on the official United States Medical Licensing Examination (USMLE), an academic study published in December found that the tool answered more complex medical questions correctly less than 45% of the time. That study has not yet been peer-reviewed, which means its findings need to be interpreted with caution, but the data point is relevant.
OpenEvidence itself makes clear in its terms of service that the platform is meant to supplement, not replace, doctors’ clinical judgment. The company also states that it is HIPAA-compliant, the federal health privacy law in the U.S., through a series of privacy protocols and safeguards. Still, some health systems are not satisfied with the overall privacy protections. MaineHealth, for example, currently asks its doctors to avoid entering protected health information into the platform.
The risk to new doctors’ training
One concern that came through strongly in the interviews is the impact of OpenEvidence on the training of students and early-career physicians. While more experienced practitioners have years of clinical expertise to fall back on as a safety net, early reliance on an AI tool can create misplaced confidence in those who are still learning.
NBC News spoke with several medical students who use OpenEvidence to study and prepare for case discussions with their professors. A mid-career doctor in Missouri, who asked for anonymity, said he is already seeing harmful effects on students’ ability to separate relevant signals from clinical noise.
According to that physician, his concern is that when you introduce a new tool that performs some of the skills the professional trained for years, those skills start to erode quickly. And it is being introduced to students from the very beginning of their training, which is worrying because most curricula still do not have a structured way to teach how to use these tools safely.
Dr. Hannah Galvin, a pediatrician and chief health information officer at Cambridge Health Alliance in Massachusetts, is conducting research specifically to fill that evidence gap. Her work will examine how early-career doctors use OpenEvidence and compare their responses with those from general-purpose chatbots, like ChatGPT and Google’s Gemini, using the same prompts.
The billion-dollar investment behind the platform
The financial market is paying attention, too. The startup raised $700 million in less than a year and is backed by the biggest names in venture capital: Sequoia Capital, Google Ventures, Nvidia, Andreessen Horowitz, Thrive Capital, among others. Valued at $1 billion in early 2025, the company skyrocketed to a $12 billion valuation in just over a year.
OpenEvidence is part of a growing industry of AI-powered medical tools, ranging from AI scribes that record and transcribe doctors’ speech during visits to competitors like Doximity and iatroX. A recent American Medical Association survey found that more than 80% of responding physicians already use some form of artificial intelligence. Nadler’s team plans to expand AI note-taking features, billing, and consultation integration in the coming years.
Bringing AI out of the shadows and into the open
Dr. Girish Nadkarni, a nephrologist and head of AI at the Mount Sinai Health System in New York, raised a fundamental point about the phenomenon known as shadow AI. According to him, there is an entire growing area of unofficial AI use, where health systems or institutions focus only on the tip of the iceberg, but there is a whole part of the system below the surface, where doctors and other professionals simply use tools on their personal computers without institutional oversight.
In March, Mount Sinai, which employs 47,000 people, announced a new corporate partnership with OpenEvidence to link the tool directly to the hospital system’s main electronic health records portal, making it available to doctors, nurses, and pharmacists. For Nadkarni, it is time to bring tools like OpenEvidence to the surface.
This stance reflects an important mindset shift: instead of ignoring that doctors are using AI on their own, the smarter approach is to work with professionals openly to discuss ethical and responsible use.
What this means for the future of healthcare with AI
The growth of OpenEvidence is not an isolated phenomenon. It is part of a larger movement integrating artificial intelligence into healthcare workflows, from diagnostic imaging to predictive risk analysis for chronic patients. What sets this particular case apart is the scale and speed at which it happened organically, without major marketing campaigns or institutional mandates. Doctors sought out the tool because it solved a real problem, and that says a lot about the potential of AI when it is developed with the end user at the center of the process.
On the other hand, the landscape also raises important questions that the medical community and regulatory bodies will need to tackle in the coming years. What level of accountability does an AI platform bear when a clinical decision based on its responses results in a negative outcome for a patient? How do you ensure models are updated frequently enough to reflect the latest evidence? How do you prevent a business model built on commercial partnerships, even transparent ones, from influencing responses in ways that are not completely neutral? These are legitimate questions, and the fact that definitive answers do not yet exist does not diminish the importance of asking them.
As Dr. Cornelius James, an internal medicine specialist and professor at the University of Michigan, put it: he knows the right questions to ask OpenEvidence and needs to combine any answer he gets with his clinical experience and intuition. For him, there is no concern about patient safety because he feels the need to check and double-check, in a trust but verify mindset.
What seems clear is that the relationship between doctors and artificial intelligence has already passed the point of no return. Tools like OpenEvidence are already part of the clinical routine of hundreds of thousands of professionals, and that number is only going to grow as the technology improves and new platforms emerge. The question is no longer whether AI will enter medicine — it already has. The question now is how to make sure that integration happens responsibly, transparently, and truly centered on the well-being of patients 🏥
