OpenEvidence offers ‘ChatGPT for doctors’ and learns only from verified medical journals

“ChatGPT for doctors.” That’s what startup OpenEvidence is quickly becoming known as.

Founded in 2022, OpenEvidence started with a simple idea: “information retrieval,” Travis Zack, chief medical officer at the company, told Healthcare Brew at the 2025 HLTH conference.

There’s a seemingly endless supply of high-quality health research out there from peer-reviewed sources like the Lancet and Nature Medicine. When clinicians need answers for patients, they often turn to these sources.

OpenEvidence is meant to make it easier and faster for them to find that reputable information. It’s similar to tools like DxGPT, developed by tech nonprofit Foundation 29, which was built off Chat GPT-4. ​​As generative AI has improved in the last few years, Zack said, it became easier to create a tool providers could use.

Two years ago, he said he was drawn to the company’s mission of making “clinical practice better” and not simply “more efficient”—plus, it was dedicated to making everything free to boost accessibility since it’s supported by ads.

“We can’t just create AI systems that are only applied to a certain group of people who can afford it,” he said. “Every doctor in the US should have access to it.”

With the New York Times reporting $200 million in Series C funding on Oct. 20, only three months after its $210 million Series B round concluded, the company is currently valued at $6 billion.

The healthcare AI market overall, valued at $26.6 billion in 2024, is expected to reach $187.7 billion by 2030, according to market intelligence company the Research Insights.

How does it work? Using it is as simple as using Google. Providers type in what they need to know, and an answer comes up within seconds. The catch is it’s only for clinicians, so patients can’t use it as their new WebMD. The company requires users to provide their National Provider Index number and verifies their identity through secondary information, like medical school records or state medical licenses.

“This is for healthcare professionals,” Zack said. “Disintermediation of the provider is already something we’re seeing. We don’t want to contribute to that.”

A diagnosis and how treatment will work “should be between a physician and a patient,” he added.

Another core principle of the company is to only pull information from trusted sources.

Large language models are trained for “general purposes,” Zack said. “That means [information] could come from the New England Journal. It could also come from Reddit forums.”

That also creates problems with hallucination, he added, which is when AI pulls together incorrect information or completely fabricates responses. This happens when an AI tries to answer questions “with an enormous black box,” Jack Lindeman, SVP of commercial strategy at OpenEvidence, told us.

“OpenEvidence surfaces evidence through a search process,” he added. “OpenEvidence develops state-of-the-art AI search algorithms that find the exact guidelines, papers, reviews, etc. that collectively address the user’s question, and then applies a layer of conversational fluency on top of these resources.”

The company carefully curates the information it puts online, then gives clinicians access to the papers it pulls from so they can fact-check responses, which is in line with the “trust, but verify” medical practice, Zack said. This curation is done through partnerships with the JAMA Network and the New England Journal of Medicine, which provide OpenEvidence full access to their trusted research.

The company has additional licensing deals and partnerships with the National Comprehensive Cancer Network and sources from other places like PubMed, the FDA, and the CDC. It adds new information to the platform as soon as resources are published, Lindeman said.

Who’s using it? Anesthesiologist and founder of health tech investing firm Langar Holdings Sanjana Vig heard about OpenEvidence from a colleague about a year ago, she told Healthcare Brew.

Vig said she uses the tool when she encounters a patient who has a disease or medication she isn’t familiar with or who doesn’t see regularly. Lately, she’s been using it to address updates to cardiac protocols to ensure she’s using best practices.

“Getting the information in aggregate on OpenEvidence has been super helpful,” she said.

Before a tool like this existed, doctors would manually reference textbooks or online resources that could take much longer to turn out answers, Antonio Forte, a plastic surgeon at Mayo Clinic in Jacksonville, Florida, and a medical advisor to OpenEvidence, told us. He’s used OpenEvidence, for example, to review the pros and cons of surgical treatments for a complex patient to make sure he made a more efficient and personalized decision.

There’s already been research about how AI use in healthcare can lead to clinical deskilling, which is when physicians lose their knowledge and skills due to overreliance on new technology. But Forte said he thinks these tools can make doctors smarter.

“Now they will just have more information at their disposal at a faster speed than before,” Forte said.