www.kaboompics.com / Pexels
According to NBC News, at least 65% of doctors in the U.S. have used the AI-powered search service called OpenEvidence in the past month, which translates to over 600,000 users in the U.S. and more than 1 million globally.
The neural network, designed to compile information from medical research articles, began development in 2022 and demonstrated over a 90% pass rate on the U.S. medical licensing examination within a year. By 2026, the startup OpenEvidence reached a valuation of $12 billion, attracting investment from companies like Google, Nvidia, and Blackrock.
OpenEvidence is described by NBC News as a neural network equivalent of a «seasoned colleague» that physicians can consult for advice on treatment options based on patient symptoms. The platform also includes tools for generating medical prescriptions.
Dr. Anupam Jena from Harvard, who is studying the tool, stated that OpenEvidence’s popularity among peers is growing at an «exponential rate.»
«60% of all search queries involve decision-making in clinical situations. Doctors often inquire about the correct treatment for a patient with specific conditions, comorbidities, etc.,» Jena noted.
According to Jena, physicians often seek information not strictly related to their specialty but about aspects that touch on other medical fields or the application of specific medications under certain circumstances.
He cited an instance where OpenEvidence helped clarify how to prescribe antibiotics to a patient without a spleen, locating a relevant 2014 article that standard search engines like Google did not yield.
OpenEvidence explicitly states in its usage guidelines that it is not a substitute for medical expertise. The service is accessible only to licensed physicians, and the developers have made no indication of plans to extend it to the broader public.
«The service is not intended as a diagnostic tool, does not aim to eliminate diagnostic inaccuracies, nor does it recommend specific drugs or treatments, or replace the judgment of qualified medical professionals,» the guidelines read.
The company acknowledges concerns about the «hallucinations» inherent in AI, where algorithms might generate or fabricate information to avoid disappointing users. Daniel Nadler, the founder of the company, emphasized in an interview with NBC News that OpenEvidence mitigates this issue by partnering with leading U.S. medical journals such as the New England Journal of Medicine (NEJM) and The Journal of the American Medical Association, as well as the National Comprehensive Cancer Network (NCCN) and the American Diabetes Association.
Despite Nadler’s assurances that the neural network functions as a «search adhesive» rather than generating text, some physicians interviewed by NBC News reported encountering instances of somewhat «exaggerated» responses from OpenEvidence, particularly on less-explored topics. However, many emphasized their surprise at the «accuracy» of the searches, noting that they strive to read original articles rather than rely solely on AI compilations.
In March, the American Medical Association (AMA), considered the largest organization of doctors and medical students in the U.S., reported from a recent survey that over 80% of physicians have begun to use AI in their work.
The head of the association, Dr. John White, emphasized that tools must be «safe, effective, and used responsibly» to «empower» physicians rather than replace them. Approximately 40% of respondents indicated they specifically use AI for compiling reviews of medical articles and the latest treatment standards.