🔎︎
More than half of US physicians now regularly use an artificial intelligence app to help answer clinical questions, according to a June 8, 2026, report in The New York Times. The app, OpenEvidence, works as a medical chatbot: a clinician enters a question or a patient's symptoms and receives a summary of likely diagnoses, a list of serious possibilities not to miss, and links to the research behind each answer. Physicians logged roughly 30 million questions last month, nearly double the volume of six months earlier, and the company's valuation has climbed to $12 billion.
The appeal to health systems goes beyond convenience. Because OpenEvidence trained its models on licensed, peer-reviewed sources such as the New England Journal of Medicine and JAMA rather than the open internet, its answers tend to be evidence-based, and the tool lets thinly staffed hospitals reach specialist-level guidance they could not otherwise afford. A community-hospital physician in Fairbanks, Alaska, said it saves her time and her hospital money on specialist consultations, calling it like having a bunch of specialists in your pocket.
That rapid uptake has outpaced oversight. Much of the early use was unsanctioned adoption by clinicians without their employers' knowledge, and systems like Mount Sinai are now pulling the tool into governed workflows, in its case, a link from the electronic record, while withholding patient data pending testing and controls. The app is built to comply with federal health-privacy law and warns against entering identifying information. Its economics matter too: the service is free because physicians see ads, many from drug companies, while awaiting answers. Competition is intensifying, with the legacy reference UpToDate adding a chatbot and OpenAI introducing a clinician version.
The reporting tempers the enthusiasm. AI has aced licensing exams and, in some cases, outdiagnosed clinicians, yet has also botched summaries and returned incorrect answers, and one cited paper found meaningful competency gaps in health AI. The cardiologist Eric Topol cautioned that the technology remains unproven in the messy reality of everyday care. For plans and systems, the takeaway is a fast-moving decision-support category promising efficiency and access gains, yet it still demands validation, privacy safeguards, and governance.
