OpenAI makes ChatGPT free for doctors

Clinicians spend almost as much time fighting paperwork, insurance pre-authorizations, and journal backlogs as they do treating patients.

OpenAI just launched ChatGPT for Clinicians, making frontier AI models completely free for verified U.S. physicians, nurse practitioners, physician assistants, and pharmacists. This expands on ChatGPT for Healthcare, the enterprise product rolled out earlier this year, by putting clinical-grade tools directly into the pockets of individual practitioners.

AI adoption on the frontlines is exploding. According to a 2026 survey by the American Medical Association, 72% of physicians now use AI in their daily practice, up from 48% last year. OpenAI reports that clinician activity on ChatGPT has more than doubled over the same period.

Reusable skills and automatic CME credits

ChatGPT for Clinicians is not just a standard web interface. Built on models like GPT-5.4, it is wired specifically for clinical workflows.

Clinicians can save repetitive paperwork—such as referral letters, prior authorization requests, and patient discharge instructions—as reusable skills to ensure the model follows identical guidelines every run.

For literature reviews, the workspace searches through millions of peer-reviewed sources, delivering cited syntheses in minutes. In a benchmark subset of 355 test cases with physician-defined ground-truth citations, the system cited primary sources more frequently than human doctors.

Crucially, researching real clinical questions in ChatGPT can now automatically count toward continuing medical education (CME) credits, cutting out manual paperwork and standalone webinars. Chats are excluded from model training by default, with multi-factor authentication and optional Business Associate Agreements (BAA) for HIPAA compliance.

To prove safety, OpenAI had physician advisors test the tool across 6,924 conversations in their actual workflows, rating 99.6% of responses as safe and accurate. OpenAI also released HealthBench Professional, an open evaluation dataset covering care consults, documentation, and medical research, where a third of the scenarios were physician-designed red-teaming tasks.

Access is currently limited to verified clinicians in the U.S., with international pilots planned alongside the Better Evidence Network.

Why it matters

For overworked clinicians, this is an immediate win. It does not replace medical judgment, nor should it, but automating prior authorizations and earning CME credits while looking up drug interactions solves very real, daily headaches.

Strategically, OpenAI is pulling off a classic bottom-up land grab. Health systems are notoriously slow at buying enterprise software. By giving frontier tools away for free to individual practitioners, OpenAI entrenches its models into clinical habits, leaving hospitals little choice but to follow along.

We think tying clinical searches directly to CME credits is brilliant product design. The lingering question, however, is liability. A 99.6% safety rating on 6,900 test cases looks impressive, but edge cases in real-world clinical care carry massive legal and physical risks. If a clinician leans too hard on a synthesized summary during a complex consult, the fallout will still land squarely on the human in the white coat.