OpenAI Says Its New ChatGPT for Doctors Outperforms Humans in Clinical Tasks

In brief

• OpenAI introduced ChatGPT for Clinicians, a specialized version designed to support clinical tasks like documentation and medical research.
• The platform is free for verified U.S. physicians, nurse practitioners, physician assistants, and pharmacists.
• OpenAI said the chatbot outperformed human doctors in certain clinical tasks, according to its own benchmark test.

OpenAI on Wednesday unveiled a free, specialized version of ChatGPT for physicians, nurse practitioners, physician assistants, and pharmacists, deepening the company's push into a healthcare sector grappling with staffing shortages, administrative overload, and burnout.

The product, called ChatGPT for Clinicians, is designed to handle documentation, medical research and care consultations—time-consuming tasks that eat into the hours clinicians can spend with patients. Access is currently limited to verified practitioners in the United States, with plans to expand internationally.

The announcement arrives as AI adoption in medicine is surging. According to a 2026 survey by the American Medical Association cited by OpenAI, 72% of physicians now use AI in clinical practice, up from 48% just a year ago. The company says clinician usage of its own platform has more than doubled over the past year, with millions relying on ChatGPT weekly.

Among the tool's features are a clinical search function drawing on millions of peer-reviewed sources, a deep research mode for medical literature reviews, reusable workflow templates for tasks such as referral letters and prior authorization requests, and the ability to earn continuing medical education credits while researching clinical questions in the platform.

Conversations will not be used to train OpenAI's models, and HIPAA compliance support is available through a Business Associate Agreement for eligible accounts.

Alongside the launch, OpenAI released HealthBench Professional, a new benchmark designed to evaluate AI performance on realistic clinical tasks across three categories: care consultations, documentation, and medical research.

The company reported that GPT-5.4, running in the ChatGPT for Clinicians workspace, scored 59.0 on the benchmark—higher than human physicians, who scored 43.7 even with unlimited time and internet access, and higher than competing models from Anthropic, Google, and xAI.

Those results, however, come with an important caveat: OpenAI built both the product and the benchmark used to evaluate it.

To develop the tool, the company says it worked with hundreds of physician advisors and reviewed more than 700,000 model responses. In pretesting, physicians rated 99.6 percent of responses as safe and accurate across nearly 7,000 conversations.