ChatGPT for Clinicians Release

OpenAI has introduced ChatGPT for Clinicians, a specialized version of the platform designed to reduce administrative burden and support clinical tasks such as medical research and documentation. The tool is currently available for free to verified physicians, nurse practitioners (NPs), physician assistants (PAs), and pharmacists in the United States.

Clinical Capabilities and Workflow Integration

ChatGPT for Clinicians provides a suite of tools specifically engineered for healthcare workflows, developed in collaboration with hundreds of physician advisors.

Core Features

• Trusted Clinical Search: Provides real-time, cited answers based on evidence from millions of peer-reviewed medical sources to accelerate case reasoning.
• Deep Medical Research: Enables clinicians to delegate literature reviews to the AI, allowing them to specify trusted sources and receive comprehensive, cited reports.
• Repeatable Clinical Skills: Allows users to convert common workflows—such as drafting referral letters, prior authorizations, and patient instructions—into reusable skills for consistent execution.
• Frontier Model Access: Provides free access to current frontier models to handle complex clinical questions and documentation more reliably.
• CME Integration: Eligible evidence reviews conducted within the platform can automatically count toward continuing medical education (CME) credits.

Privacy and Compliance

• HIPAA Support: While many tasks do not require Protected Health Information (PHI), HIPAA compliance is available for eligible accounts via a Business Associate Agreement (BAA).
• Data Security: Conversations are not used to train OpenAI's models, and the platform employs multi-factor authentication to secure sensitive work.

Performance Evaluation and HealthBench Professional

To ensure safety and accuracy, OpenAI developed HealthBench Professional, an open benchmark for real clinician chat tasks focusing on care consult, writing/documentation, and medical research.

Validation Metrics

• Physician Review: Physician advisors have reviewed over 700,000 model responses to date. Before the release of ChatGPT for Clinicians, advisors tested 6,924 conversations, rating 99.6% of responses as safe and accurate.
• Citation Accuracy: In a test of 355 examples with ground-truth citations specified by three independent physicians, ChatGPT for Clinicians cited sources more frequently than human physicians.
• Comparative Performance: Using HealthBench Professional, OpenAI reports that GPT-5.4 in the ChatGPT for Clinicians workspace outperforms base GPT-5.4, other OpenAI models, external models, and human physicians (who were given unbounded time and web access).

Benchmark Design

HealthBench Professional utilizes physician-authored conversations and rubrics, multi-stage physician adjudication, and rigorous data filtering. Approximately one-third of the examples include "red teaming" attempts to find model failures, with the most difficult conversations selected by a factor of 3.5x to ensure a high bar for measurement.

Availability and Global Expansion

Currently, the free version of ChatGPT for Clinicians is limited to verified healthcare providers in the U.S. OpenAI plans to expand access to other countries and groups, starting with a pilot for verified clinicians outside the U.S. in partnership with the Better Evidence Network, subject to local regulations.

Alongside this release, OpenAI has published a Health Blueprint, providing recommendations for the responsible integration of AI within the U.S. healthcare system.