ChatGPT launches Health for U.S. users

The launch follows earlier testing of a dedicated health space that required users to visit a separate part of ChatGPT. That design added an unnecessary step, as more than 70% of health-related conversations naturally occurred elsewhere. OpenAI has now integrated health context into the main chat experience, allowing the model to draw on connected data like medications, lab results, and sleep patterns when answering everyday questions.

To improve the underlying models, OpenAI works with hundreds of physicians to create realistic health scenarios and scoring rubrics, measuring factors like accuracy, safety, and communication. The models evaluated, GPT-5.5 Instant and the paid GPT-5.6 Sol, outperform previous versions on the company's HealthBench Professional assessments. Physicians also tested the system before release to gauge real-world performance with connected health data.

The program's success appears to hinge on user trust in privacy safeguards and the reliability of the information provided. While OpenAI emphasizes that connected medical records and Apple Health data are excluded from model training and ad targeting, it also cautions that ChatGPT can still err and does not replace professional medical care. As the feature reaches a broader audience, the critical test is whether it can deliver accurate, helpful insights that encourage users to verify details and consult their healthcare providers, rather than relying on the AI as a definitive source.