OpenAI is rolling out ChatGPT Health to every U.S. user over 18, across free, Go, Plus, and Pro plans on web and iOS. It began testing the feature in January through a dedicated hub where users could pull in data from Apple Health, MyFitnessPal, Epic, Oracle Health and other platforms. Back then, ChatGPT handled about 230 million health queries a week; now that’s up to 300 million. During testing, 70% of those questions happened outside the health hub, so OpenAI let people tap their connected medical info in any chat—whether they’re asking about food choices, allergies or workout plans. Under the hood, OpenAI says its smallest new model, GPT-5.6 Luna, outperforms the prior GPT-5.5 on HealthBench, the company’s open-source health benchmark. The company claims it doesn’t train on user data and that it works with physicians to boost accuracy. Despite those gains, the terms still forbid using ChatGPT Health for diagnosis or treatment. That disclaimer factored into OpenAI’s response to a lawsuit from a Florida pastor who says ChatGPT nearly steered him away from seeing a doctor. OpenAI tells users to verify any medical advice with a professional, even as rivals like Anthropic and Google push ahead with their own AI health tools.