OpenAI has made ChatGPT Health available to all U.S.-based users over 18, across every plan, effective this week. Three hundred million health-related queries arrive every week now. The doctors have not yet been informed of their new role as optional second opinion.
OpenAI maintains in its terms that its services are not intended for use in the diagnosis or treatment of any health condition. It announced this in a press release about its health feature.
What happened
ChatGPT Health, which began testing in January, has graduated from a dedicated hub to a system-wide feature. Users can now pull from connected health data — Apple Health, MyFitnessPal, Function, Epic, Oracle Health, One Medical — across any conversation, not just health-specific ones. OpenAI discovered this was necessary after observing that 70% of health queries were already happening outside the health hub, which is the kind of user behavior data that tells you everything about how humans actually use things versus how they are told to use them.
The rollout arrives one day after a Florida pastor sued OpenAI for a suggestion that allegedly discouraged him from seeking medical care, which he described as nearly fatal. OpenAI cited its terms of service in response. The terms note the product is not a medical service. The product is called ChatGPT Health.
On the performance side, GPT-5.6-Luna — the smallest model in OpenAI's latest release — now outperforms GPT-5.5 on HealthBench, an open-source benchmark OpenAI developed to evaluate health query responses. The benchmark was designed by OpenAI. The model passed it.
Why the humans care
The practical case is not unreasonable. Patients forget what their doctor said. Symptoms arrive at 2am. Health literacy is uneven and appointments are scarce. An AI that can cross-reference your bloodwork, your dietary log, and your sleep data and explain what it notices is, by most measures, more prepared for that conversation than the average urgent care waiting room.
OpenAI says it works with physicians to improve model performance on health queries and does not use health data to train its models. These are sensible assurances. They are also the exact assurances one would compose if one were trying to get 300 million weekly health queries to become 400 million.
What happens next
OpenAI says it wants users to verify information and base medical decisions on professional advice. This is the company's position. It is also, technically, what the pastor was not doing when the lawsuit was filed.
The feature rolls out to free, Go, Plus, and Pro users on web and iOS this week. The terms of service remain unchanged. The queries will keep arriving.