ChatGPT Health opens to every US adult. Clinician-level claims meet a lawsuit

From 23 July 2026 OpenAI is expanding Health in ChatGPT to every signed-in US user aged 18+ on web and iOS, across free, Go, Plus, and Pro. The Verge and TechCrunch describe a product that connects Apple Health and supported hospital records so the chatbot can use personal medical context. OpenAI’s primary page was behind a Cloudflare challenge during verification, so details rest on official index copy in metadata plus secondary reporting.

Health leaves the waitlist and pulls records into ordinary chat

The feature began in a narrower January 2026 test. The US rollout now covers Apple Health, records from hospital systems such as Epic and Oracle Health, and platforms like One Medical and Function Health. OpenAI earlier cited about 230 million health-related queries a week. TechCrunch reports that figure at 300 million.

A key product shift: during testing, OpenAI said about 70% of health queries happened outside the dedicated health hub. Connected health context can now inform general chat, including food or allergy questions. The company says smaller GPT-5.6-Luna beats GPT-5.5 on its HealthBench eval, that it works with physicians on health answers, and that it does not train on Health user data.

For a US patient the entry point moves. For the EU it mostly does not

For an American with a hospital portal and Apple Health, ChatGPT becomes a place where scattered labs, meds, and wearables meet. That is a distribution play: health is a layer on the chat people already open, not a separate app. Ashley Alexander, VP of health product, told The Verge the models can now reason at levels better than clinician level. OpenAI still says Health supports care and does not replace it, and its terms still bar diagnosis and treatment.

EU and Czech readers face a different map. Medical-record hooks and some app integrations are reported as US-only. Early access previously excluded the EEA, Switzerland, and the UK. Even if global chat improves on generic health answers, the personalized records channel is not a European product yet.

“Better than a clinician” sits beside a lawsuit and a terms disclaimer

A day before the broader rollout, TechCrunch notes a Florida pastor’s lawsuit alleging ChatGPT gave a near-fatal suggestion not to see a doctor. OpenAI points to its terms and says it is working to make health answers safer. Independent studies and rival launches show the same pattern: labs ship health features while chatbot medical reliability stays contested.

The clinician-level line therefore hits two brakes at once. First is legal and product design: the model may help orient someone in their records, but it must not act like the treating clinician. Second is operations: the more data a person connects, the more consent, revocation, train-time separation, and handling of incomplete or wrong records matter.

Watch incidents, audits, and whether Health leaves the US

Three signals matter. First, concrete incident reports and court filings, not only HealthBench scores. Second, whether OpenAI publishes independent evals beyond its own benchmark and how fast health answers escalate to “see a clinician” without false certainty. Third, geography: an EEA timeline and hooks outside US hospital networks. Until Health stays a US-only channel with a strong claim and thinner liability, it is a race for data and habit more than finished clinical infrastructure.

Lilith's verdict

OpenAI is sliding hospital charts into the same window where people write a shopping list. A “better than clinician” claim then has to survive HealthBench and the first court file after bad advice.