OpenAI is rolling out Health in ChatGPT to logged-in users age 18 and older in the United States, on web and iOS, across Free, Go, Plus and Pro plans. With explicit permission, users can connect supported U.S. hospital medical records, One Medical or Function Health, and Apple Health data from an iPhone.
Per the official July 2026 announcement — widely recapped in autumn product roundups — more than 300 million people ask ChatGPT health-related questions each week. The problem, OpenAI says, is that context is scattered across patient portals, apps and wearables; Health aims to gather that context when the user authorizes it.
With permission, the assistant can help compare a new result with prior tests, summarize changes since the last appointment, or explore how sleep and activity relate to a routine. OpenAI stresses the product supports, rather than replaces, clinicians’ judgment, and encourages people to verify important information with their provider.
An early-testing finding changed the design: more than 70% of health conversations among people with access happened outside the dedicated health space. Users can therefore let ChatGPT use connected Health information in everyday chats — for example when planning meals around a dietary restriction — without forcing a separate flow.
Health remains in the sidebar as the home for connecting accounts, viewing trends, exploring prompts and managing sync. By default, ChatGPT asks permission before using medical records or Apple Health in a reply; users can approve case by case or “always allow,” and change the setting anytime.
OpenAI claims additional encryption layers and says connected medical records and Apple Health data — and conversations that use them — are not used to train foundation models or target ads. When a source is disconnected, synced data is deleted from its systems within a stated 30 days; chat history already created remains until the user deletes it.
On model intelligence, GPT-5.5 Instant brought health capabilities to the free plan; GPT-5.6 Sol, for paid users, is presented as strongest on complex health tasks. OpenAI says it works with hundreds of physicians on scenarios and rubrics (accuracy, safety, communication, escalation to urgent care) and that physicians tested Health before release.
Connecting Apple Health means data from wearables or apps that already share with Apple Health can flow into ChatGPT when the user allows it; not all proprietary scores transfer. After syncing, users can review conditions and medications, correct outdated items and add family history.
The product is not available in Codex and, per the announcement, not yet outside the U.S. Regulators and clinics will watch consent closely, the accuracy of synced records, and the risk that users treat generated summaries as medical advice.
For hospitals and insurers, the competitive lesson is clear: if patients bring context to a generalist assistant, first-party portals must offer clarity and exportability, not just friction. For OpenAI, the challenge is proving that permission and encryption safeguards hold trust as scale grows.
Neural Pulse summarizes the rollout with reported facts: U.S., 18+, web/iOS, Free/Go/Plus/Pro, supported records and Apple Health, with no clinical-advice tone. Anyone using the feature should still cross-check with their care team.
In a year marked by agent incidents and self-policing pledges, Health shows the other face of the frontier: consumer products that touch intimate data. Consent quality and how errors are explained will matter as much as the model that answers.
