OpenAI Links ChatGPT to Apple Health and Medical Records

OpenAI launched a Health feature that lets eligible U.S. users age 18 and older connect Apple Health and supported medical records to ChatGPT on the web and iOS. The feature is available across Free, Go, Plus and Pro tiers. Once users grant permission, ChatGPT can access medications, lab results, recent visits, sleep and activity data and use that information in any conversation within the app.

The change follows an earlier pilot in which users had to open a separate health section to get answers based on their records. OpenAI reported that participants in that test carried out more than 70% of health-related conversations outside the dedicated area. The company adjusted the design so connected health data can be used across all chats when a user allows it. The Health sidebar remains as a place to connect accounts, review synced data and trends, browse suggested prompts and return to prior health chats.

Users can link Apple Health and, where supported, records from participating U.S. hospital systems, One Medical and Function Health. By default ChatGPT asks before using connected health data to personalise a reply; users may set an “always allow” option and can revoke access at any time. Disconnecting a data source triggers deletion from OpenAI’s systems within 30 days, while any health information already visible in chat history remains until the user deletes those conversations manually.

OpenAI said connected medical records and Apple Health data, along with any conversations that draw on them, are excluded from foundation model training and from ad targeting, regardless of a user’s broader training settings. The company also said memories can be formed from health conversations but not directly from raw synced records; users can avoid memory creation by using Temporary Chat or disabling memory in settings. OpenAI identified an edge case where sharing content generated from health data through other plugins might expose sensitive information and said extra checks apply before such actions proceed.

OpenAI linked the feature’s capabilities to newer models. The company named GPT-5.5 Instant, available to Free users, and GPT-5.6 Sol, for paying subscribers, as offering improved recognition of urgent-care needs and better reasoning across multiple data points. OpenAI said it worked with hundreds of physicians to build test scenarios and rubrics and that GPT-5.6 Sol outperformed earlier models on an internal evaluation called HealthBench Professional. The company has not published detailed methodology or full results of those tests.

Early testers reported practical benefits and limits. A technical program manager described how the tool turned scattered records into timelines and plain-English summaries that could be shared with clinicians. A portfolio manager highlighted the ability to compare lab trends over time. A nurse reported finding an unexpected chart entry through the tool and using that discovery to follow up with providers; she noted the tool helped surface an issue rather than make clinical judgments. Other testers reported the feature reduced manual data exports and made fitness and activity trends easier to review.

OpenAI acknowledged data quality and completeness are practical concerns. Synced records can be out of date, such as medications that remain listed after a patient stops taking them, and wearable data depends on what third-party apps share through Apple Health. The company advised users to flag changes in ChatGPT and to verify important information against original records or with healthcare providers.

OpenAI reported more than 300 million weekly health-related ChatGPT queries and said the Health feature is live now for eligible U.S. users via the ChatGPT sidebar.