ChatGPT Health Connects to Electronic Health Records and Apple Health, Bringing Personal Health Data into the Model Context

OpenAI began expanding the rollout of ChatGPT Health on July 23, connecting its consumer chat interface to users’ longitudinal health data. Free, Go, Plus, and Pro users aged 18 or older in the United States can link healthcare provider accounts and Apple Health on the web or iOS. They can view lab results, medications, visit records, activity, and sleep data, then use that information as context to ask about trends, prepare questions for medical appointments, or understand reports. When appropriate, responses can cite synchronized medical records instead of relying solely on excerpts manually pasted by the user.

Electronic health record connectivity is provided by b.well. The company says its data pipeline must process nearly 1,000 EHR systems, performing steps such as unit standardization, duplicate medication reconciliation, and medical coding. This normalization layer is technically critical because uncleaned, heterogeneous records can easily cause a model to confuse the units or timing of the same test or treat duplicate entries as separate events. After a user disconnects a source, ChatGPT can no longer access it, and the associated records are also deleted from b.well.

OpenAI says linked medical records, Apple Health data, and conversations that use them will not be used to train foundation models or serve advertisements, and health apps must undergo additional security reviews. However, these commitments cannot eliminate model hallucinations, errors in source data, or synchronization delays. OpenAI also positions the product as a tool for understanding and preparation, not diagnosis or treatment. For developers and health IT teams, the next areas to watch are whether citations can precisely identify original events, whether deletion covers caches and derived memories, and the verifiable error rates of cross-EHR normalization for medications, lab units, and time-series data.