OpenAI began making its ChatGPT Health feature available to all eligible users in the United States on Thursday. The tool permits users to connect personal health data, including records from Apple Health and select medical providers, directly to the chatbot. This integration allows ChatGPT to draw upon a user's health information across any conversation, not just within a dedicated health section.
During a press briefing, Ashley Alexander, OpenAI's vice president of health product, stated that the company's models are now "capable of reasoning at levels that are better than clinician level." However, Karan Singhal, OpenAI's health lead, later tempered this claim, suggesting that "there have been individual studies that have been pointing in that direction," and citing research from Harvard and Stanford. He further indicated that while the models show improved performance on health-related tasks, they have not definitively surpassed all clinicians in all contexts.
The expanded rollout includes users aged 18 and older across free, Go, Plus, and Pro plans on web and iOS. Users can link data from sources such as Apple Health, One Medical, and Function Health. OpenAI stated that connected medical records and Apple Health data, along with any conversations that use them, are excluded from foundation model training and advertising targeting. Users retain control over which data sources are connected and when ChatGPT can use that information, with the system defaulting to asking for permission before personalizing a response. Disconnecting a data source prompts deletion from OpenAI's systems within 30 days.
OpenAI attributes the feature's capabilities to its newer models. GPT-5.5 Instant, available to free users, has reportedly made gains in identifying urgent care needs and explaining uncertainty. Paid users have access to GPT-5.6 Sol, which OpenAI describes as its strongest health model to date. The company reports that GPT-5.6 Sol demonstrates stronger performance in tasks requiring reasoning across complex details, clear communication, and careful judgment. OpenAI claims that every GPT-5.6 model outperformed GPT-5.5 on HealthBench Professional, an internal evaluation benchmark for AI performance on health tasks. Physicians were involved in developing health scenarios and rubrics for this evaluation.
The feature's integration into general conversations stems from feedback during an earlier test phase. OpenAI observed that more than 70% of health-related conversations occurred outside a dedicated health section. This led to the decision to allow ChatGPT to reference health data contextually within any chat.
The launch occurs amidst legal challenges for OpenAI. The company faces lawsuits alleging that ChatGPT has provided dangerous or inaccurate medical advice. One such lawsuit, filed recently, claims that ChatGPT gave a Florida pastor "extremely dangerous medical recommendations" that delayed treatment for a pulmonary embolism, nearly resulting in death. The plaintiff in this case has requested an injunction to halt the operation of ChatGPT Health until its safety is independently verified. OpenAI maintains that ChatGPT Health is designed to support, not replace, professional medical care and is not intended for diagnosis or treatment. The company's terms of service also state that the chatbot is not a substitute for professional medical advice.