OpenAI Launches Health in ChatGPT for Connected Health Data

Quick answer

OpenAI has announced Health in ChatGPT, a feature rolling out to users in the United States that can connect selected health information to conversations. The aim is to help people understand and navigate their own health context. It is not a replacement for a clinician, emergency service, diagnosis or treatment plan.

What Health in ChatGPT is designed to do

The feature can ground health-related conversations in information a user chooses to connect. OpenAI’s announcement describes support for connected health information, wearables and other health apps, along with tools to review and update relevant details.

The potential benefit is context. A general answer about sleep, exercise or a laboratory value may be more useful when the assistant can consider a person’s recent data and stated history. It may also help users prepare questions for a medical appointment or understand terminology in a record.

However, connected context does not guarantee a correct interpretation. Medical data can be incomplete, delayed or affected by device accuracy. A model may also misunderstand a record or miss information that a clinician would consider important.

Privacy and control

OpenAI says the feature is built around privacy, security and user control. Those principles are especially important because health information can be highly sensitive and may reveal details about a person’s conditions, medication, fertility, mental health or daily routine.

Before connecting data, users should understand which records are included, how access can be revoked and whether data from a wearable is interpreted as a clinical measurement or a consumer estimate. They should also avoid sharing another person’s health information without permission.

What the feature should not be used for

Health in ChatGPT should not be treated as an emergency diagnostic system. Symptoms such as severe chest pain, difficulty breathing, signs of stroke, heavy bleeding or immediate risk of self-harm require urgent professional help rather than an AI conversation.

The feature should also not be used to change prescription medication or ignore a clinician’s advice without medical review. Even when an explanation sounds personalised, the model cannot perform a physical examination and may not have the complete clinical picture.

How users can get value safely

Lower-risk uses include translating medical language into plain English, preparing a list of questions, organising a timeline of symptoms and comparing a wearable trend with the user’s own notes. The user should verify important information against the original record and a qualified professional.

It is also useful to state uncertainty explicitly. Asking “What could this result mean, what information is missing, and when should I contact a clinician?” is safer than asking for a definitive diagnosis.

TOOLSAURA analysis

Connected health assistants may reduce the friction between raw data and understandable questions. Their value will depend on careful privacy design, transparent limitations and strong evaluation with clinicians and diverse patient groups.

The most important measure is not how fluent the response sounds. It is whether the system accurately represents the underlying information, communicates uncertainty and directs users to appropriate care when risk is high.

The US-only rollout also means availability, connected providers and legal protections may differ by state, device and account. Users outside the United States should not assume the feature is available.

What to watch next

Watch for independent safety evaluations, the range of supported health-data connections, controls for deleting or disconnecting information and evidence about how the feature performs across different populations.

This article is an independent explanation of OpenAI’s announcement and is not medical advice. TOOLSAURA did not access private patient data or conduct a clinical evaluation of the feature.