ChatGPT Health: What It Connects and How Privacy Works

OpenAI has moved ChatGPT Health from a separate early-access area into everyday conversations. The practical change is simple: with permission, ChatGPT can use selected health records, activity data and medication context without asking you to upload the same information each time.

The privacy details matter more than the headline. Connected sources, conversation history and ChatGPT memory follow different rules. This guide separates those layers and explains what users can control.

What launched

On 23 July 2026, OpenAI began rolling out Health to logged-in ChatGPT users aged 18 and over in the United States. It is available on web and iOS for Free, Go, Plus and Pro accounts. OpenAI says Health is not available in Codex.

The product can bring connected health context into ordinary ChatGPT conversations. That follows an early test in which OpenAI found that more than 70% of health-related conversations among people with access happened outside the dedicated Health area.

Health remains the place to connect accounts, review synced information, see recent trends and manage permissions. Users can also add @Health to a prompt when they specifically want connected context included.

Availability: US first, no UK date

The current launch is limited to adults in the United States. OpenAI has not published a release date for the United Kingdom, European Economic Area or Switzerland.

That distinction is important for UK readers. Screenshots and US launch coverage may make the feature look generally available, but a Health item in another user's sidebar does not mean the same connection options are available in Britain.

What ChatGPT Health can connect

OpenAI currently lists four connection routes: Apple Health, supported medical records from US hospital systems, One Medical and Function Health. Other wearables, fitness apps or nutrition services may contribute data through Apple Health when sharing is enabled.

Connected records can be incomplete or out of date. OpenAI gives the example of a medication that remains listed after someone stops taking it. Users should check important details against the original record and correct information when circumstances change.

How permission controls work

ChatGPT asks before using connected medical records and Apple Health information by default. A user can approve access for one response or choose to always allow it. The setting can be changed under Settings > Plugins > Health.

That is more useful than a single connection switch because it separates storing a connection from using its contents in a particular conversation. Someone can keep an account linked while deciding case by case whether its information belongs in a restaurant, exercise or travel discussion.

OpenAI also says it applies additional safeguards before another connected plugin takes an action that could disclose health information. Some sensitive actions may require confirmation.

Privacy, training and deletion

OpenAI says connected medical records, Apple Health information and conversations that use that connected information are not used to train its foundation models or target advertising. This rule applies regardless of the general model-training setting selected for ChatGPT.

The company says all ChatGPT conversations are encrypted at rest and in transit, while information connected in Health receives additional encryption protections. These are OpenAI's published assurances, not an independent security audit.

Disconnecting a source starts deletion of the data synced from that source, which OpenAI says is completed within 30 days. That does not automatically delete information already included in conversation history. Those conversations must be deleted separately.

The separate rule for ChatGPT memory

OpenAI says memories can be created from Health conversations, but not directly from connected medical records or Apple Health information. That means a discussion about a condition or goal could influence future chats even though the underlying record itself is not turned into a memory.

Users who do not want a health conversation to create memories can use Temporary Chat or turn memory off in Settings > Memory. This setting is separate from disconnecting a health-data source.

Where connected context may be useful

The strongest uses are organisational. ChatGPT can summarise changes between appointments, translate a visit note into plainer language, compare a new laboratory result with earlier tests, or prepare questions for a clinician.

It can also connect health context with everyday planning. OpenAI suggests examples such as considering a dietary restriction when choosing a restaurant, or adapting weekend activities around a recent injury.

These tasks benefit from continuity, but the output still depends on the quality and completeness of the source data. A confident summary can be wrong if a record is stale, a device did not sync or an important detail is missing.

Limits and risks to understand

ChatGPT Health is a product layer around an AI system, not a clinical record of truth. OpenAI says ChatGPT can still make mistakes and does not replace the care or judgement of qualified professionals.

The launch evidence is also mainly first-party. OpenAI reports extensive physician testing and improvements on health evaluations, but those claims should not be read as proof that the product is safe for diagnosis or treatment decisions in every real-world setting.

A sensible use is to organise information and improve the questions you take to a professional. Important medical decisions, urgent symptoms and changes to treatment should still be handled through appropriate clinical care.

Frequently asked questions

The bottom line

ChatGPT Health is most interesting as a permissions and context product. Its value comes from bringing scattered information into one conversation, while its risk comes from treating that conversation as more authoritative than the underlying data or professional care.

US users should review the permission setting, memory setting and deletion behaviour before connecting records. UK users should wait for an explicit availability announcement rather than assuming the American rollout applies locally.

Source: OpenAI's 23 July 2026 Health launch announcement. See our GPT-5.6 review for the model OpenAI identifies as its strongest current option for complex health questions.