Answer First
ChatGPT Health is a separate place in ChatGPT for health questions, optional connected records, and health conversations. You can ask a health question without connecting anything. If you do connect medical records or Apple Health, OpenAI says ChatGPT asks before using that connected data in a conversation.
For the connected sources and conversations that use them, OpenAI says:
- the data is not used to train its foundation models;
- the data is not used to target ads;
- you can allow connected-data access once or always;
- you can change that permission later;
- disconnecting a source starts deletion of its synced source data from OpenAI systems, with deletion completed within 30 days;
- information already copied into conversation history stays there until you delete the conversation.
Those statements do not make ChatGPT a doctor, guarantee that a record is complete, or establish that the service is suitable for every legal or clinical workflow. ChatGPT Health supports—not replaces—professional care. Verify important information against the original record and discuss medical decisions with a qualified clinician.
Need the setup steps instead? Use the separate ChatGPT Health connection guide. This page owns the privacy, availability, permission, memory, and deletion decision.
Source check: This guide was checked against OpenAI’s official Health launch announcement and product release notes on July 30, 2026. Product availability and controls can change, so confirm the live signed-in interface before relying on a setting.
Availability at Launch
OpenAI announced Health in ChatGPT on July 23, 2026 with this initial availability:
| Check | Official launch position |
|---|---|
| Region | United States |
| Age | 18 and older |
| Account | Logged-in ChatGPT users |
| Plans | Free, Go, Plus, and Pro |
| Platforms | ChatGPT on web and iOS |
| Codex | Not yet available |
This is a dated availability snapshot, not a promise that every account or country has the same product surface. If Health does not appear, verify the signed-in account, age eligibility, region, platform, app version, and the latest OpenAI release note. Do not use a third-party screenshot or an old rollout article as current access truth.
What You Can Do Without Connecting Data
A connection is optional. OpenAI says users can ask health questions in Health without connecting a medical record or Apple Health.
That distinction matters for privacy planning:
- No connection: ChatGPT works from the information you type or upload in the conversation.
- Connected source, not used in the conversation: The source may be synced, but OpenAI says permission is requested before ChatGPT uses connected data.
- Connected source used in the conversation: The connected-data privacy rules described by OpenAI apply to that source data and the conversation using it.
Do not assume that opening Health grants every conversation unrestricted access to every connected source. Review the permission prompt and the Health settings for the account.
What Can Be Connected
At launch, OpenAI listed these connection paths:
- Apple Health
- Supported medical records from U.S. hospital systems
- One Medical
- Function Health
Fitness, nutrition, sleep, and wearable apps may share data through Apple Health if the user enabled sharing into Apple Health. The information available to ChatGPT depends on what each app writes to Apple Health and what the user permits. Some proprietary scores or app-only insights may not transfer.
Connection does not prove completeness. A synced record can be missing, delayed, duplicated, or out of date. Compare medications, conditions, allergies, dates, and results with the original provider portal or source app before using them in an important conversation.
How Permission Works
OpenAI says ChatGPT asks for permission before using connected medical records or Apple Health data in a conversation. The prompt can offer two choices:
- Allow once: use the connected data for that request.
- Always allow: let ChatGPT use it in later Health conversations without asking each time.
You can also type @Health in a Health conversation to explicitly include connected health data. OpenAI points users to Settings > Plugins > Health to review or change the permission.
Use the narrowest permission that matches the task. “Allow once” is a practical default when you want to review the request each time. “Always allow” reduces prompts, but it should be a deliberate account-level choice—especially on a shared or managed device.
OpenAI also says additional safeguards apply before information is shared with other plugins, including confirmation in some cases. Still, read every prompt: a permission dialog is the point where the user can see and control the proposed data flow.
Training and Advertising: The Important Boundary
OpenAI’s launch announcement draws a specific line:
| Data or conversation | OpenAI’s stated treatment |
|---|---|
| Connected medical records | Not used to train foundation models or target ads |
| Connected Apple Health data | Not used to train foundation models or target ads |
| Conversations that use connected Health data | Not used to train foundation models or target ads |
| Conversations that do not use connected Health data | Follow the account’s standard model-training setting |
The final row is easy to miss. Merely being in a health-related conversation does not mean every conversation uses connected Health data. If you type sensitive details without using a connection, review the account’s normal Data Controls and choose the conversation mode accordingly.
This guide does not translate OpenAI’s product statements into a claim about HIPAA, a particular country’s health-data law, an employer policy, research consent, or a provider’s recordkeeping obligations. Organizations should evaluate their own legal, security, procurement, and clinical-governance requirements.
Memory Is Not the Same as Connected Data
OpenAI describes another boundary:
- ChatGPT can create memories from Health conversations when memory is enabled.
- OpenAI says memories are not created directly from connected medical records or Apple Health data.
- Information discussed in a conversation may still become a memory based on the account’s memory controls.
If you do not want a health conversation to influence later chats, use Temporary Chat or turn memory off before the conversation. Then verify the account’s saved memories and conversation history afterward. A memory control, a connected-source permission, and conversation deletion are three separate controls.
Disconnecting and Deleting Data
OpenAI points users to Health > Accounts to disconnect a source. According to the launch announcement:
- Disconnecting stops the account connection.
- Synced data from that source is deleted from OpenAI systems within 30 days.
- Health information already included in a conversation remains in conversation history.
- Delete the relevant conversations if you also want to remove that retained conversation content.
This means “disconnect” and “delete conversation” solve different problems. If you are leaving Health or cleaning up an account, do both checks:
- confirm that each source shows disconnected;
- search conversation history for chats that used the source and delete the ones you no longer want retained.
Do not promise immediate deletion when the official statement allows up to 30 days for synced source data.
Security Statements and Their Limits
OpenAI says Health data is encrypted at rest and in transit and that connected health information receives additional protections. These are meaningful product statements, but they do not remove ordinary account-security risks.
Before connecting:
- use a unique password and enable available account protections;
- avoid connecting on a shared device or shared ChatGPT account;
- review who can access the source account and the ChatGPT account;
- check employer, school, insurer, clinical, and research policies before using managed data;
- connect only the sources needed for the intended task;
- review the permission each time it appears.
Do not paste another person’s medical information unless you have a valid reason and the authority to handle it.
A Practical Privacy Decision Checklist
Use this sequence before connecting:
- Confirm eligibility. Check region, age, account, platform, and current rollout.
- Define one task. For example, summarize changes between two lab reports or prepare questions for an appointment.
- Decide whether a connection is necessary. If typed context is enough, do not connect merely for convenience.
- Check the original data. Correct medication, condition, date, and result errors at the source.
- Choose permission duration. Prefer one-time access unless recurring access is clearly useful.
- Review memory and Data Controls. Connected-data rules, standard training settings, memory, and chat history are different controls.
- Plan cleanup. Know how to disconnect sources and delete conversations before you start.
- Keep a human in the loop. Verify important information and discuss decisions with a qualified professional.
Safe Use and Medical Limits
OpenAI positions Health as support for understanding information and preparing for care—not as a replacement for professional diagnosis or treatment. ChatGPT can make mistakes, misunderstand context, or reason from incomplete records.
Appropriate tasks can include:
- organizing questions for a scheduled appointment;
- summarizing changes to verify against original reports;
- explaining unfamiliar terms in plain language;
- identifying information that may be missing from a personal record;
- reviewing sleep, activity, or workout trends without turning them into a diagnosis.
Do not use ChatGPT as the only source for urgent symptoms, medication changes, diagnosis, or treatment. For urgent or severe symptoms, contact local emergency services or a qualified medical professional immediately.
FAQ
Does ChatGPT train on connected medical records or Apple Health data?
OpenAI says no: connected medical records, Apple Health data, and conversations that use that connected data are not used to train its foundation models, regardless of the account’s normal model-training setting. Conversations that do not use connected Health data follow the standard account setting.
Does ChatGPT Health automatically read connected data?
OpenAI says ChatGPT asks for permission before using connected data in a conversation. You can allow it once or always, change the setting later, or explicitly include Health with @Health.
What happens when I disconnect a source?
OpenAI says synced source data is deleted from its systems within 30 days. Information already included in conversation history remains until you delete the conversation.
Can ChatGPT Health diagnose a condition?
Do not treat it as a diagnostic service. OpenAI says Health supports rather than replaces professional care, and ChatGPT can make mistakes. Verify important information and discuss medical decisions with a qualified clinician.
Is ChatGPT Health available outside the United States?
At the July 23, 2026 launch, OpenAI described availability for logged-in U.S. users age 18 and older on web and iOS. Check the current official release notes and signed-in product for later expansion.
Bottom Line
The safest mental model is: connection is optional, use of connected data is permissioned, and cleanup has multiple layers. Review connected-source access, standard Data Controls, memory, conversation history, and source disconnection separately.
If the privacy boundary is acceptable and your account is eligible, continue with the step-by-step ChatGPT Health setup guide. Keep the original source records and a qualified medical professional as the authority for important decisions.