OpenAI Launches ChatGPT Health: What Healthcare Teams Should Know
OpenAI has introduced ChatGPT Health, a secure environment inside ChatGPT that connects personal health data with AI to help people manage their well-being. It's meant to support care-not replace clinical judgment-with a focus on preparation, education, and coordination.
The move meets current behavior at scale: OpenAI says hundreds of millions of health questions are asked each week. Now, users can link select medical records and wellness apps to get context-aware responses inside a protected space.
What ChatGPT Health does (and does not do)
- Helps patients prepare for appointments with questions, summaries, and organized notes.
- Explains labs and reports in plain language, with prompts to follow up when appropriate.
- Surfaces insurance basics and options for next steps.
- Highlights trends across activity, nutrition, sleep, and other connected data.
It does not diagnose, prescribe, or replace clinical care. Think of it as an assistant that improves pre-visit readiness, post-visit comprehension, and ongoing self-management.
Data connections and current scope
Users can link medical records and apps such as Apple Health, Function, and MyFitnessPal to view information in one place. Integration with medical records and some app connections are available in the United States only at this time.
For context on Apple Health, see Apple's overview of its Health app here.
Privacy and security: A separate, encrypted health space
Health data lives in an isolated section of ChatGPT. Conversations, files, and app connections in this space are kept separate from regular chats and are not used to train AI models.
Encryption is on by default. Users can enable multi-factor authentication, and ChatGPT prompts people to switch into the Health space when a conversation turns medical to keep it protected.
Built with clinician input
Mira Murati, OpenAI's CTO, says the team collaborated with physicians across sixty countries. More than 260 doctors provided over 600,000 pieces of feedback to refine how ChatGPT Health communicates medical information.
The priority: accuracy, clarity, and safety-especially around when to suggest follow-ups with a clinician.
How performance is evaluated
OpenAI uses HealthBench, a framework developed with practicing clinicians. Instead of test-style quizzes, responses are scored against physician-written rubrics that reflect real-world standards like context, urgency, and clarity.
For more detail, watch OpenAI's updates on its blog here.
Data control and third-party tools
- Apps must request explicit permission before accessing health data-even if they've been connected to ChatGPT before.
- Third-party tools in the Health space undergo extra security review and must meet strict privacy criteria.
- Users can revoke access to medical records and apps at any time in settings.
Availability
ChatGPT Health is rolling out in stages to a limited number of users across Free, Go, Plus, and Pro plans. It is not yet available in the European Economic Area, Switzerland, or the United Kingdom.
Medical record integrations and certain app connections are currently limited to the United States.
Practical takeaways for healthcare organizations
- Position it as a pre-visit and post-visit companion. Use it to help patients form better questions, summarize discharge instructions, and understand labs-without crossing into clinical decision-making.
- Pilot with clear guardrails. Start with low-risk workflows (education, prep, logistics). Keep PHI handling within the Health space and review outputs for clarity and appropriateness.
- Involve compliance and security early. Map data flows, confirm permission boundaries, document revocation processes, and enable MFA by default.
- Train staff on what the tool can and cannot do. Encourage clinicians to treat it like a communication aid, not a source of diagnosis or treatment.
- Measure outcomes that matter: appointment readiness, fewer clarification calls, improved patient recall, and time saved on routine explanations.
If you're preparing your team
Build internal guidelines, run tabletop tests with de-identified scenarios, and educate patients on where the Health space begins and why it matters. If you need structured upskilling on practical AI use in clinical and operational workflows, explore our curated programs here.
Your membership also unlocks:
