OpenAI Launches ChatGPT Health across the US: Connect EHR & Apple Health Now

OpenAI has officially launched ChatGPT Health nationwide across the United States. Rolling out to all logged-in adult users (aged 18+) across Free, Go, Plus, and Pro tiers on Web and iOS, the feature unifies fragmented personal health data—including Electronic Health Records (EHRs) and consumer wearables—directly inside conversational AI. Every week, over 300 million people ask ChatGPT health-related questions; this release grounds those queries in a patient’s authentic, personal medical history.

Key Takeaways (TL;DR)

  • U.S. Nationwide Rollout: Accessible to logged-in users 18+ across Free, Go, Plus, and Pro tiers on Web and iOS.

  • Supported Integrations: Connects Apple Health, MyFitnessPal, Function Health, One Medical, and major EHR providers including Epic and Oracle Health.

  • Everyday Chat Context: Users can allow ChatGPT to draw on health data across routine conversations (e.g., meal planning or injury-aware workout routines) or invocation via @Health.

  • Model Infrastructure: Powered by GPT-5.5 Instant (Free) and GPT-5.6 Sol (Paid), evaluated alongside 260+ practicing global physicians.

  • Zero-Training Guarantee: Protected health data and connected conversations are strictly excluded from AI model training and ad targeting.

Understanding ChatGPT Health: Features, Rollout, and Architecture

ChatGPT Health is an intelligent aggregator that bridges disconnected medical records and fitness metrics into a single conversational interface. Rather than requiring patients to log into multiple hospital portals or search through PDF lab results, ChatGPT Health standardizes and contextualizes data to help users prepare for doctor visits, interpret lab work, and manage daily wellness goals.

┌──────────────────────────────────────────────────────────┐
│                Connected Health Sources                  │
│   • Wearables & Apps (Apple Health, MyFitnessPal)        │
│   • Clinical Records (Epic, Oracle Health, One Medical)  │
└────────────────────────────┬─────────────────────────────┘
                             │
                             ▼
┌──────────────────────────────────────────────────────────┐
│        b.well FHIR API & Data Refinery Layer             │
│   • Standardizes clinical records & fitness metrics      │
│   • Manages identity, OAuth consent, & auditability      │
└────────────────────────────┬─────────────────────────────┘
                             │
                             ▼
┌──────────────────────────────────────────────────────────┐
│                 ChatGPT Health Interface                 │
│   • Powered by GPT-5.5 Instant & GPT-5.6 Sol             │
│   • Encrypted context engine; zero foundation training   │
└──────────────────────────────────────────────────────────┘

The 70% Insight: Why Health Integration Moved to Main Chats

Initial pilot testing revealed that over 70% of health discussions occurred naturally inside routine conversations rather than inside a standalone health hub.

During early testing, users routinely brought up health considerations while performing everyday tasks—such as asking for recipe ideas while managing a dietary restriction, or designing a family weekend itinerary around a recent joint injury.

To eliminate unnecessary friction, OpenAI updated the platform so users can grant ChatGPT permission to draw from connected health context anywhere across standard chats.

User Query: "Plan a 3-day meal prep under 2,000 calories."
               │
               ▼
┌──────────────────────────────────────────────────────────┐
│              ChatGPT Context Evaluator                   │
│   • Detects active permissions                              │
│   • References b.well data feed:                         │
│     - Elevated LDL cholesterol (Recent Epic Lab)         │
│     - Lactose Intolerance (Self-Reported/Apple Health)   │
└────────────────────────────┬─────────────────────────────┘
                             │
                             ▼
Optimized Output: High-fiber, dairy-free meal plan designed to manage cholesterol.

Supported Integrations: Clinical Records and Fitness Apps

The feature bridges both clinical medical systems and consumer lifestyle platforms through standardized application programming interfaces.

  • Clinical EHR Systems: Epic, Oracle Health, One Medical, and Kaiser Permanente portals.

  • Consumer Apps & Wearables: Apple Health (iOS), MyFitnessPal, and Function Health.

  • Lab & Diagnostic Services: Integrated lab history across over 2.2 million U.S. healthcare providers via network data pipelines.

AI Engine & Clinical Benchmarking: GPT-5.5 Instant vs. GPT-5.6 Sol

ChatGPT Health is powered by OpenAI’s latest architecture, offering distinct reasoning tiers based on user plan subscriptions.

Feature / Metric GPT-5.5 Instant GPT-5.6 Sol
Availability Tier Free and Go plans Plus and Pro plans
Primary Design Focus Rapid lab translation, terminology clarification, basic fitness context. Multi-year longitudinal chart review, complex medication interaction checks, nuanced clinical reasoning.
Evaluation Suite HealthBench Professional HealthBench Professional
Clinical Input Built with feedback from 260+ global physicians. Built with feedback from 260+ global physicians.

HealthBench Professional and the 260+ Global Physician Network

To validate model safety and clinical clarity, OpenAI collaborated with over 260 licensed physicians across 60 countries.

Rather than relying purely on multiple-choice medical board exams, models powering ChatGPT Health are benchmarked on HealthBench Professional. This evaluation framework assesses LLMs on real-world clinical communication, including:

  1. Recognizing when vital context is missing and proactively asking follow-up questions.

  2. Clearly communicating diagnostic uncertainty without sounding overly definitive.

  3. Identifying red-flag symptoms that require immediate emergency medical care.

Diagnostic Boundaries and Risk Mitigation

ChatGPT Health is explicitly designed to support—never replace—professional medical care.

Important Boundary: The tool does not provide formal medical diagnoses or prescribe treatment plans. OpenAI terms state that all outputs are for informational and organizational use to help patients have more productive, informed conversations with their doctors.

Security & Data Privacy: How Your Health Data Is Protected

Connected medical records and health conversations in ChatGPT are strictly compartmentalized and excluded from foundation model training.

┌──────────────────────────────────────────────────────────┐
│                    Your Personal Data                    │
│      (Apple Health, Epic Records, Lab Analyses)          │
└────────────────────────────┬─────────────────────────────┘
                             │
                             ▼
┌──────────────────────────────────────────────────────────┐
│             Encryption & Isolation Boundary              │
│   • Encrypted in transit and at rest                      │
│   • Dedicated health memory isolate                      │
│   • Explicit OAuth user consent controls                 │
└────────────────────────────┬─────────────────────────────┘
                             │
                             ▼
              ┌──────────────┴──────────────┐
              │                             │
              ▼                             ▼
┌───────────────────────────┐ ┌────────────────────────────┐
│      ChatGPT Response     │ │   OpenAI Model Training    │
│    (Delivered to User)    │ │   (BLOCKED / NEVER USED)   │
└───────────────────────────┘ └────────────────────────────┘

b.well FHIR Interoperability Architecture

The data pipeline uses b.well Connected Health’s FHIR-based network to retrieve and standardize clinical data securely.

Through b.well’s 13-step Data Refinery, unstructured doctor notes, medication schedules, and lab panels are converted into an AI-ready context format. This framework ensures end-to-end auditability and identity verification across thousands of health systems.

Privacy Guarantees at a Glance

  • Zero Model Training: Your connected EHR records, fitness data, and health-specific prompts are never used to train OpenAI models.

  • Zero Ad Targeting: Health context is entirely isolated from commercial ad-targeting profiles.

  • Right to Revoke: Users can disconnect any provider or app at any time; revoked data is purged from OpenAI’s systems within 30 days.

Step-by-Step: How to Enable ChatGPT Health

Activating ChatGPT Health requires linking your verified patient accounts through the ChatGPT sidebar on Web or iOS.

1.Open the Health Home Base:Web & iOS.

Log into your ChatGPT account. Locate and select Health from the left navigation menu.

2.Authenticate Data Sources:b.well Integration.

Select Connect Data. Choose your health provider (e.g., Epic, Oracle Health, One Medical) or fitness app (Apple Health, MyFitnessPal). Complete the secure OAuth login via b.well.

3.Configure Context Permissions:Custom Privacy.

Choose whether ChatGPT can reference your connected health data across standard daily chats, or only when explicitly tagged using @Health.

4.Review & Revoke Data:Ongoing Control.

Manage connected feeds or purge saved health memories directly from the Health Settings dashboard at any time.

 

Frequently Asked Questions

Is ChatGPT Health available on Android or outside the U.S.?

ChatGPT Health is currently available for U.S. adult users on Web and iOS. OpenAI plans to expand access to additional platforms and international regions over time.

Is my medical record data HIPAA compliant in ChatGPT?

Because OpenAI is a software provider interacting directly with consumer-authorized data, traditional HIPAA rules between covered entities do not directly govern the AI interface. However, the data transmission layer relies on secure FHIR APIs, purpose-built encryption, and strict contractual guarantees that your health data will never be used for AI training or ad targeting.

Can ChatGPT diagnose my health condition?

No. ChatGPT Health is designed to help users organize, summarize, and understand their health information. It cannot diagnose medical conditions or prescribe treatments. Always consult a licensed healthcare professional for medical advice.

How do I trigger Health context in an ongoing conversation?

If you have connected your records, you can type @Health within any standard prompt to pull in relevant lab results, activity metrics, or medical histories on demand.