With health consultation becoming the most frequently used scenario for ChatGPT (over 230 million times per week globally), OpenAI has officially launched "ChatGPT Health" – an independent AI service focused on personal health management. It can not only integrate data from multiple sources such as electronic medical records, Apple Health, and MyFitnessPal, but also interpret health check-up reports, generate lists of questions for medical visits, create personalized diet and exercise plans, and even place orders for healthy ingredients via Instacart with one click. This move marks a critical step for OpenAI in transitioning from a "general AI assistant" to a "vertical domain intelligent agent."
Data Integration + Doctor Collaboration: Building a Trustworthy Health AI
The core breakthrough of ChatGPT Health lies in breaking down health data silos:
- Accessing electronic medical records (EMR) through the largest healthcare data platform in the U.S., b.well;
- Supporting authorized connections with Apple Health, Peloton, MyFitnessPal, and GLP-1 medication records;
- Users can directly ask: "What is my recent cholesterol trend?" or "What should I ask my cardiologist tomorrow?"
To ensure professionalism and safety, OpenAI collaborated with over 260 practicing doctors from 60 countries, providing more than 600,000 feedbacks to train the AI:
- Identifying high-risk symptoms (such as early signs of a heart attack) and recommending emergency medical care;
- Avoiding overdiagnosis, distinguishing between "needs observation" and "needs emergency treatment";
- Explaining medical terminology in simple language.
All health-related interactions run in an isolated encrypted environment, completely separated from regular chats, and health data is never used for model training.
From "Consultation Assistant" to "Proactive Health Management"
ChatGPT Health's functions go far beyond traditional health Q&A:
- Medical preparation: Generate personalized question lists based on medical history;
- Insurance comparison: Analyze coverage of different insurance plans;
- Lifestyle interventions:
- Postpartum recovery? Recommend Peloton-compatible courses;
- Want to build muscle using GLP-1 medications? Generate high-protein meal plans and Instacart shopping lists;
- Care summaries: Condense complex medical instructions into clear action items.
OpenAI also introduced the HealthBench evaluation framework, scoring AI on safety, understandability, and referral recommendations according to medical standards.
Executive's Personal Experience: AI Can Really Save Lives
OpenAI's application CEO, Fidji Simo (who suffers from POTS and endometriosis), shared her personal experience:
Last year, she was hospitalized for kidney stones and was prescribed a standard antibiotic. However, ChatGPT, based on her uploaded medical history, pointed out that this drug could trigger a recurrence of her previous severe infection. The attending physician confirmed and immediately changed the medication, admitting, "Each round of ward rounds only lasts five minutes, and the medical records system would never reveal such a risk."
This case illustrates the unique value of AI in information integration and risk warning.
CEO's Dilemma: Trust AI, But Not "Doctorless" AI
Despite Sam Altman repeatedly stating that "AI diagnostic capabilities have surpassed most doctors," he still emphasized, "I don't want to leave the fate of medicine entirely in the hands of a ChatGPT without human involvement."
This reveals the essential boundary of AI in healthcare: AI can be a super assistant, but it cannot replace the trust, empathy, and ethical judgment between doctors and patients.
AIbase Observation: The Health Sector Enters the "AI Butler" Era
From Ant Group's "Afu" with over 15 million monthly active users, to OpenAI's heavy investment, AI in health is shifting from "registration and consultation" to "full-cycle proactive management."
The real challenge for ChatGPT Health is not technology, but rather:
- Whether it can build user trust between convenience and reliability;
- Whether it can cross regional and data barriers (currently EMR access is limited to the U.S.);
- Whether it can scale up under compliance conditions.
