What happened
OpenAI has begun rolling out "Health in ChatGPT," a new feature that lets U.S. users aged 18 and older connect Apple Health, medical records, and wellness apps directly to the chatbot. After months of testing since January 2026, the feature now allows people to review lab results, prepare questions for doctor's appointments, and get help interpreting sleep or activity data. But there's a catch that's raising eyebrows: the quality of ChatGPT health advice now depends on your subscription tier. Free users are served by GPT-5.5 Instant, a lighter model, while paying subscribers get GPT-5.6 Sol, OpenAI's stronger flagship model built specifically to score higher on health-related benchmarks. OpenAI says connected health data won't be used for model training or advertising, and more than 260 physicians reportedly helped shape the feature.
Why it matters
The gap between the two models isn't small. On OpenAI's own HealthBench Professional test, GPT-5.6 Sol scored 88.0 percent on completeness versus 53.2 percent for the free-tier model and physician-written answers — a 35-point difference. On "health decision helpfulness," the paid model hit 83.0 percent compared to 50.8 percent for GPT-5.5 Instant and human doctors. OpenAI will likely argue this is an ethical improvement, since both models reportedly outperform physicians on the benchmark. But benchmark scores are measured in artificial test conditions, not real exam rooms. Doctors taking the same test may be rushed, fatigued, or working without access to patient records or colleague input — none of which reflects an actual clinical visit, where physicians can examine patients directly, read nonverbal cues, and draw on years of hands-on experience. More than 300 million people now ask ChatGPT health questions every week, up from 230 million in January, which means millions of free users could be getting meaningfully weaker answers than paying subscribers on the exact same questions.
How to use it today
If you're in the U.S. and 18 or older, Health in ChatGPT is available now inside any regular conversation — OpenAI found that more than 70 percent of early testers kept asking health questions outside the dedicated Health section because switching tabs felt cumbersome, so the company folded the feature into normal chat while keeping a separate Health tab for managing connected data and reviewing past health conversations. To use it, connect Apple Health, upload medical records, or link a wellness app, then ask ChatGPT to summarize lab results or draft questions ahead of an appointment. Treat the output as a starting point for a conversation with your doctor, not a diagnosis. If you're experimenting with AI tools for personal productivity beyond health tracking, it's worth exploring lightweight, free options like the ones at [mykreatool.com](https://mykreatool.com), which can help with everyday tasks without requiring a paid plan. The feature is currently U.S.-only; OpenAI hasn't said if or when it will reach Europe, the UK, or Switzerland.
Who benefits
Paying ChatGPT Plus or Pro subscribers get the clearest upside: sharper, more complete answers powered by GPT-5.6 Sol, plus the convenience of connecting Apple Health and medical records in one place. People managing chronic conditions or juggling multiple specialists may find real value in having lab results summarized and appointment questions pre-drafted. Caregivers helping older relatives track medications or symptoms could also benefit from a centralized health chat history. Businesses building on top of AI health tools gain a preview of where consumer-facing medical AI is headed — tiered model access tied to subscription level, a pattern likely to spread beyond OpenAI as competitors add similar features.
Risks
The two-tier system means free users, who may be the least able to afford a doctor's visit in the first place, get the weaker model. That's a real equity concern. More broadly, OpenAI itself repeatedly states in its announcement that ChatGPT can make mistakes and isn't a substitute for medical advice. That warning is backed by hard data: in the RadLE 2.0 radiology benchmark, none of 16 AI models tested matched human radiologists' performance. The core failure wasn't lack of knowledge — it was that chatbots gave incorrect findings with high confidence instead of admitting uncertainty, while human radiologists were more likely to flag when they'd reached the limits of what they could determine. That distinction matters enormously in medicine, where a confidently wrong answer can be more dangerous than an honest "I'm not sure." Regulatory risk is also part of the picture — the feature's absence from the EU, UK, and Switzerland likely reflects concerns that it could be classified as high-risk under the EU AI Act, alongside stricter data privacy rules in those regions.
Conclusion
Health in ChatGPT is a genuinely useful tool for organizing health data and prepping for doctor visits, but it's not a neutral upgrade — it's a tiered product where paying customers get demonstrably better answers to health questions than free users do. The benchmark numbers are impressive on paper, but real medicine involves context, judgment, and the ability to say "I don't know" that AI models still struggle with, as the RadLE 2.0 radiology results make clear. Use ChatGPT's health features as a helpful research assistant and appointment-prep tool, not a replacement for an actual doctor.
Comments 0