More than 300 million people now turn to ChatGPT with health questions every week — and OpenAI just gave them significantly more to work with. The company is rolling out Health in ChatGPT to all U.S. users aged 18 and older, a full relaunch of a feature first piloted in January that flopped quietly before being rebuilt from scratch. The bigger story, though, isn’t just the relaunch. It’s what this move reveals about AI’s expanding role in personal healthcare — and the uncomfortable trade-offs that come with it.
Summary
Key takeaways
- OpenAI Health in ChatGPT is now available to all U.S. users aged 18 and older on free, Go, Plus, and Pro plans.
- Paid subscribers access GPT-5.6 Sol, which scores 88.0% on completeness in HealthBench Professional versus 53.2% for the free tier’s GPT-5.5 Instant.
- Both AI models outperform physician-written answers on OpenAI’s HealthBench Professional benchmark — though benchmarks don’t fully capture real clinical practice.
- The feature is not available in Europe due to stricter data privacy rules and potential high-risk classification under the EU AI Act.
- OpenAI says connected health data will not be used for model training or advertising.
OpenAI Launches Health in ChatGPT for US Users
The relaunch is broader than the original January pilot. Health in ChatGPT is now open to logged-in users across all plan tiers — Free, Go, Plus, and Pro — on web and iOS in the United States. Early testing revealed a significant friction point: more than 70 percent of participants were asking health questions outside the dedicated Health section because switching to it felt too cumbersome. OpenAI responded by making the Health feature accessible within any conversation, while keeping a separate Health section for managing connected data and reviewing past health discussions.
What You Can Actually Connect
Users can link Apple Health, supported medical records apps, and other wellness platforms directly to ChatGPT. Once connected, the system can help compare a new lab result against prior tests, summarize changes since a last appointment, or analyze patterns across sleep, activity, and workout data. OpenAI frames this as a way to have “more informed, personalized conversations” about health — a pitch that positions ChatGPT somewhere between a knowledgeable friend and a pre-appointment research tool.
OpenAI says it will not use any of that connected health data for model training or advertising purposes. That privacy assurance matters, given the sensitivity of what users might share.
Tiered Quality: Not All Health Advice Is Equal
Here is where things get complicated. Free users receive health guidance powered by GPT-5.5 Instant, while paying subscribers get access to GPT-5.6 Sol, OpenAI’s new flagship model reserved for paid tiers. The performance gap between them is not trivial.
On OpenAI’s own HealthBench Professional benchmark, GPT-5.6 Sol scores 88.0% on completeness, compared to 53.2% for GPT-5.5 Instant. On health decision helpfulness, the gap is similarly wide: 83.0% versus 50.8%. Both models, OpenAI notes, beat physician-written answers across every category on HealthBench Professional — a claim it will likely lean on to justify the two-tier system.
Why Benchmark Wins Don’t Tell the Whole Story
That benchmark claim deserves scrutiny. Doctors who participate in such tests often do so without access to patient records, under time pressure, and without the collaborative input they would normally rely on. An AI model trained specifically on medical knowledge, operating without fatigue or distraction, will naturally perform well in controlled knowledge tests. What those tests cannot measure is the physical examination, the ability to notice a patient’s pallor, the nonverbal hesitation that changes a diagnosis — or the judgment call that comes from years of clinical experience.
OpenAI itself is careful to acknowledge this. The company, along with the more than 260 physicians who contributed to developing the Health features, consistently states that ChatGPT can still make mistakes and cannot replace professional medical advice. That disclaimer matters more than it might appear: in radiology, for example, the latest RadLE 2.0 benchmark found that none of the 16 AI models tested matched human radiologists — primarily because chatbots delivered incorrect findings with high confidence rather than acknowledging the limits of their knowledge. That overconfidence pattern is a known problem across AI health systems, and it has contributed to real-world harms, particularly in mental health contexts where chatbot validation of false assumptions can cause serious damage.
Ethical and Regulatory Fault Lines
The decision to offer meaningfully worse health advice to free users raises a legitimate equity concern. Health information is not a luxury feature. When the quality of AI-generated guidance on a serious symptom diverges sharply depending on whether someone pays a monthly subscription, that gap has real-world consequences — particularly for users who rely on free tools precisely because they lack access to other healthcare resources.
Europe Left Out, and for Good Reason
OpenAI has not made Health in ChatGPT available in the European Economic Area, Switzerland, or the United Kingdom. The company has not committed to a timeline for European availability either. The most likely explanations are the EU’s stricter data privacy regulations and the possibility that a feature of this nature could be classified as high-risk under the EU AI Act — a designation that would trigger significantly more demanding compliance requirements before deployment. Rather than navigate that regulatory complexity now, OpenAI appears to have chosen to launch elsewhere first.
That regulatory gap is itself revealing. The fact that a product clears the bar for U.S. deployment but remains legally ambiguous in Europe speaks to how differently the two jurisdictions assess AI risk in sensitive domains like healthcare. Europe’s caution may frustrate users there, but it reflects a broader disagreement about how much scrutiny AI health tools should face before reaching consumers at scale.
The Scale of Demand Is Already Enormous
OpenAI’s usage figures make clear why the company is investing so heavily here. Weekly health queries on ChatGPT have risen from 230 million in January to over 300 million today. That trajectory suggests people are already using ChatGPT as a first-stop health resource, regardless of whether a dedicated Health feature exists. The relaunch formalizes something that was already happening informally — and tries to make it safer, more personalized, and more structured in the process.
Research into AI in clinical settings offers a mixed but not dismissive picture. Systems like AMIE have performed comparably to primary care physicians in simulated consultations. One researcher summarized the dynamic with a useful analogy: AI agents in healthcare are like an airplane’s autopilot — capable of handling routine tasks reliably, but with ultimate responsibility always remaining with the physician in the cockpit.
That framing probably captures where things stand better than either the optimistic benchmark headlines or the more alarmed critiques. ChatGPT Health, at its best, is a more capable and personalized version of searching for health information online. At its worst — particularly with free-tier users receiving less complete answers from a model prone to overconfidence — it is a tool that can mislead people who trust it more than they should. The 300 million weekly users already relying on it for health guidance means that gap between best and worst case carries significant weight.
FAQ
Who can access the Health in ChatGPT feature?
The Health feature is available to U.S. users aged 18 and older on all ChatGPT plans, including Free, Go, Plus, and Pro.
What integrations does the Health feature support?
Users can connect Apple Health, supported medical records apps, and wellness platforms to ChatGPT’s Health feature, enabling personalized insights based on lab results, sleep data, activity, and appointment history.
What are the differences between free and paid versions of ChatGPT health advice?
Free users receive health guidance powered by GPT-5.5 Instant, which scores 53.2% on completeness and 50.8% on health decision helpfulness in HealthBench Professional. Paying subscribers access GPT-5.6 Sol, which scores 88.0% on completeness and 83.0% on health decision helpfulness on the same benchmark.
Is the Health feature available in Europe?
No. OpenAI has not made the Health feature available in the European Economic Area, Switzerland, or the United Kingdom, citing stricter data privacy regulations and the possibility of high-risk classification under the EU AI Act.
Can ChatGPT replace professional medical advice?
No. OpenAI explicitly states that ChatGPT can still make mistakes and cannot replace a doctor’s medical advice, despite the improvements introduced in the Health feature relaunch.
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.
