Greater than 300 million folks now flip to ChatGPT with well being questions each week — and OpenAI simply gave them considerably extra to work with. The corporate is rolling out Well being in ChatGPT to all U.S. customers aged 18 and older, a full relaunch of a function first piloted in January that flopped quietly earlier than being rebuilt from scratch. The larger story, although, isn’t simply the relaunch. It’s what this transfer reveals about AI’s increasing function in private healthcare — and the uncomfortable trade-offs that include it.
Key takeaways
- OpenAI Well being in ChatGPT is now out there to all U.S. customers aged 18 and older on free, Go, Plus, and Professional plans.
- Paid subscribers entry GPT-5.6 Sol, which scores 88.0% on completeness in HealthBench Skilled versus 53.2% for the free tier’s GPT-5.5 Immediate.
- Each AI fashions outperform physician-written solutions on OpenAI’s HealthBench Skilled benchmark — although benchmarks don’t absolutely seize actual medical observe.
- The function is just not out there in Europe resulting from stricter information privateness guidelines and potential high-risk classification below the EU AI Act.
- OpenAI says related well being information is not going to be used for mannequin coaching or promoting.
OpenAI Launches Well being in ChatGPT for US Customers
The relaunch is broader than the unique January pilot. Well being in ChatGPT is now open to logged-in customers throughout all plan tiers — Free, Go, Plus, and Professional — on net and iOS in the US. Early testing revealed a major friction level: greater than 70 p.c of contributors have been asking well being questions outdoors the devoted Well being part as a result of switching to it felt too cumbersome. OpenAI responded by making the Well being function accessible inside any dialog, whereas retaining a separate Well being part for managing related information and reviewing previous well being discussions.
What You Can Truly Join
Customers can hyperlink Apple Well being, supported medical information apps, and different wellness platforms on to ChatGPT. As soon as related, the system may help examine a brand new lab end result in opposition to prior checks, summarize modifications since a final appointment, or analyze patterns throughout sleep, exercise, and exercise information. OpenAI frames this as a approach to have “extra knowledgeable, personalised conversations” about well being — a pitch that positions ChatGPT someplace between a educated buddy and a pre-appointment analysis device.
OpenAI says it is not going to use any of that related well being information for mannequin coaching or promoting functions. That privateness assurance issues, given the sensitivity of what customers may share.
Tiered High quality: Not All Well being Recommendation Is Equal
Right here is the place issues get difficult. Free customers obtain well being steering powered by GPT-5.5 Immediate, whereas paying subscribers get entry to GPT-5.6 Sol, OpenAI’s new flagship mannequin reserved for paid tiers. The efficiency hole between them is just not trivial.
On OpenAI’s personal HealthBench Skilled benchmark, GPT-5.6 Sol scores 88.0% on completeness, in comparison with 53.2% for GPT-5.5 Immediate. On well being resolution helpfulness, the hole is equally vast: 83.0% versus 50.8%. Each fashions, OpenAI notes, beat physician-written solutions throughout each class on HealthBench Skilled — a declare it is going to probably lean on to justify the two-tier system.
Why Benchmark Wins Don’t Inform the Complete Story
That benchmark declare deserves scrutiny. Docs who take part in such checks typically accomplish that with out entry to affected person information, below time stress, and with out the collaborative enter they’d usually depend on. An AI mannequin educated particularly on medical information, working with out fatigue or distraction, will naturally carry out properly in managed information checks. What these checks can’t measure is the bodily examination, the flexibility to note a affected person’s pallor, the nonverbal hesitation that modifications a prognosis — or the judgment name that comes from years of medical expertise.
OpenAI itself is cautious to acknowledge this. The corporate, together with the greater than 260 physicians who contributed to growing the Well being options, constantly states that ChatGPT can nonetheless make errors and can’t change skilled medical recommendation. That disclaimer issues greater than it’d seem: in radiology, for instance, the newest RadLE 2.0 benchmark discovered that not one of the 16 AI fashions examined matched human radiologists — primarily as a result of chatbots delivered incorrect findings with excessive confidence fairly than acknowledging the boundaries of their information. That overconfidence sample is a identified downside throughout AI well being techniques, and it has contributed to real-world harms, notably in psychological well being contexts the place chatbot validation of false assumptions could cause severe injury.
Moral and Regulatory Fault Traces
The choice to supply meaningfully worse well being recommendation to free customers raises a official fairness concern. Well being info is just not a luxurious function. When the standard of AI-generated steering on a severe symptom diverges sharply relying on whether or not somebody pays a month-to-month subscription, that hole has real-world penalties — notably for customers who depend on free instruments exactly as a result of they lack entry to different healthcare sources.
Europe Left Out, and for Good Purpose
OpenAI has not made Well being in ChatGPT out there within the European Financial Space, Switzerland, or the UK. The corporate has not dedicated to a timeline for European availability both. The probably explanations are the EU’s stricter information privateness laws and the likelihood {that a} function of this nature may very well be categorised as high-risk below the EU AI Act — a designation that might set off considerably extra demanding compliance necessities earlier than deployment. Quite than navigate that regulatory complexity now, OpenAI seems to have chosen to launch elsewhere first.
That regulatory hole is itself revealing. The truth that a product clears the bar for U.S. deployment however stays legally ambiguous in Europe speaks to how otherwise the 2 jurisdictions assess AI danger in delicate domains like healthcare. Europe’s warning could frustrate customers there, nevertheless it displays a broader disagreement about how a lot scrutiny AI well being instruments ought to face earlier than reaching shoppers at scale.
The Scale of Demand Is Already Huge
OpenAI’s utilization figures clarify why the corporate is investing so closely right here. Weekly well being queries on ChatGPT have risen from 230 million in January to over 300 million right now. That trajectory suggests individuals are already utilizing ChatGPT as a first-stop well being useful resource, no matter whether or not a devoted Well being function exists. The relaunch formalizes one thing that was already taking place informally — and tries to make it safer, extra personalised, and extra structured within the course of.
Analysis into AI in medical settings presents a combined however not dismissive image. Methods like AMIE have carried out comparably to major care physicians in simulated consultations. One researcher summarized the dynamic with a helpful analogy: AI brokers in healthcare are like an airplane’s autopilot — able to dealing with routine duties reliably, however with final duty all the time remaining with the doctor within the cockpit.
That framing most likely captures the place issues stand higher than both the optimistic benchmark headlines or the extra alarmed critiques. ChatGPT Well being, at its finest, is a extra succesful and personalised model of looking for well being info on-line. At its worst — notably with free-tier customers receiving much less full solutions from a mannequin liable to overconfidence — it’s a device that may mislead individuals who belief it greater than they need to. The 300 million weekly customers already counting on it for well being steering implies that hole between finest and worst case carries important weight.
FAQ
Who can entry the Well being in ChatGPT function?
The Well being function is offered to U.S. customers aged 18 and older on all ChatGPT plans, together with Free, Go, Plus, and Professional.
What integrations does the Well being function help?
Customers can join Apple Well being, supported medical information apps, and wellness platforms to ChatGPT’s Well being function, enabling personalised insights based mostly on lab outcomes, sleep information, exercise, and appointment historical past.
What are the variations between free and paid variations of ChatGPT well being recommendation?
Free customers obtain well being steering powered by GPT-5.5 Immediate, which scores 53.2% on completeness and 50.8% on well being resolution helpfulness in HealthBench Skilled. Paying subscribers entry GPT-5.6 Sol, which scores 88.0% on completeness and 83.0% on well being resolution helpfulness on the identical benchmark.
Is the Well being function out there in Europe?
No. OpenAI has not made the Well being function out there within the European Financial Space, Switzerland, or the UK, citing stricter information privateness laws and the potential of high-risk classification below the EU AI Act.
Can ChatGPT change skilled medical recommendation?
No. OpenAI explicitly states that ChatGPT can nonetheless make errors and can’t change a health care provider’s medical recommendation, regardless of the enhancements launched within the Well being function relaunch.
Article produced with the help of synthetic intelligence and reviewed by the editorial crew.