Last year, when Fidji Simo was hospitalised for a kidney stone infection, a routine antibiotic prescription nearly became a medical crisis. The resident doctor had ordered a standard medication, but Simo, who manages a chronic illness, took a moment to check. She asked ChatGPT to review the prescription against her medical history already uploaded to the platform. The AI flagged a critical problem: this particular antibiotic could reactivate a serious prior infection.
"When I brought this up with the resident, I expected pushback, but instead, she was relieved," recalls Simo, now CEO of OpenAI Applications. "She told me this medication could have created very serious complications for me, and she was glad I caught it."
The incident captures both the promise and the problem with modern healthcare. When Simo asked why the hospital hadn't caught this potential complication, the doctor's answer was stark: she had only five minutes per patient during rounds, and health records weren't organised to make such risks clear.
Now, OpenAI is betting that artificial intelligence can help bridge these gaps. The company has launched ChatGPT Health, a dedicated space within ChatGPT designed to securely integrate medical records and wellness data, helping people navigate an increasingly fragmented healthcare system.
The scale of the problem
The numbers paint a concerning picture. According to Simo, 62% of Americans say the healthcare system is broken, whilst nearly half of physicians reported at least one symptom of burnout last year. Yet health queries have become one of the most common uses of ChatGPT, with over 230 million people globally asking health and wellness questions each week.
"The healthcare system clearly isn't working for patients or doctors, which is why both are turning to AI tools for help," Simo explained in a statement.
The shift is already underway. The American Medical Association reports that physicians' use of AI nearly doubled from 2023 to 2024, with 68% saying AI gives them an advantage in caring for patients. Meanwhile, three in five American adults have used AI tools for health or healthcare in the past three months, with 75% of those finding it very or extremely helpful.
What ChatGPT Health actually does
ChatGPT Health operates as a separate, privacy-protected space within ChatGPT where users can securely connect medical records and wellness applications including Apple Health, Function, MyFitnessPal, Weight Watchers, Peloton, and AllTrails.
The platform is designed to help with everyday health tasks rather than diagnosis or treatment. Users can ask questions like "How's my cholesterol trending?" or "Can you summarise my latest bloodwork before my appointment?" The AI pulls from connected health data to provide personalised responses.
Practical applications include understanding recent test results, preparing questions for doctor's appointments, getting advice on diet and workout routines, interpreting data from wearables, and even understanding insurance options based on healthcare patterns.
"Health is designed to support, not replace, medical care," OpenAI emphasises. "It is not intended for diagnosis or treatment. Instead, it helps you navigate everyday questions and understand patterns over time."
How to use it
Getting started with ChatGPT Health involves three main steps. First, users must request access through a waitlist, as the service is rolling out gradually to Free, Go, Plus, and Pro plan subscribers outside the European Economic Area, Switzerland, and the United Kingdom.
Once granted access, users select 'Health' from the ChatGPT sidebar menu. They can then connect their information through several methods: uploading files directly, connecting medical records through b.well (OpenAI's partner for accessing U.S. healthcare providers), or linking wellness apps through the Settings menu.
The platform supports various data sources. Medical records can include lab results, visit summaries, and clinical history. Apple Health provides fitness data including movement, sleep, and activity patterns, though iOS is required for syncing. Other integrations like Function offer lab test insights, MyFitnessPal provides nutrition advice, and Peloton suggests workout classes.
Once information is connected, conversations work just like regular ChatGPT chats, but grounded in personal health data. Users can upload photos and files, use voice mode, and even set custom instructions to focus on specific health concerns or avoid sensitive topics.
Privacy protections
Given the sensitive nature of health information, ChatGPT Health operates with layered security measures. The platform functions as a completely separate space where conversations, connected apps, and files are stored apart from regular ChatGPT chats. Health information never flows back into non-health conversations.
All data is encrypted at rest and in transit, with purpose-built encryption and isolation specifically for health conversations. Critically, conversations in Health are not used to train OpenAI's foundation models, addressing a common concern about AI platforms.
Users maintain control throughout. Apps require explicit permission to connect, even if already linked to ChatGPT for other purposes. Medical records can be disconnected at any time through Settings, immediately revoking access. The platform also supports multi-factor authentication for additional account protection.
OpenAI partners with b.well for medical record access, describing it as the largest and most secure network of live, connected health data for American consumers, adhering to the highest industry standards.
Built with medical expertise
Over two years, OpenAI collaborated with more than 260 physicians who have practised in 60 countries across dozens of specialties. This group provided feedback on model outputs over 600,000 times across 30 areas of focus, shaping not just what Health can do, but how it responds.
The collaboration addressed crucial questions: how urgently to encourage follow-ups with clinicians, how to communicate clearly without oversimplifying, and how to prioritise safety in critical moments.
This physician-led approach is built into the underlying model, evaluated against clinical standards using HealthBench, an assessment framework created with practising physicians. Rather than exam-style questions, HealthBench uses physician-written rubrics reflecting how clinicians judge quality in practice, prioritising safety, clarity, appropriate escalation of care, and respect for individual context.
Addressing systemic healthcare problems
Simo outlines four structural problems in healthcare that AI might help address. First, doctors lack bandwidth, with heavier caseloads and administrative burden leaving less time for patients and staying current with research. AI can synthesise large amounts of information, helping consider full medical histories, overlapping conditions, medications, and risk factors together.
Second, healthcare is fragmented whilst health requires a complete picture. Simo's chronic condition required seeing over 20 specialists, each examining a handful of symptoms from dozens, despite her disease being systemic. Only 16% of physicians fully exchanged and integrated electronic patient information as of 2021.
Third, cost and access create barriers, particularly in rural areas where 10 hospitals per year have closed or shifted away from inpatient care since 2010. Seven in 10 American health conversations in ChatGPT happen outside normal clinic hours, suggesting AI fills gaps when seeing a doctor isn't feasible.
Finally, healthcare is reactive rather than preventative. According to the CDC, five of the top 10 causes of death in America are associated with preventable and treatable chronic diseases, yet the system focuses on illness rather than daily health management.
Who can use it?
ChatGPT Health is available to users with Free, Go, Plus, and Pro plans, but with significant geographical restrictions. The service is currently rolling out to users outside the European Economic Area, Switzerland, and the United Kingdom through a waitlist system.
This means Indian users are eligible to request access and join the waitlist. However, there's an important caveat: whilst Indians can access the core ChatGPT Health features, medical record integrations and some wellness apps are available only in the United States. Apple Health connectivity also requires iOS devices.
For Indian users, this means the initial experience will be more limited compared to American counterparts. They can upload health documents manually, use supported wellness apps like MyFitnessPal and Peloton, and have health conversations with the AI. However, the seamless integration with electronic medical records through b.well—a key feature—remains unavailable outside America.
OpenAI hasn't specified a timeline for when full medical record integration will expand to India or other countries. The company states it's "starting by providing access to a small group of early users to learn and continue refining the experience" before broader rollout in the coming weeks, but this appears to refer primarily to expanding access within currently eligible regions rather than geographical expansion.
The phased approach suggests OpenAI is prioritising the American market first, likely due to partnerships with U.S. healthcare providers and the complexities of navigating different countries' healthcare systems and data protection regulations.
The bigger picture
ChatGPT Health represents OpenAI's attempt to transform ChatGPT into what Simo calls a "personal super-assistant" supporting users across all aspects of life. The platform currently handles over 40 million daily health questions globally, indicating substantial demand for AI-assisted health navigation.
The geographical limitations highlight a broader challenge in healthcare AI: what works in one country's healthcare system may not translate easily to another. India's healthcare landscape, with its mix of public and private providers, different electronic health record systems, and distinct regulatory environment, would require separate partnerships and infrastructure.
Whether AI can truly address healthcare's systemic challenges remains to be seen, but for now, hundreds of millions are already asking the question—even if the full answer isn't yet available everywhere.