I unexpectedly spent a lot of time sick in hospitals and doctors’ offices this summer, so, as a reporter covering AI, I was well-prepared to stress-test the new version of ChatGPT Health.

ChatGPT Health is a specialized experience inside the chatbot, designed with solid privacy restrictions and safety guardrails to help answer a variety of health questions. It’s rolling out today in the US for free and paid personal ChatGPT accounts for folks over 18 on the web and iOS. It was first announced in January in beta.

I had my appendix removed earlier this summer, and a bevy of postoperative complications kept me feeling pretty miserable for several weeks. I’m feeling much better nowadays, thankfully, and I knew the bucketload of digital paperwork I’d racked up would be useful during my early access to the program.

As with any AI program, the more information you’re willing to give it, the better the results turn out. You can connect your Apple Health account to ChatGPT, as well as your electronic health record for many different providers. The idea is to have all your health info, from your doctor notes to data to info from your Apple Watch and Oura ring, all compiled inside ChatGPT. This could be helpful for people who have a lot of health records, like folks with chronic illnesses, or those who have their records spread out across hospitals and doctors’ offices.

OpenAI says it does not use your connected health information to train its foundation models or advertisements. You can request that your data be deleted, set a limited time window for which ChatGPT Health has access to your outside accounts and disconnect each account as needed. (Disclosure: Ziff Davis, CNET’s parent company, in 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)

Even so, there will certainly be instances when you don’t want to share your most sensitive data. Tech companies have a history of cooperating with law enforcement, and as some governments ramp up highly politicized healthcare policies criminalizing some procedures, like abortion and gender-affirming care, you may want to keep your most sensitive information offline. ChatGPT Health is not HIPAA-compliant, meaning you don’t have the kind of confidentiality you have with your doctor.

Setting up ChatGPT Health

I connected ChatGPT with one of my patient portals, selecting the provider from a list and signing into my account, going through the usual two-factor authentication process. I was then asked to select my current health conditions and medications from the options it gleaned from my electronic health record, as well as broadly indicate any family medical history.

ChatGPT pulled a lot of repetitive data – generic and prescription names for the same drug, for example – so I had to double-check its results. The whole process took less than five minutes.

Over 300 million people ask ChatGPT health-related questions every week, OpenAI said on Thursday, a huge leap forward from January’s reported 40 million weekly queries. But unsurprisingly, most of those queries (70%) happen outside the dedicated Health tab. Now, once you set up ChatGPT Health, it will ask you if you permit the chatbot to use your health information to answer your question, no matter where you ask it.

I asked ChatGPT a series of health questions, and each time I was pleasantly surprised by the results. Its answers were clearly grounded in my medical history, referencing lab results regularly. I used it to brainstorm a decent weekly meal plan and fitness regimen, rooted in my specific recovery phase. I was irked that it wanted me to connect more plugins before it could help me with an insurance question, but why should anything about insurance be straightforward?

Maybe most reassuringly, everything ChatGPT told me was something I had already been told by my doctors. This is on purpose – ChatGPT isn’t supposed to diagnose issues or create treatment plans. It can explain those conditions very well, thanks to the more than 260 physicians working with OpenAI. They’ve reviewed over 700,000 AI responses over the past two years, across 49 languages and 26 medical specialties.

ChatGPT Health uses the newest models from the company, including the GPT-5.6 series. The newer models, 5.5 and 5.6, are better at recognizing when urgent medical intervention is needed, seeking additional information from you when necessary, and avoiding overreach in their answers. To put it another way, the AI is better at knowing when its answers aren’t good enough and when to push you to seek out doctors for their human medical expertise.

Managing health anxiety with tech

Tech in healthcare is nothing new, and ChatGPT Health follows a growing trend of AI tech being used in the field, from insurance to charting to laboratory and imagery software. Microsoft introduced a similar tool called Copilot Health earlier this year. Many of my doctors recorded our conversations with AI so the tech could write up our visit summary (with and without my permission, I later learned).

But I would be remiss if I didn’t mention that while I was going through surgery and weeks of recovery, I only used an AI chatbot twice: Once to understand a particularly scary lab report that came in at 11 p.m. and once more to prep for a doctor appointment to discuss the scary lab report. I didn’t use ChatGPT Health, but those experiences only turbocharged my anxiety. I worried that the chatbot was hallucinating – that is, making stuff up – or pulling from dodgy sources online like Reddit.

AI did what it did best and tried to reassure me, but it was no replacement for the real human support I needed in those moments. Unsurprisingly, it was the real humans who made my experience bearable – my nurses, doctors, friends and family.

OpenAI, to its credit, isn’t claiming its AI is a substitute for any of these things. And its specialized health experience, when connected with your data, is a big step up from general chatbots. The improvements have me wondering if I would’ve had a less anxiety-ridden experience if ChatGPT Health had been available at that time. But as with a lot of telehealth innovation, I fear ChatGPT Health is just another Band-Aid we are putting on the gaping wound that is the US healthcare system.