OpenAI’s release notes for September 28, 2026 say you can select a chart, metric, or record in the ChatGPT Health tab and get a personalized explanation of it. That turns a screen of numbers from Apple Health or a hospital portal into something you can ask questions about in plain language. The catch is who gets it. OpenAI’s help center lists Health for web and iOS only, for adults in the US, and Android is not on that list.
TL;DR: ChatGPT’s Health tab can explain any chart, metric, or medical record you connect, according to OpenAI’s September 28, 2026 release notes. It works on the web and iOS for eligible US users aged 18 and older, and it can only read your data, not change it. OpenAI says connected health data is not used to train its foundation models or target ads. It is not for diagnosis and is not HIPAA eligible.
What the ChatGPT Health tab can explain
The Health tab is the place where you connect accounts and see recent data and trends, per the help center article on Health in ChatGPT. The new piece is a shortcut inside that view: pick a chart, a single metric, or a record, and ChatGPT explains what it shows.
Picture the moment this is built for. You open a lab result, see a value flagged with an arrow, and have no idea whether it deserves a phone call or a shrug. Asking about that one line, with your own history sitting beside it, is a different experience from pasting numbers into a blank chat.
ChatGPT Health first launched on January 7, 2026, and OpenAI’s launch post for ChatGPT Health describes a July 23, 2026 update that opened it to US adults on web and iOS. The release note itself is short and skips plans and regions, so the details below come from the help center article it links to.
| Detail | What the help center says |
|---|---|
| Platforms | Web and iOS |
| Plans | Free, Go, Plus, and Pro |
| Age | 18 or older |
| Region | US only, gradual rollout |
| iOS app version | 1.2026.188 or later |
| Android | Not listed |
Who can use it and who is left out
Health is open to eligible, logged in Free, Go, Plus, and Pro users who are 18 or older. It is limited to US users, and the rollout is gradual, so the tab may not show up for everyone yet.
Apple Health can only be connected from the iPhone app, which needs version 1.2026.188 or later. Apple’s own AI features also depend on which iPhone you own, as our iOS 27 Siri AI device list shows. Android has no equivalent listed, so Pixel and Galaxy owners are looking in from outside.
That gap is easy to overlook in a feature list, and it is the first thing an Android user will hit. The other question is what you hand over in exchange for the explanations.
What the ChatGPT Health tab connects to and keeps private
The help center lists Apple Health on iOS, medical records from supported US hospital systems, and apps including One Medical, Function Health, WHOOP, Oura, Garmin, Strava, and MyFitnessPal. Several accounts can be connected at the same time.
On privacy, OpenAI states that connected medical records and Apple Health information, and conversations that use them, are not used to train its foundation models or target ads, regardless of your model training setting. ChatGPT can only read from these sources. It cannot update your records.
Disconnect a source and data synced from it is deleted from OpenAI’s systems within 30 days. The conversations themselves stay until you delete them, which is easy to forget. You can keep health chats apart from your other memories through project settings or Temporary Chat.
The help center also notes that, as of September 2026, new Health plugin connections default to your global plugins permission setting, which is “Allow low-risk actions” unless you have changed it. You can adjust it under Settings, then Plugins, then Health, and our guide to phone tracking privacy settings covers the wider cleanup.
What to do before you trust an explanation
OpenAI says Health is designed to support, not replace, medical care, and is not intended for diagnosis or treatment. It is also not HIPAA eligible and does not offer a Business Associate Agreement, so it is not meant for clinical use.
The launch post says more than 260 physicians who have practiced in 60 countries provided feedback on model outputs over 600,000 times, using OpenAI’s HealthBench evaluation framework. That is useful context for how the answers were tuned. It does not promise that any single explanation of your own lab result is right.
Use the explanation to prepare better questions for your doctor, not to skip the appointment. If a number looks alarming, the useful move is to bring the chart and the question to someone who can examine you.
