You use ChatGPT Health by opening Health from the ChatGPT sidebar, connecting a source like Apple Health or your medical records, and then asking questions in a workspace that keeps that data separate from your regular chats. It rolled out to all US users the week of July 23, 2026, and it's built to summarize test results and prep you for appointments, not to diagnose anything.
Short answer: Open ChatGPT, tap Health in the sidebar (under "More" on mobile), then Connect Health to link Apple Health, a supported hospital system, One Medical, or Function Health. ChatGPT asks permission before using that data in a chat. It's free on every plan — Free, Go, Plus, and Pro — for US users 18 and older, on web and iOS.
In my testing, the setup itself takes about two minutes — the part that actually matters is what you do with the permission prompts afterward. OpenAI says roughly 300 million people already ask ChatGPT health questions every week, per TechCrunch’s coverage of the July 23 rollout, and Health is the company's answer to the fact that most of those questions were already happening without any of your actual records attached. Here's what connecting them changes, what it doesn't, and where I'd stop trusting it.
What you'll need
A ChatGPT account on any plan — Free, Go, Plus, or Pro — works, since Health isn't gated behind a paid tier. You need to be 18 or older and located in the US; OpenAI hasn't extended the feature outside the US as of this writing. If you want to connect actual records rather than just Apple Health, check that your provider is in the supported network first — not every hospital system is hooked up yet, and Health also connects to One Medical and Function Health directly. You don't need anything installed beyond the ChatGPT app itself; the web version and iOS both support it, though voice input for Health questions is iOS-only right now.
Step-by-step: how to use ChatGPT Health
1. Open Health from the sidebar
On desktop, Health sits in the left sidebar. On mobile, tap More to find it if it's not already showing. This opens a separate space from your normal chat history — OpenAI describes it as running on its own memory, isolated from the rest of ChatGPT.
2. Connect a data source
Tap Connect Health or Get started, then choose what to link: Apple Health, a supported medical-records provider, One Medical, or Function Health. Apple Health pulls in whatever you've already synced there — workouts, sleep, medications logged through other apps like MyFitnessPal. Medical records pull in labs, visit notes, and diagnoses directly from your provider's system.
3. Approve what ChatGPT can see
The first time you ask a health question, ChatGPT asks for permission before reading connected data. You get three choices: allow it for this one question, allow it going forward, or say no. In my testing, the one-time approval was the better default to start with — you can see exactly what it pulled in before deciding whether to open the door permanently.
4. Add anything that isn't synced automatically
Conditions, medications, and family history that aren't in Apple Health or your provider's system can be typed in directly inside Health. This matters more than it sounds — a lot of relevant context (an allergy, a past surgery, a parent's diagnosis) never makes it into a synced record at all.
5. Ask a real question, not a symptom dump
Once something's connected, ask it to do something specific: summarize your last three cholesterol readings, draft questions for an upcoming appointment, or explain a term from a lab report. Vague prompts like "what's wrong with me" produce vague, hedging answers even with your records attached.
6. Manage or revoke access any time
Go back into Health settings to see what's connected, switch permissions from one-time to always-on or back, or disconnect a source entirely. Nothing here is a one-way door.
Example prompts you can copy
- "Here's my last three A1C readings from my labs — summarize the trend in plain English and tell me if it's moving in the wrong direction."
- "I have an appointment with my cardiologist next week. Based on my connected data, write five questions I should actually ask."
- "Explain what 'elevated ALT' means on this lab result, and what usually causes it."
- "Look at my Apple Health sleep data from the last two weeks and tell me if anything stands out."
- "I'm starting a new medication [name]. What does it typically interact with, and what should I watch for the first week?"
Common mistakes to avoid
The biggest one: treating an answer as a diagnosis instead of prep material for an actual doctor. OpenAI's own disclaimer is blunt about this — ChatGPT "is not a doctor and should never be used as a substitute for medical care, diagnosis, or treatment," and it can still get things wrong. Second, granting always-on access before you've seen what a one-time approval actually surfaces; start narrow, then widen it once you trust what it's doing with the data. Third, assuming a connected record is complete — OpenAI itself notes connected health information may be outdated or incomplete, so anything that changes a real decision is worth checking against the original chart or portal. Fourth, connecting records and then still asking generic questions instead of ones grounded in the data you just linked, which wastes the entire point of setting it up. And fifth, forgetting this exists in a US-only, 18-and-over rollout right now — if you're outside the US or helping a minor, Health simply isn't available yet, synced data or not.
ChatGPT Health vs. asking ChatGPT normally
| Regular ChatGPT chat | ChatGPT Health | |
|---|---|---|
| Sees your actual labs/records | No — you'd have to paste them in | Yes, once connected |
| Data used to train the model | Yes, unless you opt out in Data Controls | No, never, per OpenAI |
| Memory | Shared with your other chats | Separate, isolated space |
| Cost | Included on every plan | Included on every plan (Free–Pro) |
| Availability | Global | US only, 18+, as of July 2026 |
| Best for | Quick, general health questions | Trend-spotting and appointment prep using your own data |
The privacy separation is the actual reason to use Health over just pasting a lab result into a normal chat: OpenAI says Health data runs on its own encrypted memory and is never used to train its models, a stronger default than the standard consumer tiers, where training is on unless you switch it off yourself — the same setup I found when I looked at ChatGPT’s work-account privacy settings.
Tools that make this easier
If you're not sure ChatGPT is worth paying for at all before you start connecting medical records, my guide to ChatGPT’s free plan covers exactly what the no-cost tier includes — Health works on it too, so there's no reason to upgrade just for this. Health is web and iOS only right now, so if you're mainly on your phone, my how to use ChatGPT on iPhone guide covers the app basics worth knowing first. If your interest is really in getting more organized, useful answers out of ChatGPT generally, not just the health workspace, my broader how to use ChatGPT guide is the place to start. Worth comparing before you commit to one assistant for anything sensitive: how to use Claude AI covers Anthropic's equivalent privacy defaults, which are relevant if you'd rather keep health questions off OpenAI's infrastructure entirely. And if multi-step research tasks are more your use case than a health workspace specifically, how to use ChatGPT agent mode covers the feature built for that.
My take
Health is a genuinely useful addition if what you want is help making sense of your own data — trend-spotting across labs, turning a confusing result into plain English, building a list of questions before an appointment you'd otherwise walk into cold. What it isn't, and what OpenAI is careful to say it isn't, is a replacement for the actual visit. Start with one-time permission instead of always-on, verify anything that would change a real decision against your provider's portal, and treat every answer as a draft to bring to a doctor rather than a final word. Used that way, it's a solid prep tool. Used as a stand-in for medical advice, it's exactly the wrong tool for the job.
Frequently Asked Questions
Is ChatGPT Health free?
Yes. It's available on every plan — Free, Go, Plus, and Pro — at no extra cost, for logged-in US users 18 and older.
How long does it take to set up ChatGPT Health?
A couple of minutes to connect a source like Apple Health. Connecting full medical records can take longer if your provider requires you to confirm the link on their end first.
Is my data private in ChatGPT Health?
OpenAI says Health runs in a separate, encrypted space and that connected data is never used to train its models — a stronger default than standard ChatGPT chats, which train on your conversations unless you opt out.
Can ChatGPT Health diagnose a condition?
No. OpenAI states plainly that ChatGPT is not a doctor and shouldn't replace medical care, diagnosis, or treatment. Use it to understand and prepare, not to decide.
Does ChatGPT Health work outside the US?
Not yet. As of the July 2026 rollout, it's limited to logged-in users 18 and older located in the United States, on web and iOS.