On July 23, 2026, OpenAI launched Health in ChatGPT to all US users across Free, Go, Plus, and Pro plans. The feature lets you connect Apple Health and supported medical records so ChatGPT can reason about your actual health data — not generic medical knowledge, but your specific lab results, medications, sleep patterns, and visit history. More than 300 million people already ask ChatGPT health-related questions every week. Now the tool can actually ground those answers in your personal information.

If you have ever tried to explain a complex medical history to a new doctor, you already understand the problem this feature is trying to solve. Your health data is scattered across patient portals, wearable apps, PDF lab reports, and the half-remembered details of what your doctor said during a visit you didn't take notes for. Health in ChatGPT attempts to pull that into one place where the model can reference it.

What it actually does

Once you connect a data source through the sidebar in ChatGPT, the model can use your information across all conversations — not just inside a dedicated Health tab. OpenAI tested a separate Health experience with early users earlier this year and found that more than 70 percent of health-related conversations happened outside it. People did not want to switch contexts. They wanted ChatGPT to know about their food allergy while helping pick a restaurant, or account for a recent injury when planning weekend activities. So OpenAI folded the data into the main conversation flow.

In practice, you can ask ChatGPT to compare a new lab result with previous tests, summarize what has changed since your last appointment, or explore how your sleep and activity data relate to how you feel. The model references your connected data with your permission. You control what gets connected and can revoke access at any time.

The privacy model

OpenAI is being explicit about the privacy guarantees here. Connected medical records and Apple Health data, and the conversations that reference them, are not used to train foundation models and are not used for ad targeting. That is the baseline promise, and it matters — health data is among the most sensitive information anyone can share with a software company.

The feature is available to logged-in users 18 and older on web and iOS. It is not yet available in Codex, which suggests OpenAI is keeping a tighter leash on the agentic side of this integration for now.

What early users report

The early feedback points to something specific: people with complex or overlapping health conditions find the most value. One early tester, a technical program manager named Blake, described using ChatGPT to connect scattered diagnoses, imaging results, and surgery notes into a coherent timeline. The model translated medical jargon into plain English and created shareable summaries for a physical therapist. That is not a hypothetical use case — it is someone using the tool to advocate for themselves more effectively in a medical system that does not always make that easy.

The model behind it

Health conversations are running on GPT-5.5 Instant for free users and GPT-5.6 Sol for paid tiers. OpenAI says GPT-5.6 Sol performs better on complex health questions, which makes sense — multi-condition reasoning across lab values, medications, and timeline data is exactly the kind of task where a stronger model earns its keep. The company says it is continuing to improve the models behind health conversations, specifically around recognizing when professional care is needed and asking for missing context rather than guessing.

What this means for how you use AI

Three hundred million health queries a week is a staggering number. It tells you that people already trust ChatGPT for health questions — the question was never whether they would use it, but whether it could give answers grounded in something real instead of generic medical text pulled from training data.

The practical takeaway is straightforward. If you have Apple Health data or can pull your medical records from a supported provider, connecting them gives ChatGPT the context to give you better answers. Not perfect answers — this is not a replacement for a doctor, and OpenAI is careful to say that. But better-informed questions for your next appointment, a clearer picture of trends in your lab work, and a tool that remembers what you told it about your health across conversations.

The interesting tension is that this feature only works if you trust OpenAI with your most sensitive data. The no-training, no-ads promise is the floor. Whether that is enough depends on your personal risk tolerance. For a growing number of people, 300 million times a week, the answer appears to be yes.

Sources: