300 million people ask ChatGPT about health every week. What that means for medical practices.

On 7 January 2026 OpenAI launched ChatGPT Health, a separate space inside ChatGPT where users can connect medical records and wellness apps and ask questions in the context of their own data. At launch the company said more than 230 million people asked health and wellness questions on ChatGPT every week. On 23 July 2026 it rolled the feature out to all adult users in the United States, including the free plan, and put the weekly figure at more than 300 million.

What ChatGPT Health is — and is not

OpenAI describes the tool as support, not care: it “does not replace the care and judgment of qualified medical professionals”, and at launch it was presented as a tool for navigation, not diagnosis or treatment. It was developed over two years with more than 260 physicians, and OpenAI says health conversations are not used to train its models. Medical-record integrations are US-only.

One detail from the July rollout matters more than the feature list. According to TechCrunch, around 70% of health-related questions during testing were asked outside the dedicated health space — in ordinary chats. People do not open a special mode to ask about symptoms. They ask the assistant they already use.

Why European practices should care anyway

At launch, ChatGPT Health was available only to users outside the European Economic Area, Switzerland and the United Kingdom, and the July rollout was announced for the US. A practice in Prague, Munich or Vienna could conclude that this is an American story. That would be the wrong conclusion. What is restricted in Europe is the feature that connects personal medical records — not the habit of asking a general-purpose AI assistant about health. That habit is already here, and every symptom conversation tends to end in the same place: see a doctor. The next question is which one, where, and whether they take new patients.

What the assistant knows about your practice

An AI assistant can only describe a practice from what it can find and read. If the practice website is outdated, if opening hours differ between the website and the Google Business Profile, or if the specialty and services are buried in images and PDFs, the assistant either guesses or names someone else. The practical checklist is short:

• Plain, current facts in text. Specialty, services, opening hours, contracted insurers, whether new patients are accepted, address and how to get there — written on the page, not inside images.
• Consistent data everywhere. Name, address and phone number identical on the website, Google Business Profile and medical directories.
• Healthcare structured data. Markup such as MedicalClinic and Physician tells machines unambiguously what the practice is and whom it serves.
• Answers to the questions patients actually ask. What to bring, how to book, what happens at a first visit — in plain language, on dedicated pages.

The rules still apply

None of this licenses symptom-checker content or promises of outcomes. Medical advertising rules — the Heilmittelwerbegesetz and professional codes in Germany, the advertising act and the Czech Medical Chamber’s code of ethics in Czechia — continue to apply to everything a practice publishes. The good news is that they point in the same direction as AI visibility: factual, verifiable, specific information is both what the regulators allow and what AI systems prefer to cite.

How FIB Consulting approaches it

We build practice websites so that the facts a patient — or an assistant — needs are current, consistent and machine-readable, and so that the practice team can update them without a developer. See what our medical IT work includes, read how AI assistants choose what to recommend, or get in touch.

Sources

OpenAI, “Introducing ChatGPT Health”, 7 January 2026 · Fortune, 7 January 2026 · Tom’s Guide, January 2026 (regional availability, physician involvement) · OpenAI, “Launching Health in ChatGPT”, 23 July 2026 · TechCrunch, 23 July 2026.