ChatGPT Ads for Healthcare and Clinics: 2.4 Billion Weekly Messages, With a Policy Gate

OpenAI's September 2026 stat pack counts over 2.4 billion ChatGPT messages a week in healthcare, defined as care, fitness, nutrition, tests and results, and triage or symptom questions, up 31% between January and July 2026. It is the third-largest category OpenAI published, behind media and consumer goods, and the one with the most caveats attached before a clinic should spend on it.

What the 2.4 billion is

A weekly average of global user messages classified with health as the primary topic, across 6 to 26 July 2026. The category is wide on purpose: it includes the person asking what a blood test result means, the person asking for a training plan, and the person describing a symptom at two in the morning. OpenAI's own footnote applies to all of it: "Volume is not ad inventory or verified purchase intent." A large share of these messages will never see an ad because they are not commercial, and a further share sits in countries where ads do not serve.

What is left is still the largest pool of pre-appointment conversations that exists anywhere. The person who asks ChatGPT whether a persistent knee pain needs a specialist is doing, in one conversation, what used to take a search, three forum threads and a call to a friend. A clinic that can appear under that answer, in its own city, reaches the patient before the patient has a name to search for.

The policy gate

OpenAI opened ChatGPT Ads to health and finance advertisers in August 2026, and the same September pack that carries the volume number describes self-serve access in Europe as opening to "advertisers in approved categories whose ads clear its policy checks". For a clinic that means three things. The account goes through a category review as well as the standard business verification. Individual ads are reviewed against brand-safety policy, and our own experience on a different category is that plain descriptive copy can be rejected with a reason that is only visible in a popover. And the creative rules for health are stricter than for retail: no outcome claims, no before-and-after framing, no urgency around symptoms.

None of that is a reason to skip the channel. It is a reason to build the review stage into the timeline, to write the 100 launch ads conservatively and to keep the claims to what the clinic's own site already says.

Tracking, with a privacy decision attached

Conversion bidding needs a conversion event, and for a clinic that is the appointment request or the call-back. OpenAI's benchmark floor is five click-attributed conversions on the event before switching to Conversion Optimization, and the pixel plus Conversions API pairing added 57% to measured conversions across more than 1,000 advertisers in OpenAI's data. A clinic has one extra decision the retailer does not: automatic advanced matching is on by default and matches hashed form data against signed-in ChatGPT users. For a regulated practice that setting should be a written choice, made by whoever owns the privacy posture, before the pixel goes live. The conversion-tracking guide covers where the switch is.

What a clinic campaign looks like

Context hints describe the patient's situation. For a physiotherapy practice, as an example: recovering from a knee operation and looking for rehab nearby; recurring lower back pain from desk work; a runner with shin pain who wants to keep training; choosing between physio and osteopathy. Each is an ad group, each has creative that answers the situation and a landing page that opens on it, and each reports separately, so the practice learns within a month which situations its city actually asks about. That demand map is also the content plan for the organic side, where the same prompts decide whether ChatGPT recommends the practice without an ad at all. The eligibility checker confirms whether the practice's registered entity can open an account.