The AI Visibility Brief, Issue 1: Win the Citation, Lose the Patient

This week's theme: you can win the AI citation and still lose the patient.

Matt's Take

Twenty-five years in healthcare marketing taught me one thing about new channels: the first number everybody reports is usually the wrong one. With AI search, that number is referral traffic. Three studies just showed it's small, and plenty of CMOs are reading that as permission to wait.

Here's what those studies didn't capture. When a patient asks ChatGPT where to get a hip replaced, the answer is the decision. No click required. And this month, Google started writing its own summaries of the pages it cites, so even winning the citation doesn't guarantee your message survives.

So here's my question: do you know what AI said about you this morning?

What changed in AI search this week for healthcare?

Google is writing its own summary of your page

On October 23, SEO consultant Brodie Clark spotted AI-generated summaries inside the citation cards beside Google AI Overviews, and Search Engine Roundtable reproduced it in its own tests. The card no longer shows your page's text. It shows Google's summary of your page, shaped by what Google thinks matters for that query. Google hasn't commented, and it doesn't appear on every search yet.

For a health system, this splits the scorecard in two. Getting cited is step one. Step two is what Google says about you when it cites you. If your orthopedic page reads like a brochure and a competitor's states outcomes, access, and specialties plainly, Google's summary will reflect that difference, and so will the patient's choice. Start capturing citation card text when you check your priority queries, not just whether you appear.

Structure predicts citation, even against bigger brands

SISTRIX compared 96,418 cited and non-cited pages across ChatGPT, Google AI Mode, and AI Overviews, matching each cited page with a similar page on the same website. That design largely removes brand strength from the comparison. Tables of contents, Article schema, five or more subheadings, genuine FAQ schema, author boxes, and visible last-updated dates all raised the odds of citation. Pages with many of these features were cited about 2.1 times as often, but only 10% of cited pages had them. llms.txt showed no benefit.

The study covers German-speaking markets and shows correlation, not cause. Still, it undercuts the most common excuse we hear: "We can't compete with Mayo." Page-level structure is a gap you can close.

What else should healthcare marketers know this week?

Young adults use AI for sensitive health questions. A Rutgers study in JMIR Formative Research surveyed 995 US adults aged 18 to 29 and found 32% used ChatGPT, Gemini, or Meta AI for drug and alcohol information. Of those, 93% also used Google for the same purpose. For behavioral health and addiction treatment programs, where patients research privately, both AI Overviews and chatbots now shape the first impression.

Gemini now recommends places using your Maps history. With permission, Gemini can draw on a user's Maps searches, saved places, and calendar to make day-specific recommendations. A recommendation like "an urgent care near your 3pm meeting" depends on Google Business Profile accuracy and reviews. For multi-location systems, location data now feeds AI recommendations directly, not just the map pack.

ChatGPT ads may soon be bought programmatically. Search Engine Roundtable reported that some ChatGPT ads now route clicks through Amazon's DSP tracking, which suggests programmatic buying. OpenAI hasn't commented. If that inventory opens to the tools media agencies already use, spend will show up next to AI answers the brand doesn't control. Know what the organic answer says before the media plan lands.

Where is the Trust Tax showing up?

Seer Interactive ran an identical ChatGPT prompt twice, eight minutes apart. The two runs shared zero exact search queries. In one capture, ChatGPT retrieved 40 pages and cited only 6. The model even switched versions mid-session. The author notes this is a single-prompt sample.

Two lessons for health systems. First, the "I checked ChatGPT and we showed up" report from a board member means very little; one run is a coin flip. Second, look at the 34 pages that were retrieved and dropped. AI found them and chose something else. When a health system's signals conflict, with different provider details on the website, the directory, and Google Business Profile, it ends up in that discarded pile. That's the Trust Tax: the penalty AI applies to an inconsistent digital presence.

What does the AHS benchmark show?

In version 1 of the AHS National AI Visibility Benchmark Study, we scored 50 top-ranked programs across five specialties. The national average was 6.2 out of 10, and no program reached 9.0. These are the most recognized names in American medicine, and none of them has AI visibility locked down.

That finding is the backdrop for this week's theme. If the strongest brands in the country leave this much on the table, regional systems have a real opening, and a real exposure. Version 2 of the study releases Monday, November 2, with the 2026 to 2027 U.S. News top programs in cardiology, orthopedics, oncology, women's health, and neurology.

What is the industry saying about AI visibility?

The debate over whether AI visibility matters shifted this week. Rand Fishkin of SparkToro, responding to three skeptic studies, concluded that AI sends almost no traffic and should be budgeted as a brand channel. Mike King of iPullRank argued the studies grade a brand channel with a traffic ruler. Two of the loudest voices on opposite sides now agree on the frame: measure presence in the answer, not referrals.

Content Marketing Institute, citing Conductor's AI Brand Recommendation Study, added the detail that matters most for healthcare. Up to 72% of AI answers to educational questions mention no brand, while one brand led 91% of comparison answers. Symptom questions rarely name a provider. "Best hospital for" questions are winner-take-most. We break down what this means for health systems in this week's post.

Want this in your inbox every week? Subscribe to The AI Visibility Brief, weekly intelligence for healthcare marketing leaders.

If your health system's AI visibility is a question mark, that's worth a conversation. Contact us.