“Who is a good dentist near me who can see me this week?” That sentence, or one very close to it, is being typed into AI models thousands of times a day by prospective patients. The model returns a paragraph with a short list of practices and a sentence on each. The patient calls one of them, or opens the website of whichever one got mentioned first.
That prompt is doing more work than it looks like it is doing. Practices trying to understand AI search often approach it as a ranking problem, the same way they approached Google in 2015. That framing gets in the way. The AI does not rank practices against each other so much as read the available information about each one and decide which ones fit the patient’s stated criteria. What practices should be paying attention to is what those criteria actually are, and whether the AI can see that a specific practice meets them.
What Is Actually Inside That Prompt
Three constraints, spoken in one sentence. “Good” is a reputation criterion. “Near me” is a geographic and location-services criterion. “This week” is an availability criterion. A patient searching Google in 2018 would have typed “dentist 60540” and done the filtering themselves by opening tabs, scanning review counts, and calling around to ask who had an opening. The 2026 patient is asking the AI to do that filtering before showing them anything at all.
More elaborate prompts get more specific. Insurance carrier, procedure type, sensitivity concerns, provider gender preference, hours that work around a work schedule, whether the practice sees children. Patients are telling the AI what they need in the same way they would tell a friend who is recommending someone, and they expect the AI to filter accordingly.
There is a side effect worth naming. In the Google era, a practice never saw the reasoning behind a search, only the keyword. AI queries expose the reasoning. When practices watch how patients are actually phrasing their questions to AI models, they are watching a live feed of the criteria patients care about most, in the patients’ own words.
What the AI Needs to Answer Well
For any given prompt, the AI has to find practices where the criteria are visibly true. Not implied, not likely, not inferable from a homepage that talks about “gentle, family-focused care.” Visibly true, in a form the model can read and cite.
That means recent reviews with substantive text, not just star counts. It means a Google Business Profile with current hours, current services, current photos, and a confirmed new-patient status. It means dental schema markup that describes the practice as an entity, with specialties, service area, and provider information formatted so a language model can parse it without guessing. It means the practice’s own website carrying enough substantive content, in real sentences, that the AI has something to draw from when writing that one-paragraph summary.
AI models read text, not design. A beautiful site with three paragraphs of copy gives a language model less to work with than a plainer site that answers real patient questions in real sentences. Most practices have some of the technical pieces in place. Very few have all of them, and the gaps are where the AI’s answer either omits the practice entirely or hedges in a way that costs the practice the click.
Why Some Practices Get Named and Others Do Not
The consistency question is the one that separates practices AI models mention from practices they do not. When the practice name, address, phone number, hours, and specialties are the same on Google, on the practice website, on directory listings, and in structured data, the language model treats the practice as a coherent entity and speaks about it with confidence. When those signals conflict, the model hedges or leaves the practice out of the answer to avoid contradicting itself.
The work of getting recognized in AI answers is unglamorous. It is auditing citations. It is fixing hours that changed six months ago and never got updated in three places. It is writing longer, more substantive answers to questions patients actually ask on the phone. Practices doing this work are being named. Practices assuming the model will figure it out from a homepage are not.
What to Actually Do About It
The action for a practice this quarter is straightforward. Query ChatGPT, Perplexity, and Google’s AI Mode from a location in the practice’s target area, and ask each of them for a good dentist in the practice’s zip code. Read what they say. If the practice is not named, or is named with the wrong information, the gap is now diagnosed rather than theoretical.
The fix is usually not one thing. Google Business Profile fields need to be current: hours, services, photos, provider information, new-patient status. The website needs to be readable to a language model, which means substantive content in real sentences and not marketing lines wrapped in graphics. Schema markup needs to describe the practice as an entity the AI can cite with confidence.
The uncomfortable part is that most of this work is not glamorous, and none of it produces a screenshot that looks impressive at a partner meeting. What it produces is a practice that AI models can find, understand, and recommend when a patient asks a question in real sentences.
Schedule a discovery call to see what this looks like for a specific practice.
