Does ChatGPT Recommend My Dental Practice? The New-Patient Test

Dentistry is the trade where the AI answer is filtered before it is ranked. A new patient does not ask “who is a dentist in Gilbert.” They ask “which dentist in Gilbert takes Delta Dental and is accepting new patients,” and any practice the engine cannot confirm on both counts is dropped before the list is even ordered. This guide is how to run that test for your practice, the reasons dentists specifically are not named, and what to fix first.

Ten minutes for the test. Each two-minute check below is something the front desk can do between patients.

Why dentistry is different from every other local business in AI search

Two patients, and the engines treat them differently:

• The toothache patient. A broken tooth, a lost crown, pain that started overnight. Their question is “emergency dentist near me open today,” and it is answered from listing data: hours, phone, whether the practice says it takes same-day emergencies. They call the first name that fits.
• The new patient. New to town, new insurance, unhappy with the last office, a child who needs a first visit. Their question has conditions attached: the insurance they carry, whether new patients are accepted, whether the practice does sedation, whether it sees children. The engine filters on those conditions first and only then chooses.

The filter is what makes dentistry unusual. A plumber is named or not named. A dentist is first checked against the patient’s constraints, and a practice whose insurance list and new-patient status are not written anywhere an engine can read them fails the filter silently. Test both patients.

How to run the test properly

Do not ask the AI about your practice by name. That tells you whether the model has heard of you, not whether it would send a patient to you. Ask the questions patients ask, with the conditions patients attach. Four rules:

1. Use the patient’s words, conditions included. “Dentist in Gilbert AZ that takes Delta Dental and is accepting new patients.” “Emergency dentist in Mesa open today for a broken tooth.” “Pediatric dentist in Chandler that takes AHCCCS.” “Gentle dentist near Queen Creek for someone with dental anxiety.”
2. Name the city, and the neighbouring one. Patients will drive fifteen minutes; the engines answer by town. Absent for the town next door is common and fixable.
3. Ask each question three or four times in fresh chats. The answers vary run to run. One run is an anecdote.
4. Ask in more than one engine. ChatGPT, Perplexity and Google’s AI features read different sources. Dental answers lean on Google’s own listings and on the health directories, and each engine has its own favourites among them.

Write down every practice named and how often. That list is your answer, and every practice on it has told the engines, somewhere it can read, that it takes that insurance and those patients.

Reason one: the engines cannot read your website

What it is: your site refuses the crawlers that decide citation, through a robots.txt rule or a firewall setting. The crawler that decides whether you can be cited is a different crawler from the one that collects training data, for OpenAI, Perplexity and Anthropic, and blocking the wrong one is common.

Why it hits dental practices: dental websites are very often built and hosted by a specialist dental marketing company on a platform the practice never logs into, and a large share sit behind Cloudflare, which has blocked AI crawlers by default for new domains since 1 July 2025, according to its own announcement. Nobody at the practice is told, and the vendor’s monthly report does not mention it.

Reason two: your practice has more than one Google profile, and they disagree

What it is: Google allows a listing for the practice and a separate listing for each dentist in it. Most practices end up with both, often created by different people years apart: the practice listing with one phone number and set of hours, a listing for the founding dentist with another, and one for an associate who left in 2022 and still shows up as practising there.

Why it hits dental practices: the toothache patient’s question is answered from listing data, and when the engine finds three listings at one address with different hours, phone numbers and categories, it does not pick the best one. It treats the practice as unclear and names one that is not. The category matters too: the practice listing should be “Dentist” with additional categories for what you actually do, such as “Cosmetic dentist,” “Pediatric dentist” or “Emergency dental service,” because the patient’s conditions are matched against them.

Reason three: the two facts the filter needs are not written anywhere an engine can read

What it is: your website says you are gentle, modern and family-friendly. It does not say, in plain text, which insurance plans you accept by name, or whether you are accepting new patients. Sometimes it says “we accept most major insurance,” which is a sentence that answers no patient’s question.

Why it hits dental practices: this is the filter. A patient carrying Delta Dental PPO asks for a dentist who takes it, the engine looks for practices whose pages say “Delta Dental,” and a practice that says “most major insurance” does not match. Same for new patients: an engine will not guess. The practices that get named are the ones with a page that lists plan names, states new-patient status, and gives the new-patient exam price or what a first visit includes, all as plain text a machine can lift. That page is worth more to AI visibility than any other page on a dental site.

Reason four: your details disagree across the health directories

What it is: your name, address, phone number and provider list are not identical across your website, your Google listings, Healthgrades, Zocdoc, Yelp and the directories your insurance networks publish.

Why it hits dental practices: dentistry has a layer of directories most trades do not: the health-specific ones, and the insurance networks’ own provider finders, which are created from credentialing data and are often years out of date. An engine gathering from all of these finds a provider listed at your practice who left, a phone number from before the office moved, and a practice name with and without the doctor’s name in it, and does not reconcile the differences. A practice with one consistent identity across all of them gets named; one with three gets skipped.

Reason five: your reviews say “great dentist” where the engines need “took my insurance and got me in the same day”

What it is: you have reviews, and they are warm, and they are useless to the filter. “Best dentist ever, so gentle” tells an engine nothing about insurance, new patients, emergencies or children.

Why it hits dental practices: for a question with conditions, the engines read the review text on the sources that rank for your town, usually Google and Healthgrades, looking for the conditions. A review that says “got me in the same afternoon for a cracked tooth and they took my Cigna” is evidence for two patient questions at once. You cannot write those reviews, and you must not ask patients to say specific things, but you can ask at the right moment, after a same-day emergency or a smooth first visit, and let the specifics come naturally. Two cautions: no review count unlocks anything, and marking up patient testimonials as review schema on your own site risks removal from the listing the toothache answer depends on. Health privacy rules also mean you never reply to a review with any detail about the patient’s care.

What to do first, in this order

1. Confirm the engines can read you. Free, ten seconds, and nothing else matters until it passes.
2. Consolidate the Google listings. One primary practice listing with the right categories; every practitioner listing agreeing with it; departed dentists removed.
3. Write the filter page. Insurance plans by name, new-patient status, what a first visit includes and costs. Plain text, linked from the navigation, echoed on the listing.
4. Make your identity identical everywhere, including the insurance networks’ provider finders.
5. Earn reviews that carry the conditions, by asking at the right moment and never scripting them.
6. Re-run the test in a month, same questions, same engines, several runs each.

Steps two and three cost an afternoon and no money, and between them they answer the two questions every new patient attaches to their search. Nothing here needs a developer, and all of it is wasted if step one fails.

---

Questions, answered

Written by Tyson McKay, founder of AEOCheck. We publish how we check things, including our own claims. Last verified 14 September 2026.