Someone types “dentist in Denver” into Google. More and more, the first thing they see isn’t a list of websites or a map pack. It’s an AI-written answer that names about eight dental practices, describes each one, and tells them where to start. That answer is the new top ranking, and there is no way to buy a place in it today.
We wanted to know what Google’s AI looks at when it picks a dentist, what actually matters, and what agencies can do to make a client the dental practice it keeps choosing.
- The AI recommends dental practices it can find, verify and describe. It searches for top-rated dentists, checks a few city pages, and repeats whatever specific details those pages and the dental practice’s homepage give it.
- The homepage is the page that gets read. Practices recommended almost every time had their own site read 78% of the time, vs 24% for rarely picked ones.
- Recognition is what it repeats. It cites awards and credentials 6 to 7 times as often when describing its favorites.
- The signals stack. Dental practices with all four winning signals (outside listings, visible awards, the city in the homepage title and the dentist’s full name) are recommended in 31% of answers, nearly 3× the 12% for dental practices with one or none.
- Most dental markets are wide open. Each answer names about 8 dental practices, but only about 2 per city are locked in.
- The usual technical checklist doesn’t move the needle. Schema, booking widgets and rating markup are equally common on winning dental practices and on those the AI rarely picks.
To learn how dentists get recommended, we asked Google’s AI for one 2,500 times.
AI answers change from one run to the next, so a single answer is an anecdote. Fifty is a measurement. We asked the same unbranded question, “dentist in [city]”, 50 times in each of the 50 largest US cities. Every query then went through three passes:
- Ask. Every query went to Google’s Gemini model with live Search grounding, likely the same retrieval system behind AI Overviews. We recorded the full answer, every source it cited and every search it ran on its own.
- Extract. A second, independent model read each answer and pulled out every dental practice and dentist it recommended, in order.
- Compare. We traced all 21,957 citations to the exact page, crawled 809 dental practice websites, and compared dental practices the AI picks in most answers (“winners”) with those it picks in under 10% (“occasional picks”).
Dental practices are anonymized throughout. The findings are about the playing field, not any one player.
Before recommending a dentist, Google’s AI searches for ratings and reads dental practice homepages.
First, it searches for reputation. Before answering, the AI runs about two Google searches of its own. 98% of them ask for “top rated”, “best” or reviewed dentists in the city. Only 3 of 4,687 mentioned cost. Ratings likely decide who enters the candidate pool, yet no answer ever quoted a star rating.
Then it reads a short list of local pages. Over half its citations are dental practice websites, and 73% of those are homepages. The rest come from the same city-level pages in nearly every market.
More than half of what the AI reads is dental practice websites
Dental practices get recommended again and again when the AI can read, verify and quote them.
Dental practices the AI recommends in most answers differ from those it picks only occasionally in four ways. These are patterns, not proof of cause, but they line up across what the AI read, what other sites say and what the AI wrote.
Winners are read, vouched for and recognized. The technical basics are the same for everyone.
Winners are dental practices recommended in 40%+ of their city’s answers; occasional picks appear in under 10%.
Dental practices with all four signals are recommended nearly 3× as often
Among dental practices whose website the AI had already read, we counted four signals:
Average share of answers recommending the practice, by how many signals it has:
When the AI explains why it recommends a winning dental practice, it mentions awards 5.9× and credentials 6.7× as often, and writes 40% more about them. Location, price and emergency hours are said about everyone, so they don’t help anyone stand out.
The AI describes its favorites with awards, credentials and technology, not convenience or price
Becoming a fixture. Dental practices recommended in 80%+ of answers, compared with those at 40–79%, get their own site cited in 71% vs 41% of the answers that recommend them, and display recognition on their site 87% vs 63% of the time. They are also listed earlier.
The AI recommends the dental practices it can find, verify and describe. It finds them through top-rated searches, verifies them through city pages and recognition, and describes them with whatever specific detail their homepage provides. Schema markup, booking widgets, review widgets and listed prices don’t separate winners; nearly every dental practice already has them.
Seven ways to get a dental practice recommended by Google’s AI.
- 01
Make the homepage say everything that matters
Practice name, city and neighborhood, each dentist’s full name and credentials, standout services and recognition, in visible text the AI can read.
WhyDental practices recommended almost every time had their own site read 78% of the time, vs 24% for rarely picked ones, and 73% of those reads were the homepage. - 02
Get onto the city pages the AI already cites
Check which directories, “top rated” lists, magazine features and subreddit threads it reads in the client’s city, and earn a genuine place on them.
WhyWinners were named on those pages 42% of the time vs 24%. - 03
Earn recognition the AI can quote, then display it
City-magazine Top Dentist selections, board certifications and fellowships.
WhyThe AI cites awards 5.9× and credentials 6.7× as often for winners, and dental practices picked every time display recognition 87% vs 63%. - 04
Own a neighborhood and a patient need
Name the neighborhood and the one or two needs the practice serves best (families, cosmetic, anxious patients, emergencies), plus the specific technology and services behind them.
Why61% of answers are sorted by neighborhood and 67% by need, so each heading is a smaller list that is easier to win. Generic claims like “convenient” and “affordable” don’t separate anyone. - 05
Keep reviews growing and insurance listings accurate
Grow reviews steadily, and make sure the client is listed correctly in every insurer directory it accepts and in Zocdoc’s insurance settings.
Why98% of the AI’s searches ask for top-rated dentists, and 87% of answers send patients to their insurance directory before booking. - 06
Make every office read like a local dental practice
For a multi-location or DSO-owned client, give each office its own page with its neighborhood, its dentists and its recognition.
WhyNational chains are nearly invisible (one of the largest appeared in 5 of 2,500 answers), while multi-location groups that read as local do well. - 07
Measure position across many runs, and check what the AI says
Track the recommendation rate and where the client is listed over time, and correct wrong names or details at the source.
Why43% of dental practices appear in just one answer out of 50, so a single screenshot misleads, and locked-in dental practices take 48% of first positions.
How we measured dental recommendations, and what the data can’t tell us.
2,500 answers from gemini-3.8-flash with Google Search grounding, collected September 25–26, 2026, 50 per city for the 50 largest US cities by population. Winners are dental practices recommended in at least 40% of their city’s answers; occasional picks appear in under 10%. Rates at 50 runs carry about ±7 points of sampling error, and two separate 50-run samples of the same city agreed closely (correlation 0.97). Reddit and Zocdoc block automated reading, so presence on those sites is undercounted. Google review counts and ratings were not collected. Findings are associations, not proof of cause.