I don't need any dental work, at least right now. Having said that, I know I will one day. As will everyone. So I decided to run an experiment. One afternoon last week, I asked five AI tools -- ChatGPT, Copilot, Gemini, Google's AI Overview and Google's AI Mode -- three questions a prospective dental patient might actually ask. Who's the best implant dentist in Los Gatos? I need an implant consultation near Los Gatos, who should I call? Which cosmetic dentist do people recommend for veneers?
Then I asked every question again, same words, same tools, a few minutes later.
Two sets of three inquiries, five chatbots, minutes apart from each other. Thirty answers. Thirty-one different practice names. And in 13 of the 15 question-and-tool pairs, the second answer named a different set of practices than the first. On average, only about half the names carried over.
Ask the machine twice, get a different dentist.
That sounds like a problem. It's actually the most useful thing I've learned this year, and I'll get to why. First, who won.
Who AI kept naming
Los Gatos Oral & Facial Surgery was named in 15 of the 20 answers to the two implant questions -- on all five tools. It was named first more often than any other practice. It wasn't named once for veneers, which is exactly right: it's a surgical practice, and veneers aren't surgery.
Vasona Family Dentistry was named in 11 of the 30 answers, again on all five tools, and for both implants and veneers. Here's the interesting part: it was never named first. It's the dependable second or third name -- the practice the machine always seems to remember, even when it leads with someone else.
Los Gatos Dental Specialists was named in 9 of 30, and led the AI Overview's answer on one implant question.
Everyone else was a sometimes.
What didn't explain it
The easy story would be "most reviews wins." It didn't, at least not here. The most-recommended practice has 104 Google reviews and a 4.6 rating. Several practices with more reviews and higher ratings were named a fraction as often. One of the busiest, best-reviewed practices in town showed up in barely one answer in five.
My read -- and it is a read, from 30 answers, not a law of physics -- is that the machine is matching the question to the practice that is most unmistakably about that question. Ask about implants, and the surgical practice whose entire identity is implants and oral surgery wins. Ask about veneers, and the names change completely. Reviews still matter -- a 10,000-business study of local search this summer found review volume and map presence were the strongest signals of who AI recommends -- but they're the price of admission, not the whole game.
One more contrast worth sitting with: Google's own map results, the familiar three-pack under the search bar, didn't change at all between runs. The AI answer sitting right above it kept shuffling. The old scoreboard holds still. The new one never stops moving.
Why the shuffle is good news
If AI picked a winner once and locked it in, there'd be nothing to do. The practice that got there first would own the answer, and everyone else would be paying for ads around it.
That's not what's happening. The answer is a probability, not a ranking. It gets recalculated every time someone asks, from whatever the machine can find about you that day. Which means it can be moved -- and it means it can slip.
Visibility in AI isn't a trophy. It's a pulse.
That's also the catch. There's no "fix it once" here. The practices showing up consistently are the ones that keep giving the machine something to work with:
None of that is a campaign. It's a habit -- the same unglamorous upkeep that keeps a good practice's reputation good, now with a new reader on the other end.
What I'd do today if I owned a dental practice here
Ask ChatGPT and Google's AI Mode the three questions a new patient would ask about the thing you most want to be known for. Ask each twice. Write down who comes up, and whether it's you. Then look at what the practices that do come up have that you don't: the reviews, the plain-language listings, the mentions elsewhere.
It takes fifteen minutes. It's the most honest market research you'll do this year, because the machine doesn't know who your friends are.
How I did this: three questions, asked twice each on ChatGPT (with web search), Copiltot, Gemini (signed out), Google AI Overview and Google AI Mode on Monday, Oct. 5, 2026, between 12:50 and 1:10 PM Pacific -- 30 answers. A practice counted as named if its name appeared in the AI's answer text; map listings and ordinary search results weren't counted. Google review counts are as of Oct. 5. One afternoon is a snapshot, not a study -- that's coming next month for all of San Jose.
I'm Hani Durzy. I spent 25 years running communications at LinkedIn, eBay, and Realtor.com. Red Dog Strategies helps owner-led businesses control what customers, and the AI tools they now ask first, hear about them. Want to know what the machines say about your business when you're not in the room? Message me and I'll run your free Red Dog AI Visibility Score.
