Which dental clinics do AI assistants recommend in Tijuana?

In July 2026 we asked four AI assistants the forty questions a US patient actually asks before booking dental treatment in Tijuana — about safety, cost, credentials, logistics, and which clinic to choose. Every question was asked repeatedly, on every assistant, and every answer was recorded with a timestamp.

That produced 440 recorded answers, 2,407 citations, and one finding that overrides all the others.

Nobody owns Tijuana

Across those 440 answers the assistants named 61 different clinics, an average of 4.6 per answer and never more than seven at once.

Every single one of those 61 clinics was the top recommendation in at least one recorded answer.

Not one of them was the top recommendation reliably. We grouped the answers by question and assistant — 160 groups, each asked two or three times in a fresh session — and compared the first clinic named in each repeat. In 159 of the 160 groups, the top recommendation changed between repeats.

This is the entire story of the market. There is no clinic that AI assistants recommend in Tijuana. There is a pool of roughly sixty, and each answer draws a handful of them in an order that does not survive being asked again five minutes later.

For a clinic owner, that cuts both ways. The good news is that no competitor has locked up the category, and nothing you are seeing is a rival who bought the top spot. The bad news is that checking your own visibility by asking ChatGPT once, which is what nearly everyone does, tells you essentially nothing — you sampled one draw from a pool of sixty.

What the assistants are actually reading

If no clinic controls the answer, something does. The citations are far more concentrated than the recommendations: 96 domains supplied all 2,407 citations, and the top ten supplied half of them.

By category, across all 2,407 citations:

Three things in that table are worth sitting with.

Clinics are citing clinics. The largest single category is content published by dental clinics themselves — and when an assistant cites ilovemydentist.com.mx or advancedsmilesdentistry.com to answer which clinic should I choose, it is not treating that page as a competitor's marketing. It is treating it as a neutral guide to the city. Those clinics wrote the article that recommends the shortlist, and the assistant repeats the shortlist.

Reddit outweighs the entire independent press. One forum supplied 4.4% of all citations; every independent editorial and news source in the market combined supplied 3.9%. What patients say to each other in public is a larger input to these answers than everything written about Tijuana dentistry by journalists.

Four aggregators carry more than a quarter of the market. Dental Departures, Bookimed, WhatClinic, and DentalMexico together account for 21.6% of every citation recorded. A clinic's entry on those four sites is not a marketing nicety — it is a quarter of the evidence base the assistants draw on.

What this means if you run a clinic in Tijuana

Your listing on four or five aggregators matters more than your website redesign. Not because those sites send traffic, but because they are what the assistants read. An outdated price or a closed service on one of them is being repeated to patients at scale, and it costs nothing to check.

The clinics winning are the ones publishing guides, not ads. The highest-cited clinic-owned pages in this market are articles that compare options, explain credentialing, and answer safety questions — the format assistants need. The clinic that writes the honest guide to choosing a Tijuana dentist gets quoted in answers about every clinic, including its competitors'.

Being absent from Reddit is a measurable gap. Not something to manipulate — astroturfing a subreddit is both against the rules and easy to spot — but worth knowing that patient conversation carries more weight here than press coverage.

One check tells you nothing. With 61 clinics in the pool and the top answer changing on nearly every repeat, a single conversation with ChatGPT is a coin toss you will misread in whichever direction it lands.

Method, and what this report deliberately does not say

Forty questions, written to match real US-patient intent across five stages: discovery, comparison, trust, price, and logistics — eight questions each. Asked on 30 July 2026 in English, to OpenAI (chat-latest), Anthropic (claude-sonnet-5), Google (gemini-3.5-flash), and Perplexity (sonar). OpenAI, Gemini, and Perplexity were asked every question three times; Anthropic twice. 440 answers completed, none failed. Every mention, position, and citation in this report is extracted from those recorded responses.

Two limits, stated plainly because the whole point of this exercise is that findings should be checkable.

We are not publishing a clinic ranking, and the mention counts should not be read as one. Clinic names were not resolved across spelling variants, so at least three clinics in this data appear under two names each and have their counts split between them. A leaderboard built on that would be wrong in a way that flatters whoever happens to have one consistent name. Fixing it is straightforward and we have not done it yet.

This run was commissioned by one Tijuana clinic, whose own figures and domain are excluded from every table above. Their questions were market-level and generic, which is why the run describes the city rather than the client — but a market report should not double as a customer's private diagnostic, so their numbers stay out.

Both limits argue for the same follow-up: a clean, uncommissioned run with resolved clinic names, published as a real ranking. That is the version worth arguing about, and it is the one we intend to publish next.

If you run a clinic named in this data and want to know your own numbers — how often you were named, on which assistants, at what position, and which sources put your competitors ahead of you — that is the audit, and it is the same 440 answers pointed at one business.