How do patients find a dentist in 2026? Through six overlapping paths: a friend's recommendation they verify in Google, the map pack, the AI Overview above it, a direct question to ChatGPT or Perplexity, the insurance directory, and a final pass through your reviews. No single path decides it; patients rarely stop at one before booking. And every path reads the same evidence about your practice: your reviews, your profile, your treatment pages, and what third parties say about you.
Here's each path, and what decides who wins it.
The referral now ends in a search box
A friend says "go see Dr. Patel." That used to be the whole path. Now it's the first half, because the patient searches the name before calling, and what comes back decides whether the referral survives. Your site, your profile, a wall of recent reviews: referral confirmed. A thin homepage, a three-year-old review, a competitor's comparison page: the sure thing starts leaking.
The mechanism is corroboration. A name from a friend is one data point, and the patient is looking for a second source that agrees. You win the referral path by making your brand search boring: your site, your profiles, and your reviews all telling the same story. Mismatched names or addresses across listings break that agreement, for patients and for machines alike.
Who wins the map pack?
Type "dentist near me" and the map pack is the first real estate. Three practices, ratings attached. Proximity you can't control. The rest you can.
Completeness first. Primary and secondary categories, every treatment listed as a service, photos, hours, Q&A answered. Most dental profiles sit half-finished, so a complete one is a genuine edge rather than table stakes. Then review signals: a steady stream of recent reviews beats a big stale pile, because recency tells Google, and the patient, that you're still the practice those reviews describe.
There's a quiet compounding effect here too. The profile doesn't just power the map pack; it feeds the directories, roundups, and AI answers built downstream of it, so every empty field starves more than one path at once. The full build order for this evidence layer is in local SEO for AI search.
What decides the AI Overview above it?
Question-shaped searches increasingly return an AI Overview before anything else: "how much does invisalign cost," "is a cracked tooth an emergency," "veneers vs bonding." Those answers are assembled from pages that answer the question directly, near the top, in plain language.
That's what treatment pages are for. One page per treatment you want patients from, with the question answered in the first sentences: what it costs in your market (ranges you're willing to publish), how long it takes, who it's for. A paragraph about invisalign buried on your homepage gives the Overview nothing to cite. Publishing cost ranges feels risky to most practices, which is exactly why it works: the practice that answers the money question becomes the source, and the patient arrives pre-qualified instead of sticker-shocked. The structure that earns those citations is the same one that wins AI Overview placement in every category.
What happens when a patient asks ChatGPT?
Assistants don't rank dentists. They describe them. Ask for a dentist in your city and the model builds a shortlist from reviews, maps data, directories, and whatever local "best dentist" lists it retrieves, then names two or three practices with reasons attached. "Patients mention gentle cleanings and easy scheduling" is a sentence assembled from your reviews, not your website.
Two things decide this path. Evidence, meaning the sources above mention you with specifics worth quoting. And entity confidence: the model has to be sure you're one real practice with one name, one address, one phone number, before it puts your name in an answer. That confirmation machinery is covered in how LLMs pick brands, and if assistants draw a blank on your practice entirely, that has specific causes. Closing those gaps for practices is exactly what we build in our AI SEO for dentists work.
The insurance directory is still a front door
Plenty of patients start from coverage, not quality: open the insurer's find-a-dentist tool, filter by city, and shortlist from whoever appears. Two failure modes lose this path. Your directory listing is wrong, an old address or a name variant that doesn't match your Google profile, so the cross-check fails when the patient searches you. Or you're invisible for the follow-up search, "dentist that takes [insurance] in [city]," because your accepted plans live in a PDF or nowhere at all.
The fix is unglamorous: audit your listing with every insurer you accept, match it to your canonical name and address, and publish a plain page listing the plans you take, in text a search engine can read. Few practices bother, which means the ones that do collect an entire discovery path with an afternoon of work. That's the point.
Reviews get read twice
Every path above ends in the same place: the patient reading your reviews before booking. And your reviews now have two audiences, humans and machines. The patient skims the most recent ones, looks for their situation ("crown," "kids," "dental anxiety"), and checks how you answer the bad ones. The machine quotes patterns: specific treatments, specific outcomes, recency.
Which means the same review does double duty. "Great dentist!" helps neither reader. "Fit me in same day for a cracked crown, and the billing matched the estimate" gives a patient their situation and a model a sentence to build a recommendation from. Ask at checkout, make it one tap, and nudge toward specifics with the question itself: "would you mention what we did for you?" Respond to every review, especially the rough ones, because your reply is the only part of a bad review you control. And never buy or fake a single one; fabricated reviews poison the whole evidence chain you just spent months building.
Test your own practice tonight
You've never read what the machines tell patients about you. Fifteen minutes fixes that. Paste this into ChatGPT with web search on, then again into Perplexity:
Act like a patient looking for a dentist in [your city].
Answer each separately and list the sources you used:
1. Which dentists in [your city] would you recommend, and why?
2. What do you know about [your practice name]? Is it reputable?
What do patient reviews say?
3. [your practice name] vs other dentists in [your city]:
how do they compare?
For each answer, tell me whether [your practice name] came up,
and if not, what the recommended practices had that it lacked.
Then open an incognito window and search "best dentist in [your city]." Note whether you're in the map pack, and if an AI Overview appears, expand its sources. Screenshot everything. The sources the assistants cite are the exact pages your next patient's answer will be built from.
Whatever gap showed up first is next month's work. Half-finished profile: finish every field this week. No reviews since spring: ask your next ten happy patients at checkout. No treatment pages: write the first one for the treatment you most want to book, answer-first, cost included. One path at a time, and the referral you get next month won't leak.