What did we actually test?
Most writing about AI search visibility is theory. We wanted observed data, so we ran a small, repeatable test and wrote down exactly what the AI did. This is a pilot, not a national index, and we are deliberate below about what it does and does not prove.
What we asked, and where
We ran ten patient-style searches through Perplexity on June 3, 2026, phrased the way a real patient would type them: “best Botox in Miami,” “best plastic surgeon in Dallas,” “best lip filler in Scottsdale,” and so on. We chose three strong but different aesthetic markets – Miami, Dallas, and Scottsdale – and four query types – Botox, med spa, plastic surgeon, and lip filler. For every search we recorded which practices the AI named, the Google rating and review count it showed for each, and the websites it cited in its written answer.
Why we used Perplexity
Perplexity does live web retrieval on every query and shows its source list openly, which makes a test like this reproducible and verifiable. You can run the same searches and check our work. That transparency is also why it is a useful proxy: the 2026 Medical Aesthetics AI Visibility Index, which measured ChatGPT, Claude, Perplexity, and Google AI Overviews, noted that Perplexity “surfaces the broadest range of publishers” because of that visible source list. It shows you the plumbing the other engines hide.
What this study does not claim
One engine. Ten queries. One day. We did not test ChatGPT, Claude, or Google AI Overviews on these exact searches, and we only tested head local “best [treatment] in [city]” queries, not treatment-education or safety questions where the source mix is different. Local pack contents also change over time. Treat the numbers below as a verified snapshot of how one transparent engine behaved, not as a universal law of AI search. Where we move from what we observed to what we think it means, we say so plainly.
What did the AI name across all 10 searches?
The consistency was the surprising part. Across three cities and four treatment types, the answer was built the same way every single time.
Every query returned a local pack of eight practices
All ten searches produced a Google-style local pack and named eight practices each before the “see more places” cutoff. There was no case where the AI wrote a thoughtful essay weighing dozens of providers. It surfaced a ranked local list, then wrote a few sentences explaining the top picks. For a patient, that means roughly eight names exist and the rest of the market does not.
The AI told us its own criteria
On the Miami Botox search, Perplexity opened its answer with its reasoning, word for word: “Here are some of the strongest Botox options in Miami based on recent ratings and review volume.” That is not us guessing at a black box. The engine named its tiebreaker. Across all ten queries, the named practices carried Google ratings from 3.1 to 5.0, and all but one sat between 4.4 and 5.0. Review counts of the named practices ran from a few dozen to one Miami med spa with more than 2,200 reviews.
The same names kept reappearing
Within a market, the winners repeated across treatments. NuYou and 4Beauty showed up for both Botox and lip filler in Miami. Lemmon Avenue Plastic Surgery appeared in all three Dallas searches we ran. Young Again and Nude Aesthetics anchored both Scottsdale lists. A strong Google Business Profile did not just win one query. It won the whole category in that city.
The full query-by-query breakdown looked like this:
So what actually decides whether AI names a practice?
Here we move from what we observed to what we think it means. This section is interpretation, grounded in the data above and in broader 2026 reporting, not a guarantee.
Google Business Profile strength gets you into the pack
The single clearest pattern was that the named practices were the ones with strong Google ratings and meaningful review volume. The AI said as much. In practical terms, your Google Business Profile, with its star rating and review count, appears to be the gate you pass through to be considered at all for these searches. A practice with a 3.9 rating and 18 reviews was not in any of the lists we saw. This is not exotic AI work. It is the same review and profile hygiene that has mattered for local search for years, now feeding a new surface.
A readable, citeable website gets you quoted
Getting named in the list is one thing. Getting quoted is another. On the Miami Botox search, where we captured every source, all eight named practices were cited to their own websites. The AI did not just list them. It pulled descriptive detail from each practice’s site to explain why they were a good pick. A practice with a thin, slow, or unreadable website can still appear as a name pulled from Google data, but it hands the descriptive paragraph, the part that actually persuades the patient, to whoever has a site the AI can read and quote. Making your pages easy for an engine to read and lift is the heart of generative engine optimization for med spas and plastic surgeons.
What this means for on-page GEO copy
This is the honest nuance. For these specific head local queries, on-page content was not the gate. Reviews and a readable site were. That does not make content irrelevant. It means content earns its keep on the deeper, non-local queries this study did not cover: “is Botox safe while breastfeeding,” “Juvederm versus Restylane for lips,” “what to ask at a rhinoplasty consult.” On those questions the source mix shifts toward treatment pages, directories, and editorial, and answer-first writing is what gets lifted. The lesson is sequencing, not either-or. Fix the local gate first, then build the content authority that wins the rest.
We run a free audit across ChatGPT, Perplexity, Claude, and Google AI Overviews for your top treatments and your city, then send a written report in 48 hours with your named-appearance baseline and the first three fixes to make.
Get a Free AI Visibility Teardown →What should your practice do with this?
The data points to a clear order of operations. None of it is glamorous, and that is the point. The work that moves these searches is foundational, not clever.
Audit your Google Business Profile first
Before anything else, look at your Google Business Profile the way the AI does. What is your star rating, how many reviews do you have, and how recent are they? If your rating is below the mid 4s or your reviews are thin or old, that is the gate keeping you out of the pack, and it is fixable with a consistent review request process at checkout. In our test, recency and volume tracked closely with who got named.
Make your site readable to the AI
Open your top treatment page and ask whether a machine could read it and quote a useful sentence. If the first paragraph is “our caring team is passionate about your beauty journey,” there is nothing to lift. Replace it with direct, factual answers: what the treatment is, which products you use, how long results last, what it costs. Add clean medical schema so the engine can identify you as one clear entity. This is the structural work at the center of our GEO audit framework for aesthetic practice schema and treatment-page visibility.
Then optimize for the queries this study did not cover
Once the local gate is handled, the next tier of value is the educational and comparison queries patients ask before they ever search “best in city.” Those are won with answer-first content, a complete RealSelf profile, and editorial mentions. They take longer and compound. But they are wasted effort if your reviews and site fundamentals are not in place first, which is why we sequence them second.
Where does this study stop?
We would rather under-claim than oversell, so here is the boundary line drawn clearly.
We only tested one engine
This was Perplexity. ChatGPT, Claude, and Google AI Overviews weight sources differently and may name different practices for the same query. Published 2026 reporting suggests they lean even harder on third-party authority like RealSelf and editorial, which would change the picture. We did not measure that here.
We only tested head local queries
“Best [treatment] in [city]” is one query family. The conclusions about reviews and local packs likely do not transfer to treatment-education or safety questions, where content and directories carry more weight. Different question, different gate.
Local packs change
The exact practices named on June 3, 2026 will drift as reviews and rankings move. The pattern – review-ranked pack plus own-site citations – is the durable finding. The specific names are a snapshot.
Frequently asked questions
Does this mean on-page SEO and content do not matter for AI search?
No. It means that for one specific query family, “best [treatment] in [city],” Google Business Profile strength and a readable site mattered more than page copy in our test. Content is what wins the larger universe of educational and comparison queries that come earlier in the patient journey, and answer-first writing is consistently the format AI engines lift. The right read is that reviews and site fundamentals are the first gate, and content authority is the next.
Why did a practice with a 3.1 rating still get named?
One Miami plastic surgery result included a low-rated multi-location brand, which suggests the local pack is not a pure rating sort and that brand recognition or category breadth can pull a known name in. It was the clear exception. Every other named practice across all ten searches sat between 4.4 and 5.0, so the rule held and the outlier is worth noting rather than ignoring.
Can I reproduce this study myself?
Yes, and you should. Open Perplexity, run “best [your treatment] in [your city],” and note who gets named, what their Google ratings are, and which websites get cited in the answer. Repeat for your top three treatments. It takes about fifteen minutes and tells you immediately whether you are inside or outside your market’s AI shortlist.
How is this different from the brand-level AI visibility studies I have seen?
The 2026 Medical Aesthetics AI Visibility Index measured which product brands, like Botox and Juvederm, dominate AI citations, and found the top 15 brands hold roughly 62 percent of citation share. That is the manufacturer layer. Our study looked one level down, at which individual practices get named when a patient searches locally. Both matter. Ours is the one a practice owner can act on this week.
Is review volume or review recency more important?
Our data cannot fully separate the two, because the practices that win tend to have both. Perplexity’s own phrasing was “recent ratings and review volume,” naming recency first. Broader 2026 commentary points the same way: a steady flow of recent reviews appears to read as relevance more than a large pile of old ones. The practical move is a consistent, ongoing review request process rather than a one-time push.
If I fix my reviews and site, how fast could I appear?
Perplexity does live retrieval and tends to reflect website and profile changes within days to a couple of weeks. The other engines move on a days-to-weeks cycle, with Google AI Overviews usually slower because it depends on Google recrawling. We frame any timing as an expectation based on observed patterns, not a guarantee, since results vary by market and competition.
What to do this week
Run the test on yourself. Open Perplexity and search “best [your top treatment] in [your city].” Write down whether your practice is named, what your Google rating and review count are next to the practices that did get named, and whether your website is among the sources cited. Do it for your three highest-margin treatments. That fifteen-minute check tells you which gate, if any, is keeping you out.
If you are not named, the fix order is the same one the data points to: shore up your Google Business Profile rating and review recency, then make your top treatment pages readable and quotable for the engines. If you would rather have it done for you across all four major AI engines with a written diagnostic, that is the free AI Visibility Teardown we run for aesthetic practices. You can also read more about our GEO services for what the full program looks like.
Sources
Our own query data was collected directly from Perplexity on June 3, 2026 and is reproducible. Every external stat cites a primary or industry source. Open any link to verify.
