Short answer: more Google reviews did not make a med spa more likely to appear in AI answers in our test. Better information on the clinic's own website did.
We analyzed 174 med spas in September 2026, and clinics whose websites included more concrete facts were named in AI answers much more often: 38% versus 21% for clinics with sparse information.
Review count and review content, by contrast, showed no reliable relationship with whether a clinic was named once we accounted for other factors.
That does not mean reviews are unimportant. They still help with trust, conversion, and how a business looks in search. But our data does not support the idea that simply getting more reviews is a reliable way to appear in AI recommendations.
OpenAI and Google also do not say that Google review volume is a direct ranking factor for these AI results. Below, we show what the data does and does not tell us, plus one review practice that can put your Google Business Profile at risk.
Where this question started: the Scottsdale surprise
We did not start with that answer. We started with a small test that pointed the other way.
In our Scottsdale field test, a 25-clinic sample (August 2026), the clinics an AI named had far more Google reviews: a median of about 558 versus 245, and the gap held in a second, stricter run (567 versus 244). Star ratings were nearly equal (5.0 versus 4.9). Review volume looked like it mattered a lot.
But it was one small sample, and even there the signal was shaky. The two clinics with the most reviews (around 1,185 and 1,050) were not named at all, and most of the picks traced back to the same few directory pages the model cited. Volume tracked naming without deciding it.
A 25-clinic sample from one city cannot carry a general claim, so we ran a larger test to see whether the review-count signal held up. It did not.
Count, content, or the page?
We expanded the test to 174 clinics across several cities (September 2026). For each one, we checked whether an AI search model actually named it when asked for local med-spa recommendations. 48 of the 174 clinics were named. Then we looked at what set the named clinics apart: how many Google reviews they had, what those reviews talked about, or the facts on their own website.
We double-checked the result with a statistical model that weighs all three at once and accounts for the city. Page facts were the only one that still predicted whether a clinic got named. Review count and review content dropped out. (The full model, with the exact numbers, is in the methodology below.)
The plain takeaway: what most tracked with getting named by AI was not the reviews other people wrote about the clinic. It was the concrete, checkable facts the clinic put on its own pages, the one input a clinic fully controls. This is a pattern from one dated test, not a guarantee. Here are the three comparisons, side by side:
Review count did not separate the groups
The clean, near 2.3x median gap Scottsdale showed did not reproduce across 174 clinics. There was no consistent pattern where more reviews meant more naming.
Review content did not separate them either
We scanned the review text and scored how often reviews named specific treatments and outcomes. Treatment-specific content did not distinguish named clinics from not-named ones.
The clinic's own page facts did
Grouped by a 0 to 5 count of concrete decision facts on the clinic's own page, clinics with more facts were named far more often than those with few.
Controlling for review count, review content, and city, only page facts stayed statistically significant.
Does AI search read med spa reviews at all?
Yes, AI can pick up review-related information. That is not the same as reviews deciding who gets named.
Here is what the companies actually say. OpenAI's help page for ChatGPT search says ChatGPT can use your location and trusted third-party providers, such as nearby business listings, to find local results, and that "placement is not guaranteed." It does not say Google review count or review text is a ranking input. So treating reviews as a direct input is a guess, not a stated fact.
Google's AI-features documentation says a page becomes eligible for AI Overviews through normal Search indexing, with no special markup or payment. Google does recommend keeping your Business Profile current. Separately, Google's local-search guidance says more reviews and better ratings can help your local ranking. But that is the local map system, not AI Overviews, so "reviews drive AI citations" goes further than Google states.
In our Scottsdale runs, the model backed almost every pick with a star rating, a review count, and a directory. That is the wording of its answers, not a confirmed ingredient of its choice. In the same runs, most picks traced back to the same directory pages, and some high-review clinics were left out.
Does review count matter more than review content?
For the Google map pack, the wording of a review can add a visible detail. For getting named by AI in our test, neither the count nor the wording did the work.
When a patient writes "I got Botox here and it looked completely natural," Google can show that phrase as a highlighted snippet under your listing, something like "customers mention Botox." BrightLocal documents these snippets, pulled straight from your reviews. But a controlled Sterling Sky test found that adding keywords to reviews did not change map-pack rankings, so it is a display effect, not a ranking lever.
One thing you cannot do: ask a patient to name a specific treatment in their review. Google's review policy prohibits requesting specific content. Encourage honest, detailed reviews, and treatment names show up on their own. Just do not build your AI plan on that text: in our 174-clinic test, review wording did not separate the named clinics from the rest.
Why reviews are a poor substitute for publishing your own prices
A review that says "had fillers done, ended up around $650, really happy" names a procedure and appears to give a price. But a review is a weak source of record for price: the number may be old, may include a discount or a membership rate, may cover several syringes, or may belong to a different year entirely. If you do not publish your own price, an AI tool or a patient may pick up whichever figure sits in a review or a directory, and you hand control of that fact to third parties.
From our price transparency study, roughly 55 to 59 percent of med spas (180+ clinics, August 2026) show no procedure price as plain text on any page we checked. That gap is exactly where reviews and third-party sources step in by default. This connects straight back to the page-facts finding above: a plain-text price is one of the concrete facts a clinic controls, and publishing it makes your own page the first source of record instead of a stranger's review.
What you absolutely cannot do with reviews
This section is not my opinion. It is Google policy and federal law.
Review gating is asking only satisfied patients for a public review, or routing unhappy ones to a private form before they reach Google. Google's review policy explicitly prohibits discouraging negative reviews or selectively soliciting positive ones. (Google uses "selectively solicit positive reviews," not the word "gating," which is an industry term for this practice.)
The platform consequences include review removal, a temporary block on new reviews, unpublishing of existing reviews, or a consumer warning on your profile. In one documented case, Sterling Sky reported a business dropping from 464 reviews to 78 (about 83 percent) after gating was detected.
The FTC layer is separate. The Trade Regulation Rule on the Use of Consumer Reviews and Testimonials took effect October 21, 2024, and covers fake reviews, paid reviews, compensation tied to review sentiment, and suppressing negative ones. Civil penalties for knowing violations run up to $53,088 per violation for covered conduct.
Send the same review request to every patient, regardless of how you expect them to feel about the visit.
How to check what AI says about your clinic
Ask an AI search model with web search on: "Is [your clinic name] good for Botox?" Does it cite a rating or review count? Does the answer name specific treatments? Does it get something wrong? AI systems can surface inaccurate information when the public data on a clinic is thin or inconsistent. Logging what AI says about you once a month takes five minutes and gives you a baseline. More on reading AI recommendations is at getting recommended by ChatGPT.
What to do with reviews this week
No tactics that promise AI placement. Just the practices that hold up when AI systems read your public profile, plus the one that our data pointed to: get your own page facts right.
Fix the concrete facts on your own treatment pages first. In our 174-clinic test (September 2026), page-facts count was the only input that predicted AI naming after controls. This is the input you fully control.
Ask everyone, the same way. A short follow-up text or email after each visit, sent to every patient regardless of how you think the appointment went. Google supports sharing your review link or QR code in follow-up messages.
Do not coach the content. Ask for a review, not for a specific wording or a treatment name. Detailed, honest reviews produce justification snippets on their own.
Reply to reviews, including critical ones. In our spot-check of 40 med spa profiles (Miami and Atlanta, August 2026), 13 of 40 (32 percent) had no owner reply anywhere in their 12 most recent reviews, and 23 of 40 (58 percent) had replied within the prior 90 days. Google frames replies as a way to show patients you value their feedback, though it does not list replies as a local ranking factor.
Build reviews as a normal part of follow-up. Rather than running artificial review bursts, keep a steady, genuine flow; Google watches for unusual review volumes and patterns in the context of manipulation. We have not measured that any particular cadence helps ranking or AI visibility.
No incentives. Google prohibits incentivizing reviews. No discount, no gift card, no offer attached to a review ask.
Method and limits
Sample
174 med spa clinics across several US metros, September 2026. The earlier Scottsdale run was a separate 25-clinic convenience sample from August 2026, used here only as an opening counter-example.
Definitions
A clinic counted as AI-named if an AI search model with live web access named it in local med-spa prompts across three separate passes, with a stability threshold of at least two of three; 48 of 174 met it. We compared three inputs: Google review count, review-content specificity (how often review text named specific treatments and outcomes), and a 0 to 5 count of concrete decision facts on the clinic's own pages. The page-facts count uses a frozen 5-factor set on the clinic's own treatment page: procedure, city or metro, candidacy or indications, price or scope, and a named provider with credential.
Classifier validation
We scored review-content specificity across 7,389 review texts with a keyword classifier and validated it. A blind LLM re-labeling of a random 500-review sample agreed 81% of the time, with 96% precision and 63% recall. The classifier is conservative: it flags 31% of reviews as treatment-specific versus the LLM's 48%, so it under-counts rather than over-claims. We repeated the comparison using the LLM labels, and the result stayed the same: more treatment-specific review content still did not separate the clinics AI named from the ones it skipped.
Results
A logistic regression modeling AI naming against review count, review content, and page-facts count, with city controlled, left only page-facts count significant (OR 2.05, 95% CI 1.39 to 3.03, p=0.0003). Review count (p=0.13) and review content (p=0.37) were not significant. The 38% versus 21% page-facts headline reproduces from a single anonymized file pairing each clinic's page-facts count with its named or not-named outcome.
Limitations
One AI model, one window (September 2026), and a limited set of cities. AI systems change frequently, association is not causation, and the model's own selection process is not published. Treat these as dated observations, not rules.
Provenance and usage
The row-level dataset is retained by MedSpaCopy, versioned with timestamps, and available for verification or research inquiries. Copyright MedSpaCopy 2026. Methodology, analysis, classifications, and database compilation are proprietary; aggregate findings may be quoted with attribution and a link to this study; redistribution of the underlying dataset is not permitted without written permission.