The setup
Two complete panels in two Sun-Belt metros, run on the same protocol as our family-law field study: a frozen set of real customer questions (“best med spa near me,” “best place for Botox,” “who does the best lip filler,” device-specific questions, “affordable” questions) across ChatGPT, Gemini, Perplexity, Claude, Google AI Overviews, and Google AI Mode — the highest-intent questions run three times each on the automated engines and twice on the chat apps, the supporting questions once. 200 scored observations per metro, 400 in total, every run logged and archived. All firms below are anonymized.
The headline is the same one every panel produces: the subject business — well-reviewed, established, genuinely good — was praised whenever a customer asked about it by name, and almost never present when a customer asked who to go to. That gap between reputation and visibility is the product of four levers the winners carry and the invisible don’t.
Lever one: the engines recommend a person, not a clinic
The single most consistent pattern across both metros: “people follow the injector, not the clinic.” One engine said nearly exactly that in a live run, recommending practices where “the strongest reviews have the same named injector coming up over and over.” Eponymous practices — the clinic named after the person doing the injecting — over-indexed everywhere. On device questions, every top recommendation in one metro was a named physician with a stated specialty, not a brand.
The catch: in one metro we audited, nearly half the spas that do have credentialed injectors name them only on Instagram or a booking app — off-site, where the recommendation engines that read websites never see them. The person exists; the page doesn’t say so. That’s an afternoon of website work worth more than a year of generic content.
Lever two: the published price sheet wins the money questions
On “affordable” and cost-flavored questions, the winners were the businesses that publish actual numbers. In one run, the top pick was chosen explicitly because new-client Botox “is advertised at $10 per unit” — the engine did the math for a typical treatment in the answer. Another clinic earned its slot two runs out of two “because they publish standardized rates.” A published tier card ($200 / $250 / $350) that had merely earned a mention one week was quoted as the number-one selection reason the next.
Nothing about this is subtle: engines quote price sheets because a price is a fact they can lift whole. “Competitive pricing” is unquotable. “$12 per unit, 20-unit minimum” is a sentence an engine names you for.
Lever three: name the machine
On device questions — laser hair removal, RF microneedling, branded facials — the engines displayed real device-platform literacy: recommendations cited the specific machine (“uses the newer [platform],” “the gold standard for darker skin tones”) and warned against clinics that don’t say what they run. A clinic that names its device models, on indexed pages, enters answers that a clinic with a generic “laser treatments” page cannot.
Lever four — the new one: the engines run verification checks
This is the finding that surprised us most, and it sharpened during the panel itself. On device and injectable questions, the reasoning-tier engines didn’t just rank — they verified:
- The state license registry. Top picks were justified with “has a clear, active license” — sourced to the state’s official license-lookup portal — and in one run all three recommended injectors were named with their individual license status printed in the answer.
- The manufacturer’s provider directory. On a branded-facial question, the engine told the user to “confirm the provider appears in the official provider directory” and cited the device maker’s own site — twice, in two independent runs.
- The regulator’s warning. On RF microneedling, the answer cited the FDA’s safety communication as the reason to choose a credentialed provider.
Put together, that’s a four-layer trust stack — state license, board credential, regulator, manufacturer — standing between a med spa and a recommendation. A spa whose named people, licenses, and devices are independently checkable passes through it. A spa marketing “luxury treatments” with an anonymous team does not, no matter how good its reviews are.
What reviews actually do (less than you think)
Review scores mattered — but not the way most owners assume. Engines stated flatly that ratings are “close to worthless” at five-star saturation and ranked instead on review text: who is named in the reviews, which treatments come up, what the recurring complaints are. A ~190-review solo operator beat 1,000-review generalists on the treatment she’s named for. Sub-4.5 ratings were excluded outright in multiple runs. And review-driven visibility proved the most fragile kind: it can evaporate when measured from a clean, logged-out vantage, while published on-page facts survive.
What to do with this
If you run a med spa, the order of operations from these panels is unusually clear: put your named, credentialed injectors on your website; publish your prices; name your machines; make sure your license, your people’s licenses, and your manufacturer-directory listings are current and findable. Then measure — per engine, because the engines don’t agree with each other — and hold the positions, because nothing holds unassisted. That measurement, for your market, is what a Nameworthy panel produces.