How to Get Your Med Spa Recommended by ChatGPT and AI Search

A patient in your city types “best med spa near me for lip filler” into ChatGPT and gets three clinic names with a sentence on each. If yours is one of them, you just got a warm lead who arrived pre-sold. If it is not, you did not lose a ranking. You were never in the running, and nothing on your analytics dashboard will tell you that.

This is the part of clinic marketing that changed in the last eighteen months and that most owners have not caught up with yet. It has a name now, generative engine optimization, and a lot of agencies are selling it with the same confidence they once sold “page one of Google.” Before you spend a dollar on it, it is worth understanding how these recommendations are actually built, because the mechanism decides what works.

We have data to work from. In September 2026, Altaway Solutions, an AI-visibility agency in Utah, asked four assistants, ChatGPT, Perplexity, Gemini and Google AI Mode, to recommend a med spa in Salt Lake City, four prompts each, sixteen answers in total, and logged every source each assistant cited. The full study is here. We wrote about what it means for patients in Should You Trust AI to Pick Your Med Spa?. This post is the owner’s side of it.

What the study found, read as a clinic owner

The sixteen answers drew on 30 distinct sources. The most-cited category was clinics’ own websites. After that came directories and review platforms: MedSpaScout was cited five times, Medical Spa Locator four, Google Maps four, a local Salt Lake City directory three, Yelp twice, ClinicCompass twice, and one other directory twice.

Three things in that list matter more than the rest.

Your own website is the primary source. When the assistant builds its sentence about a clinic, the facts in that sentence come from the clinic’s own pages more than from anywhere else. If your site does not state what you do, who does it, and roughly what it costs, the assistant either skips you or fills the gap from a directory.

No single source dominated. The study notes that no source appeared in even 40 percent of the answers. The assistants sample widely and inconsistently. That means there is no one directory to “win.” There is a set of places the assistant tends to read, and you want to be described accurately on as many of them as you reasonably can.

The engines disagreed with each other. Six clinics were recommended more than once across the sixteen answers, but the lists changed by engine and by prompt. The authors are careful to say the results “vary from run to run and by account history” and that the whole thing is “one dated snapshot.” Treat it as a window into the mechanism, not a league table.

For transparency: ClinicCompass was one of the cited sources, twice out of sixteen. Two other directories were cited more often. Nothing about how a clinic appears on this site, free or paid, has any bearing on what an assistant cites, and any directory that tells you otherwise is guessing.

How an AI assistant actually builds a clinic recommendation

It helps to drop the idea that ChatGPT “knows” your clinic. For a local question it does something closer to what a diligent assistant would do with a browser: it runs one or more web searches, opens a handful of the pages that come back, pulls the facts it can find, cross-checks them against each other, and writes a summary. Perplexity shows you this happening. ChatGPT and Gemini do it behind the curtain.

That has three consequences for a clinic.

First, you cannot rank in an AI answer directly. You get cited because the pages the assistant opened said good, specific things about you. The pages are the lever.

Second, the pages it opens are mostly pages that already rank in ordinary search for the same question. Google Maps, the big directories, the “best med spas in [city]” roundups, and your own site if it ranks for your name and services. Ordinary local SEO is the foundation of AI visibility, not a separate discipline.

Third, assistants prefer facts they can extract cleanly. “Botox from $12 per unit, performed by our nurse practitioner Sarah Lee, medical director Dr. Amit Patel” gets pulled into an answer. A homepage that says “luxury aesthetic experiences tailored to you” gives the model nothing to say about you, so it says nothing.

Everything below follows from those three points.

Seven things that get a clinic into AI answers

1. Make your website answer the questions the assistant is asking

The questions are always some version of: what do they do, who does it, what does it cost, where are they, and is there any reason to trust them. Put the answers in plain HTML text on pages the assistant can open.

• A services page per treatment, not a single “menu” image or a PDF. Each page names the treatment, the typical price or range, how long it takes, and who performs it.
• A team page with every provider’s name, credential (RN, NP, PA, MD), and what they focus on. Name the medical director. Assistants answer “who is a good injector for X” with names, and they can only use names that are written down.
• Address, hours and phone in text, matching your Google Business Profile exactly.
• A short FAQ on each service page in real question-and-answer form. The assistant is literally looking for question-shaped text.
• LocalBusiness or MedicalBusiness schema with your services and providers. It is not required, but it removes guesswork.

If your website is built as a slideshow of hero images with two sentences of text, this is the single change that will do the most. It is also plain old SEO, which is the point.

2. Be on the directories the assistants read, with a complete profile

The study’s list is a reasonable starting map: Google Maps, Yelp, the med spa directories, and whichever local roundup site covers your city. Claim the profile on each, fill every field, and make sure the treatments and prices match your website. An assistant that finds “Botox $10 per unit” on one page and “$14 per unit” on another has a conflict, and conflicts get you described vaguely or dropped.

Your free ClinicCompass profile is one of those pages. It lists your Google rating and review count, your treatment menu, your photos, and your booking link, and it is what an assistant reads if it opens our page for your city. Claiming it takes a few minutes and costs nothing. We will say it once more because it matters: paying for a Featured listing changes where you sit on our city page and gives you direct website and booking buttons. It does not and cannot change what an AI cites.

3. Treat your Google Business Profile as the source of record

Google Maps was cited four times in sixteen answers, and both Gemini and Google AI Mode lean on it more heavily than the study’s raw count suggests, because Google’s own listing data feeds its assistant. Get the primary category right (Medical Spa, not Beauty Salon), list every service with a price, keep hours current, post weekly, answer the Q&A section yourself before a stranger does, and reply to every review. ScaleHaven’s guide to the med spa Google Business Profile walks through each field and is the most complete checklist we know of.

4. Build review volume that says what you want the assistant to say

When an assistant writes “known for natural-looking lip filler,” that phrase came from somewhere, and it was almost always review text. Assistants summarize reviews, so the content of your reviews becomes the content of your recommendation.

Two changes. First, volume and recency: a 4.9 from 40 reviews is a thinner signal than a 4.8 from 400, both to a patient and to a model deciding how much weight your clinic deserves. Ask every happy patient, every visit, with a link. Second, specificity: when you ask, ask them to mention the treatment and the provider by name. “Sarah did my lips and they look natural” is worth ten “great place, friendly staff” reviews, because it is the sentence the assistant is going to reuse.

5. Get onto the lists the assistants love

Roundups are catnip for AI. A page titled “Best med spas in Denver” is exactly the shape of the question being asked, so it gets opened first and cited often. The local directory in the study was cited three times for that reason.

Find every “best med spas in [your city]” page that ranks in the top ten and check whether you are on it. Where you are not, ask. Local lifestyle magazines, city blogs, and directory roundups all have an editor or a submission form. Our own ranked city lists are built from public Google rating and review data, so the way onto those is the review work in point four, not a request.

6. Put your providers in front of the camera and give them their own presence

This is the one most clinic owners resist and the one with the biggest gap between the clinics that do it and the clinics that do not.

Patients increasingly ask assistants about people, not businesses: “best nurse injector for a lip flip in Austin.” The assistant can only answer with names that exist on the web with evidence attached. A provider who has a named page on your site, a public Instagram where she shows her work and explains her approach, and reviews that mention her by name is a provider an assistant can recommend. A provider who exists only as “our expert team” is invisible.

The practical version, which we have seen work in clinics we have run and worked with: each injector and provider runs their own professional Instagram, posts once on every day they work, and that expectation is written into their contract from day one along with who owns the account and what happens to it if they leave. Educational content, before-and-after with consent, a face and a voice explaining a treatment. It builds the provider, it builds the clinic, and it produces exactly the kind of named, specific, cross-referenceable evidence assistants look for. It also happens to be the best-performing raw material for paid ads, which is a separate post.

7. Say the same thing everywhere

Assistants cross-check. If your website says you offer Morpheus8, your Google profile does not list it, and a directory says you closed in 2024, the model resolves the conflict by hedging or by leaving you out. Once a quarter, open your website, your Google profile, and your top five directory listings side by side and reconcile the name, address, phone, hours, services, prices and providers. It is dull work. It is also the difference between “a med spa offering Morpheus8, Botox and fillers led by Dr. Patel” and “a med spa in the area.”

How to measure whether it is working

You will not find AI referrals in a neat report, so build a small one yourself.

• Monthly prompt audit. Once a month, ask ChatGPT, Perplexity, Gemini and Google AI Mode the same four questions a patient in your city would ask (“best med spa in [city],” “best place for Botox in [city],” “best nurse injector for lips in [city],” “med spa in [city] with good reviews”). Log which clinics are named and which sources are cited. Sixteen answers, ten minutes. Watch the trend, not any single run, because the same prompt can return a different list tomorrow.
• Ask patients. Add “an AI assistant (ChatGPT, etc.)” as an option on your “how did you hear about us” question, on the booking form and at the front desk. Within three months you will know whether this channel is real for your clinic.
• Watch branded search. Patients who get your name from an assistant usually Google you next. A rise in branded searches and direct visits to your site, with no matching ad spend, is often this channel showing up sideways.

Set expectations accordingly. The changes above take weeks to appear in search indexes and months to work through into assistant answers, and the answers are noisy. Anyone promising you a spot in ChatGPT by the end of the month is describing something that does not exist.

What not to spend money on

• Paid placement in AI answers. As of this writing there is no ad slot inside a ChatGPT, Perplexity or Gemini recommendation for a local clinic. Anyone selling one is selling something else.
• “AI SEO” retainers that skip the basics. If the proposal does not start with your website copy, your Google profile and your reviews, it is skipping the three things the study says matter most.
• Fake or incentivized reviews. Assistants summarize review text, and platforms are getting better at removing the fake ones. You would be building your AI reputation on sand.
• Keyword-stuffed pages written for robots. The assistant is reading for facts to reuse in a sentence to a human. Write the sentence you want it to reuse.

Where this leaves you

AI recommendations are not a new channel you have to buy into. They are a new reader of the channels you already have. The clinic that describes itself clearly on its own site, keeps its Google profile and directory listings complete and consistent, earns specific reviews, gets named on local lists, and puts real providers with real names in public will show up in AI answers over time, because those are the pages the answers are made from.

Start with your website and your Google profile this week. Then claim your free ClinicCompass listing and make sure it matches. If you would rather have someone run the whole visibility system, ads, search and all, that is what ScaleHaven does for aesthetic clinics, and the same principles apply there: named offers, real faces, and facts a patient or an assistant can act on.