What ChatGPT is doing when a patient asks
Patients have stopped typing two words into a search box. They ask a whole question: where do I get Botox near me, how much does it cost, and is the person injecting me a nurse or a doctor. The assistant answers in one screen with a short paragraph and two to five clinics by name. There is no page two to be on.
To write that paragraph the assistant needs facts it can state without hedging: what you treat, where you are, what it costs, who performs the treatment, and whether other people say the visit went well. It collects those facts from websites, from business listings and from reviews. When it cannot find them for your clinic, it names a clinic where it can.
So this is not a beauty contest. It is a legibility contest. The clinic that is easiest to describe correctly gets described.
Readable, verified, trusted
Every AI answer about local businesses rests on three things. Miss one and you drop out of the answer even if the other two are strong.
Readable
One page per treatment, written in the words patients use. Price ranges in plain text. Injector credentials on the page and a medical director named. MedicalBusiness, Service and FAQPage structured data so a machine reads the same facts a patient does.
Verified
The same clinic name, address, phone, hours and treatment list on Google Business Profile, Bing Places, Apple Maps, Yelp, RealSelf and the Allergan Alle and Galderma Aspire provider locators. ChatGPT search runs on Bing, which is the listing almost no med spa has ever touched.
Trusted
Recent reviews that mention the treatment by name, before and after galleries labelled by treatment, and a medical director named on the site. Trust here means checkable, not famous.
A clinic with a gorgeous site and no listings is unverifiable. A clinic with perfect listings and a site that hides prices in an image is unreadable. Both get skipped.
Start with the ten questions, not with keywords
Write down the ten questions a patient in your city would actually ask an assistant before booking. They are not keywords. They are sentences, and they are treatment level.
Ours for every city follow the same ten shapes: best med spa in the city, where to get Botox there, how much Botox costs there, best place for lip filler, laser hair removal reviews, a med spa near a neighbourhood, a good clinic for first time Botox, best microneedling or Morpheus8, medical weight loss, and top rated aesthetics clinic. Every page you write should answer one of them completely.
Run all ten yourself in a fresh, logged out session before you change anything. Screenshot the answers and date them. That is your baseline, and without it you cannot tell later whether the work moved anything.
Fix the website first
The website is the only source in this list you fully control, and it is where the answer engine goes to confirm what a listing claims. Give every treatment its own page: Botox, dermal filler, laser hair removal, microneedling, medical weight loss. A single services page with five paragraphs cannot be quoted for any one of them.
On each page answer the five things patients ask before they book: what it costs as a range, whether it hurts, how much downtime there is, when results show and how long they last, and who is a candidate. Write it as text, not as a graphic. Assistants cannot read your price list if it is a JPEG.
Then add structured data. MedicalBusiness for the clinic, Service for each treatment, FAQPage for the questions you just answered. This is the step most clinics skip entirely.
Then make the listings match
Assistants cross check. If your Google Business Profile says one suite number and your site says another, or your Yelp page still lists a treatment you stopped offering two years ago, the safe move for the model is to name someone else.
Claim and complete the profiles in this order: Google Business Profile, Bing Places, Apple Maps, Yelp, RealSelf, then the Allergan Alle and Galderma Aspire provider locators. Make the name, address, suite, phone, hours, website URL and treatment list identical, character for character, on all of them.
Then get reviews that say the treatment
A five star average with 200 reviews that say nothing is weaker here than 40 reviews that say the word Botox. The assistant is matching a treatment question to evidence about that treatment.
Build a simple request flow: after a visit, the front desk sends one message with a direct link, and it asks the patient to say what she came in for and how the visit went. Never offer anything in exchange, never write the review for her, and never reveal in a public reply that someone is a patient.
How to tell whether it is working
Re run the same ten questions every 30 days in a fresh logged out session, on ChatGPT, Perplexity and Google AI Mode, and keep the dated screenshots in one folder. Count how many of the ten name your clinic. That single number is the scoreboard.
Listings and structured data changes tend to show up within weeks. Consistent naming across all ten questions usually takes 60 to 90 days, because reviews and pages have to accumulate before an assistant treats you as the obvious answer.