Get your med spa cited by ChatGPT and AI Overviews

What is AI search visibility, and how is it different from SEO?

AI search visibility is whether an assistant names your clinic inside the answer it writes. SEO earns you a place in a list of links that a patient then clicks. Here there is no list: you are quoted, or you are not in the answer at all. There is no position five to settle for.

So what gets optimised changes. In SEO you optimise a page to rank. Here you structure facts so a machine can restate them without mangling them. A clinic can rank perfectly well in Google and still have ChatGPT quote the wrong hours, the wrong service area or a price it stopped charging two years ago.

That last case is the one owners underestimate. Being invisible is a missed opportunity. Being confidently misquoted is an active leak.

Why does ChatGPT name a directory instead of your clinic?

Two reasons that need two different fixes. For comparison questions, your site was never a candidate. For questions about your clinic specifically, the directory states the fact in a plain sentence and your site does not.

Comparison questions were never yours to win. When someone asks for “the best med spa in Scottsdale”, the assistant needs a page that compares clinics. Your site cannot honestly do that, and it should not try. The pages that answer that question are directories, review aggregators, city listicles and Reddit threads. If you are absent from those, you are absent from the answer — no amount of on-site work changes it.

Questions about you were lost on structure. “Does [clinic] do consultations on Saturdays?” is a question only your site can answer. Assistants still fall back to a directory for it, because the directory states the hours in a sentence and your site puts them in a footer image, or states them differently on three pages, or hides them behind a booking widget that never renders for a crawler.

So the realistic goal is two-part: own every factual question about you, and be present in the third-party sources that answer the comparison questions. Anyone selling you only the first half is selling half a mechanism.

What does a quotable page look like?

One page, one question, answered in the opening forty words, in complete sentences that survive being cut out of their surroundings — and answered with facts only you can state: hours, price, duration, policy, who performs the treatment. The atmosphere paragraph goes below the answer, where it was always doing less work than you thought.

Then the boring half. Headings phrased the way patients ask, not the way marketers title things. Numbers with units. Prices as text, never as an image or a PDF. A provider page that names who performs each treatment and what licence they hold — for anything health-adjacent, provenance is weighted heavily and an anonymous site is a weak source.

Which crawlers have to reach your site?

Six, and they do not all do the same job. Googlebot, OAI-SearchBot and PerplexityBot decide whether you can be cited at all. ChatGPT-User decides whether a patient’s own prompt can reach your page in real time. Google-Extended and GPTBot are about model training, not citation — and those last two are the ones owners block by mistake, believing it removes them from AI answers.

Googlebot

Feeds Search and, from the same index, AI Overviews. There is no separate opt-in and no way to appear in Search while sitting out AI Overviews — the nosnippet controls that limit one limit the other.

OAI-SearchBot

Fetches pages so ChatGPT can cite and link them. Block this and you have opted out of being a source there. Distinct from the training crawler below.

PerplexityBot

Indexes for Perplexity's answers, which cite sources visibly — the easiest place to see whether your fixes landed.

ChatGPT-User

Visits a page in real time because a user's prompt sent it there. If this one is blocked, a patient asking about your clinic gets an error where your answer should be.

Google-Extended

A control for Gemini and grounding, not for AI Overviews. Blocking it does not remove you from AI Overviews and does not affect ranking. Widely misunderstood, including by agencies.

GPTBot

OpenAI's training crawler. Blocking it is a defensible decision about your content, and it is not the same decision as blocking OAI-SearchBot.

Three things break the fetch in the wild, in descending order of frequency: a staging robots.txt with a blanket Disallow shipped to production and never noticed; a bot-protection setting at your CDN or host, several of which now block AI crawlers by default without telling you; and treatment information that only exists after a JavaScript interaction, which is content a crawler never sees. Check all three before you write a single new page.

What should you change first?

Kill the contradictions between your site and your listings, rewrite each page to open with its answer, then publish the facts nobody else has. Those three come before schema, before new pages, and cost nothing but attention. The full order:

1. Kill the contradictions. Your site, Google Business Profile, Yelp, RealSelf, old directory entries and your booking page should state one set of hours, one address, one phone number and one service list. Assistants hedge or skip when sources disagree, and stale directory entries are the usual culprit.
2. Give every page one question. Open with the answer, in the first forty words, before any brand copy. Move the atmosphere paragraph below it. It was never doing the work you thought.
3. Publish the facts nobody else can. Price ranges, consultation length and cost, cancellation and deposit policy, financing, parking, who performs each treatment and their credentials. This is the material an assistant cannot invent — and if you do not publish it, someone else’s guess becomes the answer.
4. Mark it up. LocalBusiness or MedicalBusiness, Service, FAQPage, and a Person entry for each provider. Structured data does not make you quotable on its own, but it removes ambiguity about what your page is claiming.
5. Get the facts out of images. A price list as a JPEG, a menu inside a PDF, hours baked into a hero graphic: all invisible. Text, on the page, in the markup.
6. Only then publish new pages. New content on a contradictory foundation just adds a fourth version of your hours.

None of this touches patient records, and it should stay that way. The only patient data anywhere near this work is what you publish voluntarily — photos, reviews, testimonials — and that carries its own rules. Read before and after photo rules before a single new image goes up. Being quoted by an assistant does not soften a single FTC or state advertising requirement.

How do you know whether any of it worked?

You run the same ten questions across the same four assistants every month and write down what comes back. That is the method, and it is the method because there is no dashboard: not one major assistant reports how often it named your clinic.

0

The dashboards that promise otherwise are sampling. An “AI share of voice” tool runs a set of prompts on your behalf and extrapolates from the sample. Some do it carefully. None of them are measuring, and none can tell you what happened inside the answers they did not run.

What you genuinely can measure is thin but real. Referral sessions from chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com show up in your analytics like any other referrer — small numbers for a local clinic, worth counting anyway because they are the only hard evidence in the room.

Which leaves the manual test. It is unglamorous, it takes twenty minutes, and it is the only method that answers the question you actually have.

Two rules keep the test honest. Run it signed out or in a fresh session, because your own history skews what you get back. And record what was wrong, not just whether you appeared — a confident answer with your old pricing in it is a worse result than not being mentioned.

Can anyone guarantee you a spot in ChatGPT?

No. Not us, not anyone.

Answers shift by account, by location, by model version, and sometimes between two runs of the identical prompt an hour apart. There is no ranking to buy, no submission form, no ID to verify. A vendor who guarantees you a position in ChatGPT either has not looked closely or is counting on you not checking.

What can be committed to is the input side: that the facts an assistant finds about your clinic are correct, complete, consistent across every source, and readable by machine. That is engineering, and it is verifiable in an afternoon. Whether the model then names you is not in anyone’s control, and any honest description of this work says so out loud.

Treat the size of the prize the same way. Assistants are a growing slice of how people research treatments, and nobody publishes a trustworthy figure for how large that slice is in aesthetics specifically. It is not yet your main channel. It is cheap to fix, it compounds with the local search work you are already paying for, and the wrong answer is costing you patients today regardless of how the slice grows.

Is this worth it for a single-location clinic?

Yes, and most of it is a one-off cleanup rather than a permanent retainer. A single location has one set of hours, one address and one provider list to make consistent — the same job as a ten-location group without the multiplication. If you have a spare weekend and the list above, do it yourself. That is not a rhetorical offer.

Do something else first if your phone is not being answered. A patient the assistant sends to a clinic that rings out is a patient you paid attention to lose twice. Fix intake first — that is a different guide, and the AI front desk exists for exactly that — then come back to visibility.

If you would rather it were handled, we run it as a monthly retainer at the price in the box at the top of this page: month to month, 30 days notice. Fact layer, schema, the content that answers real questions, and the monthly test above logged and reported. The details are on AI visibility and content and the numbers are on pricing.

One thing to be straight about, since this page asks you to trust a name you have not heard. Oris is a team in Wroclaw, Poland, building patient acquisition systems for US clinics. We have no US clients yet, which is exactly why there is no case study on this page, no client logo row and no percentage next to our name. What we have instead is a method you can verify without us — run the ten questions on your own clinic tonight and see what comes back. If you would rather see it written up, the free clinic audit includes that test on your clinic, in writing, within two business days. No call required to receive it.