Health Tourism ChatGPT Visibility: Becoming a Source
How do clinics get cited in ChatGPT answers? What is actually controllable across the access, verifiability and third-party corroboration layers.
By Roozbeh Nazari · CEO
The first point of contact in health tourism is shifting. A prospective patient no longer only types into a search box; they describe their question to an assistant and get a composed answer back instead of ten blue links. Underneath that answer sit a handful of source links, and for clinics the real question becomes: how do you get into that short source list?
This article takes that question specifically to the clinic vertical. We covered the mechanics on the Google side in a separate article, and how to measure visibility at the prompt level in the GEO playbook. The focus here is different: the layers a clinic can genuinely control on the assistant side, and why health topics behave differently.
A frame up front: no work can guarantee that a brand will be cited in an assistant's answer. These systems produce different answers from session to session, source selection is probabilistic rather than deterministic, and providers keep changing the mechanism. What follows is not a guarantee. It is the set of work that raises the probability and is at least auditable.
Layer 1: access, meaning your content can be read
The most basic layer is the most frequently skipped. OpenAI's own documentation defines several crawlers with different functions: OAI-SearchBot is used to surface sites in ChatGPT's search features; GPTBot collects content for training foundation models; ChatGPT-User handles user-initiated actions.
That distinction has a practical consequence. A clinic that wants to opt out of training collection can block GPTBot while allowing OAI-SearchBot, preserving its visibility on the search side. A robots.txt rule that blocks all of them in one line also closes off the possibility of appearing in ChatGPT's search answers. The most common mistake we see in practice is exactly this: a well-intentioned broad blocking rule producing an unintended side effect.
Another practical point: these crawlers behave independently of one another. A clinic not wanting its content used in model training is an entirely reasonable preference, and that preference does not have to mean giving up visibility on the search side. Rather than collapsing the decision into a single "we are closed to AI" line, it works out better to read what each crawler does and decide separately for each. The same distinction applies to other providers; you need to read each provider's own documentation, because the names and the behaviours are not identical.
The rest of the access layer is the same as classic technical SEO: the page should return meaningful HTML server-side, critical text should not be rendered client-side only, redirect chains should be short, and language versions should point at each other correctly. If something is broken in this layer, no work in the other layers produces a result. We treat setting this side up as the first step of our technical SEO work.
Layer 2: verifiability, meaning the sentence can be quoted
Assistants do not use a single page whole when composing an answer; they take parts of the answer from different sources. So what gets cited is not the page but the passage. What makes a passage quotable in clinic content is structure far more than style.
A few things make a practical difference:
- Ask the question in the heading and answer it in the very first sentence below. A section that delays the answer by three paragraphs does not produce a quotable unit.
- Write variables rather than figures. "The price is X" versus "these are the variables that determine the price" — the second is both safer on the regulatory side and longer-lived information.
- Define things on the page. A page that states plainly what a procedure is, who it is not suitable for and how the process works gets quoted more often than a page that merely implies the same information.
- Cite your source. If a medical claim appears, the source it rests on should be visible on the page; otherwise the claim should either be removed or framed explicitly as an observation from clinical practice.
There is an additional reality in health topics: these systems behave more cautiously on medical questions and weight institutional, verifiable sources in their selection. Google's AI features documentation likewise describes original, useful, people-first content as the basic condition; it does not propose a separate "AI optimisation" technique.
Layer 3: corroboration, meaning someone other than you says it too
What a clinic writes on its own site is a weak signal on its own. The pattern we see repeatedly on the assistant side is this: a brand that also appears consistently in sources beyond its own site shows up more often in answers.
For clinics this corroboration layer means something concrete: the doctor's professional profiles being current and consistent, the clinic's name and address written identically everywhere, a verifiable presence in trade publications and association pages, and patient reviews collected genuinely and in line with platform rules.
None of these items is a quick fix and none of them is something you can buy. Bought corroboration is not corroboration; it is also a separate risk item on both the search and the advertising platforms. We describe how we build this layer in the clinic vertical on our clinics page.
Why the health vertical behaves differently
Where assistants summarise comfortably on a general topic, the tone of the answer shifts noticeably on a medical question: more caveat sentences, more "consult your doctor", and a drift towards institutional publications in source selection. That is both a disadvantage and an advantage for clinics.
The disadvantage is obvious: it is not easy for a commercial clinic page to carry the same weight as a health authority or a professional association page. The advantage is this: what stands out in this environment is not marketing language but verifiable information. To differentiate you do not need to speak more boldly; you need to write more clearly and more honestly. What you already have to do on the regulatory side and what works here point in the same direction.
In practice this means the clinic's strongest content asset is not the marketing page describing a procedure; it is the informational page explaining how the process works, who it is not suitable for, in which situations it should be postponed, and what determines the cost. Because those pages do not look like conversion pages, most clinics do not prioritise them — yet they are overwhelmingly the pages selected as sources on the assistant side.
Measurement: what is worth tracking
Looking at a rankings report to see the result of this work does not help, because there is no ranking. The only meaningful approach is to fix the questions in advance, repeat the same questions at regular intervals, and record whether the clinic appears in the answers and, if so, in what context.
We explained how to set up that measurement in detail in the GEO article. The one clinic-specific note to add here: write your question list in the sentences a prospective patient actually forms, not in procedure names. "Rhinoplasty prices" is a keyword; "does it make sense to go to Turkiye for a nose job, and what should I watch out for" is a real question, and the latter is what has a counterpart on the assistant side.
One final warning: some of what is sold in this space promises unmeasurable outcomes. An offer that guarantees being cited in an assistant's answer is making a commitment that cannot technically be kept. Our own AI search visibility work is likewise not about guaranteeing an outcome but about building a measurable process.