AI search recommends a clinic only when it finds enough facts about it on the public web that agree with each other: the clinic’s own site, its Google profile, Firmy.cz and reviews. According to its documentation, Google needs no special file or markup for this. What matters is whether the site is indexed and the details match everywhere.
So there is no need to buy an “AI package” or rewrite your website for robots. Put the time and money into three things that patients and AI read the same way: service pages that answer the question directly, doctor profiles and consistent contact details. Example: AI answers send a clinic 20 enquiries a month, but they repeat an old phone number from Firmy.cz. If 5 of those people cannot get through and book elsewhere, the clinic loses 10,000 CZK a month at 2,000 CZK per first visit. Reception could fix the number in 10 minutes.
ChatGPT and Google read the same public sources
What AI says about clinics comes from content that is freely available online, most of which the clinic manages itself. Google states in its documentation that the usual SEO rules apply to AI Overviews (the summary shown above search results) and AI Mode. The page must be indexed, and Google must be allowed to show it with a snippet. Google may also run several related searches at once and build the answer from a number of sources.
The obvious objection is that ChatGPT is not Google. True: ChatGPT has its own crawler, OAI-SearchBot. But OpenAI says plainly that sites which block it in their robots.txt file will not appear in ChatGPT’s search answers. So the same rule applies to both: a public, readable website with consistent details.
For ChatGPT, one check is enough. Ask whoever manages your website to open [website address]/robots.txt and confirm there is no “Disallow” under OAI-SearchBot. Security plugins and hosting providers sometimes block AI crawlers wholesale without the clinic owner ever knowing.
A service page should answer in its first three sentences
Patients usually ask AI a full question, such as “how much does teeth whitening cost in Prague 5 and who does it”. The best answer comes from the page where AI finds the facts right at the top. Yet clinic websites often open with a paragraph about their modern approach, while the price list sits in a PDF on another page.
Start every service page with four facts: who the treatment is for, what it costs, how long it takes and who performs it. Here is a template to fill in:
[Treatment name] at [clinic name] is for [who it is for]. The price is [price from] to [price to] CZK, depending on the scope the doctor sets at the first examination. The treatment takes [length] minutes and is performed by [doctor’s title, name and specialty]. You can book on [phone number] or online at [link].
Write prices as text, not as an image. Both a crawler and a patient on a phone can read text. Below the opening, add the 3 to 5 questions reception hears most often about that service, each with a short answer.
AI may shorten whatever you write on the website and quote it out of context. So do not promise treatment results. The Czech advertising act (Act No. 40/1995 Coll.) forbids this in § 2(1)(b), as it treats such promises as an unfair commercial practice.
Publish a specific patient’s review or story on your website only with their consent. The fact that someone is your patient is protected by confidentiality (§ 51 of Act No. 372/2011 Coll., for doctors also § 2(9) of the ČLK Code of Ethics) and by Article 9 of the GDPR, which covers health data. We sum up the whole framework in our article on the ČLK Code of Ethics and advertising.
Doctor profiles and structured data add the facts AI looks for
Every doctor needs a page or a block of text of their own: title, name, specialty, board certifications, languages and the treatments they perform. AI can only answer “dentist in Brno who speaks English” if that is written down somewhere. A photo captioned “our team” won’t tell it.
Structured data is markup in the page code, following schema.org. It tells machines which part of the page is the address, the phone number or the opening hours. Google requires no special markup for AI, but it does point out that structured data must match the visible text. Your web developer can add the MedicalClinic or Dentist type with address, phone number and opening hours in an hour or two. Check the result in Google’s Rich Results Test.
Consistent contact details in five places do more than new copy
When sources contradict each other, AI may pick up the outdated detail. So make the details match everywhere the clinic appears:
| Place | What to check | Who does it |
|---|---|---|
| Website (contact page and footer) | phone, address, hours, name | website manager |
| Google Business Profile | the same, plus categories and services | reception or owner |
| Firmy.cz | phone, hours, service descriptions | reception |
| Mapy.com | entrance location, hours | reception |
| Facebook and Instagram | phone and booking link | social media manager |
Reviews belong on this list too. Google says of local results that more reviews and positive ratings can improve a business’s ranking. Our article How to get patient reviews explains how to ask for them without breaking the rules.
Our free audit shows whether your details are out of step and how your reviews compare with three competitors nearby. We go through your profiles on Firmy.cz, Google and Mapy.com and email you a PDF within one working day.
You can find out what AI says about you in 20 minutes
Ask ChatGPT, Perplexity and Google’s AI Mode the same 5 questions a patient would: “good dental hygiene in [district]”, “how much does [treatment] cost in [city]”, “[clinic name] reviews”. For each answer, note three things: whether it mentions the clinic, whether the price, phone number and hours are right, and which source it cites. Ask every question twice, because answers vary from one attempt to the next.
A wrong detail in an answer points you to the source that has it wrong. Fix it there, not in the chat. Repeat the test once a month and compare the tables.
The second measure is how often people search for the clinic’s name, which Google Search Console shows. On the Angel Smile project we used it to track brand awareness: during Google Ads and Meta Ads campaigns, searches for the clinic’s name rose from 250 to 480. The same figure will show whether patients are also hearing about you from AI answers. Our article How to measure clinic marketing explains how to link it to booked patients.
This approach has limits. If three competitors nearby have 300 reviews each and you have 20, consistent details alone won’t change the ranking. And nobody can promise you a place in an AI answer, because the answer depends on how the question is worded and where the patient is asking from.
Frequently asked questions
Does a clinic need an llms.txt file?
Not for Google. Its documentation says AI Overviews need no new machine-readable files. If your web developer can add one in a few minutes, it does no harm. Just don’t pay for it as a separate job.
How quickly does a fix show up?
According to OpenAI, ChatGPT picks up a change to robots.txt in about 24 hours. A corrected phone number or price only appears in answers once the crawlers have read the page again. That usually takes days to weeks, so repeat the test after a month.
Is it worth paying for a fixed place in ChatGPT’s answers?
Google says there are no special techniques for AI Overviews. Anyone promising a specific place in an answer is promising something they cannot control. Spend the money on rewriting your 5 most visited service pages instead.
Four steps are enough this week
On Monday, have reception write down the 5 questions patients ask most often. On Tuesday, put those questions to ChatGPT, Perplexity and Google and note every mistake. On Wednesday, ask your website manager to check robots.txt and turn prices in images and PDFs into text. On Friday, align the phone number, address and hours in all five places from the table.
