AI Engines · 5 min read · September 24, 2026 · Hafsa Tariq, gaash.ai
A dental practice, ChatGPT and Gemini: which crawlers reach the site, and what they find there
ChatGPT reaches a dental practice's site through OpenAI's crawlers: OAI-SearchBot for its search answers, ChatGPT-User when a person asks about a page. Google's AI Overviews and AI Mode read what Googlebot indexes, and the Google-Extended token decides whether Gemini may use that content for training and grounding. None of them can read the patient portal, the booking widget's slots or a new-patient form in a PDF, so the facts patients ask about have to be on public pages as text.
What each operator documents
| Token | What the operator documents | For the practice |
|---|---|---|
| OAI-SearchBot | Indexes pages for ChatGPT search; a site that disallows it is not shown in those answers, though it can still appear as a navigational link | Keep it allowed on the public pages |
| ChatGPT-User | Fetches a page when a user asks about it; OpenAI says robots.txt rules may not apply, because a person asked | Nothing public needs hiding from it; private data stays behind a login |
| GPTBot | Collects training data; disallowing it does not remove a site from ChatGPT search answers | A separate decision about training |
| Googlebot | Feeds Google Search, including AI Overviews and AI Mode | Keep it allowed; Google's AI results are built on its Search index |
| Google-Extended | Controls whether crawled content is used for Gemini training and grounding; it does not affect Search | Blocking it keeps the practice's pages out of Gemini's grounding too |
Sources: OpenAI's crawler page, Google's common crawlers and Google's page on AI features, which says a page needs to be indexed and eligible for a snippet to appear as a supporting link, with no special markup required. None of these pages says how an assistant chooses one practice over another, and this article does not claim to know.
What they find on a typical practice site
| Part of the site | What a crawler receives | What to do |
|---|---|---|
| Service, team and contact pages in HTML | The text, if nothing intercepts the request | Keep the facts patients ask about here |
| The patient portal | A login page | Nothing to change; say publicly what a patient needs before signing in |
| An online booking widget | Often an empty frame drawn by a script | Write beside it which visits can be booked online, how far ahead, and the phone number |
| New-patient forms and the plans list in PDFs | A separate file that may not be opened | Put the plans list and what to bring on an HTML page; keep the PDF as a printable copy |
| Emergency hours in a banner image | No text | Write the hours as text on the contact page |
| The whole site behind bot protection | A challenge page instead of the site | Ask whoever runs it to let the answer-side crawlers through by name |
The last row is the one nobody chooses. In our audit of 9,483 small business websites, 7.2% of the sites served our crawler a challenge page instead of the site, 3.0% delivered their content only through JavaScript, and 2.1% kept real content in a PDF. Those figures cover all 40 trades; a practice site with a booking widget and a forms page is exposed to the second and third.
What the public pages have to state
Once the crawlers can reach the pages, the pages have to carry the answers. A patient's question to ChatGPT or Gemini is usually a filter on a dentist, and each filter needs a sentence of the practice's own.
- Whether the practice is accepting new patients, with the date that was last true.
- The plans the practice bills, by the names on patients' cards, on an HTML page.
- Same-day and emergency arrangements, and the hours they apply, as text.
- A page for each treatment you want patients for, with who does it and a price range.
- Which visits can be booked online, beside the booking widget, and the number to call for the rest.
- One name, one address and one phone number, the same on every page and listing.
Each is a fact only the practice can supply, and none of them depends on anything the crawlers' operators have not documented. Treatment pages covers the fourth item in detail.
How to check, in a few minutes
- Run the crawler access check on the practice's domain and note any answer-side token that is turned away.
- Fetch the new-patients page as the search crawler would, following redirects, and read the status code: curl -sL -A "OAI-SearchBot" -o page.html -w "%{http_code}" https://www.example.com/new-patients. Then open page.html and confirm it is your page, not a challenge.
- View the source of the contact page with scripts switched off and search for the phone number, the address and the emergency hours.
- Search the source of every page you want in Google's AI results for noindex, which keeps a page out of the index and so out of those results.
- Ask ChatGPT, with search on, a question the new-patients page answers, and look for the page among the sources.
These checks settle whether the crawlers can read the practice. Whether the answers then name it is what AI visibility for dentists measures, and the facts the pages should carry are in the dental practices overview.
Common questions
Will blocking GPTBot keep our practice out of ChatGPT?
Not out of ChatGPT search. OpenAI's crawler page says disallowing GPTBot does not remove a site from ChatGPT search answers; disallowing OAI-SearchBot does. GPTBot is about training, and it is a separate decision.
Does blocking Google-Extended affect our Google listing or search results?
Google says it does not affect Search. It decides whether Google may use your crawled content for Gemini's training and grounding, so blocking it keeps your pages out of what Gemini answers from, which a practice that wants to be named may not intend.
Is patient information at risk from these crawlers?
Not if it is where it belongs, behind the portal's login, which no crawler passes. robots.txt is a public request, not a lock, so nothing about a patient should ever rely on it. Keep public pages to facts about the practice.
Do we need special markup or a file for Google's AI results?
Google's page on AI features says no: a page needs to be indexed and eligible for a snippet, with no additional requirements. Business markup that matches the page still describes the practice clearly to any parser.
Sources
What each token does is taken from its operator's documentation, linked where used. The reachability figures are this site's own and cover all 40 trades.
Sources: OpenAI, crawler documentation · Google, common crawlers · Google, AI features and your website · our audit of 9,483 small-business sites, with its method
Read on