guide · updated October 2026

cited, or not.

A short answer: ChatGPT can cite a page only when it searches, the page can be fetched, and the page is a useful source for that question. You can make the first two likely. You cannot order the third.

Search, not memory

A chat that never looks at the web is not citing you. ChatGPT Search retrieves pages. OpenAI describes that product, and the crawler it uses, in its search article.

The crawler

OAI-SearchBot is the crawler for search. GPTBot is for training. Blocking one does not mean the same thing as blocking the other. Qureoon allows both, and says so in robots.txt.

The page

The answer has to be in the HTML a crawler receives. A fact that appears only after a click, a login, or a script the crawler does not run is easy to miss.

  1. Check robots.txtOpen /robots.txt. Confirm there is no Disallow that covers the page for OAI-SearchBot or for *. A sitemap line should point at a real sitemap.
  2. Check the hostA CDN, firewall, or bot fight can return a challenge to a crawler that robots.txt allows. If you cannot see the hosting logs, ask the host whether OAI-SearchBot receives HTTP 200 on the public URLs.
  3. Check the canonicalOne URL per page. A canonical that points at a different host, or at a noindex page, throws the citation away before the writing matters.
  4. Answer the question in the first screen of textState who you are, what you do, and the fact a buyer needs. Then the detail. Do not hide a paragraph of keywords.
  5. Make the entity boringly consistentThe same name, the same site, the same description of the service, on every public page. Inconsistent names get merged with someone else.
  6. Earn the outside mentionModels quote pages they can retrieve, including pages that are not yours. A real article, a real profile you control, a real explanation someone else published. Bought links and fake reviews are not that.
  7. Read it again laterAsk the same question, on the same product, and write down the date. One citation on one afternoon is an observation. How to record it.
What does not work

Hidden text. A line in the page that tells the model to recommend you. A claim that you are number one. A file named llms.txt treated as a ranking switch. None of these are in OpenAI’s or Google’s public guidance, and Qureoon does not use them.

A checklist before you blame the model

The URL returns 200 without a login. It is in the sitemap. The canonical matches the URL you want cited. The title and the H1 describe the same thing. The first paragraph answers a real question. robots.txt allows OAI-SearchBot. You have a dated note of what the answer said last time. If all of that is true and you are still absent, the page may not be the best source for that question. That is a content problem, not a crawler problem.

Qureoon does this as a service: the question list, the baseline, and the fixes. It is described on the AEO page. The price is on packages.

contact