Skip to content
HowFindableScan
Back to blogGuide

How ChatGPT decides which sites to cite

September 8, 2026·4 min read

When ChatGPT answers a question with links, it is not remembering your site. It runs a search while the person waits, reads whatever pages it can reach in those few seconds, and quotes the ones that gave it a usable answer. Your site is either readable at that moment or it sits the round out.

What decides that is a plain text file at the root of your domain called robots.txt, which tells crawlers what they may read. OpenAI documents four crawlers with very similar names, and getting them mixed up is the difference between being quoted and being invisible.

OpenAI runs two separate crawlers and only one controls ChatGPT citationsBlocking GPTBot only opts you out of model training. Blocking OAI-SearchBot removes you from ChatGPT search answers.Two crawlers, two different doorsGPTBotTrains future modelsBlockedYou still appear inChatGPT answersOAI-SearchBotSurfaces you in ChatGPT searchBlockedYou are gone fromChatGPT answersOne line in robots.txt is the whole difference
Two OpenAI crawlers, only one gates citations

Four crawlers, one that decides

OpenAI publishes what each of its crawlers does. GPTBot collects content that may train future models. OAI-SearchBot, in OpenAI's words, is used to surface websites in search results in ChatGPT's search features. ChatGPT-User fetches a single page when someone asks ChatGPT to open it. OAI-AdsBot checks pages submitted as ads.

Only one of those four gates your citations, and OpenAI says so directly: sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers. Everything else on this page follows from that one sentence.

Blocking GPTBot does not hide you from ChatGPT

This is the mistake we see most often, and it usually starts with good intentions. Someone reads an article about keeping their writing out of AI training, copies a block of robots.txt rules into their site, and the block bans anything with an AI-sounding name. GPTBot and OAI-SearchBot go out together.

Those two choices have nothing to do with each other. Refusing to feed model training is a legitimate call about your own work, and it costs you no visibility today. Refusing OAI-SearchBot removes you from the answers people read, which is the traffic you were hoping for in the first place. You can block the first and welcome the second.

Being readable is the ticket, not the prize

Once ChatGPT can read you, you are competing with every other page it found. Most pages that get cited in AI answers already rank in ordinary search, so the classic work still carries you: a clear title, a page that loads its content in HTML rather than assembling it with JavaScript, and a topic you actually cover in depth.

The last step is where most sites lose it. ChatGPT lifts short passages, not whole pages. If your answer sits in paragraph nine, after the story of how the company was founded, the model reads the top of the page, finds no answer, and quotes the competitor who put theirs in the first two sentences.

Check yours in two minutes

You do not need a tool for the first look. Type your domain followed by /robots.txt into your browser and read what comes back.

If nothing loads, you have no robots.txt, and that means nothing is blocked. That is a fine place to be. If a file does load, three things are worth finding:

  • OAI-SearchBot with a Disallow line under it. That is the one closing the door on ChatGPT.
  • PerplexityBot, same story, for the other big answer engine that cites live sources.
  • User-agent: * followed by Disallow: / which blocks every crawler on the web, Google included. That is rarely intentional on a live site.

What to do if you find a block

Removing the line is a small edit, and in most website builders it is a settings field rather than a file you upload. The bigger job is the one underneath: making sure that when the crawler comes back, there is a page worth quoting waiting for it.

HowFindable scans your site for free and tells you which AI crawlers your robots.txt is turning away, along with the rest of the signals ChatGPT and Google read. If you want the corrected files, we generate them ready to upload, with steps for your platform.

Can ChatGPT read your site today?

Paste your URL and get two scores, one for Google and one for AI, plus the exact signals you are missing. If you want, we generate the fixed files for you, ready to upload. The scan is free.

Scan your site