Get cited by AI / ChatGPT

How to get cited by ChatGPT

ChatGPT can only cite a page its crawlers are allowed to fetch and can read without JavaScript, so being permitted and readable comes first, and any citation follows from there.

OpenAI reaches the web through named crawlers, and each one fetches the HTML your server returns rather than the fully rendered page a person sees in a browser. GPTBot gathers pages that may inform future models, OAI-SearchBot crawls pages so they can appear in ChatGPT's search results, and ChatGPT-User fetches a page when someone's action in ChatGPT triggers live browsing.

You cannot control whether ChatGPT cites you, but you can control whether it can reach and read you at all. Get the order right: be fetchable, be allowed for OpenAI's tokens, and be readable without JavaScript first. Citation is a consequence that can follow once a page is retrievable, permitted, and structured, and it is never guaranteed.

How ChatGPT reads the web

OpenAI publishes three crawler tokens, each with a distinct job, its own user-agent string, and separate handling in robots.txt. GPTBot fetches pages that may be used to improve future models. OAI-SearchBot crawls pages so they can be surfaced in ChatGPT's search results. ChatGPT-User fetches a page when a user's action in ChatGPT triggers live browsing, rather than crawling the web automatically, and OpenAI notes robots.txt can apply to it differently for that reason. Independent testing of large volumes of AI crawler traffic has found these crawlers retrieve the HTML your server sends and do not execute client-side JavaScript, so text and links that only appear after JavaScript runs may never reach them. When ChatGPT answers, it can draw on model training data, on its search results, or on a page fetched live during the session, and a surfaced link reflects content the system was able to retrieve and read, not a guaranteed placement.

What it takes to be readable and citable

  1. Be fetchable

    Before anything else, OpenAI's crawlers have to complete a clean request. Return a 200 status, avoid redirect chains and login walls on content you want read, and keep the page a reasonable size. Lantad's Retrievability check fetches the page the way a crawler does and reports the status, redirects, and size a bot actually encounters.

  2. Allow the OpenAI crawlers in robots.txt

    robots.txt controls whether OpenAI's automated crawlers may fetch a page, and OpenAI publishes separate tokens: GPTBot, OAI-SearchBot, and ChatGPT-User. A rule that blocks one does not block the others, so name each token explicitly and set it deliberately. Blocking GPTBot opts a page out of training use, and blocking OAI-SearchBot opts it out of ChatGPT's search results. ChatGPT-User is user-initiated rather than an automatic crawler, so OpenAI notes robots.txt can apply to it differently. Lantad's Access check reports the exact robots.txt rule that applies to each OpenAI token.

  3. Serve the content without JavaScript

    Because these crawlers do not run JavaScript, any text injected client-side reaches them as an empty shell. Render the meaningful content in the HTML your server returns. Lantad measures this as Prose Parity: the share of the rendered, human-visible text that is present in the raw HTML a crawler receives. At 100 percent, a crawler sees everything a visitor sees, and the gap is content ChatGPT never gets.

  4. Give the page a machine-readable structure

    Semantic headings give content a clear outline, and structured data labels what each part is. Use real heading levels in order, mark up lists and articles with semantic HTML, and add schema.org JSON-LD where it fits. Lantad scores Structure and Schema separately so you can see whether a machine can both parse the page and classify what is on it.

  5. Be substantive and specific

    Being readable is the prerequisite, and being worth reading is the content's job. Answer the question the page claims to answer, in plain language, with specifics a reader cannot get from a generic summary. Lantad does not grade prose quality, but readable, allowed, and structured pages are the ones that have any chance of being drawn on, and that is the part you control.

Common questions

Does allowing GPTBot get my page cited by ChatGPT?
No. Allowing the OpenAI crawlers and serving readable HTML makes a page eligible to be retrieved and read, which is a prerequisite for citation, not a guarantee of it. No configuration can promise a citation. What you can verify is whether ChatGPT's crawlers are able to reach and read the page at all.
Should I block GPTBot but allow OAI-SearchBot?
They are separate tokens for separate jobs, so decide each on its own. Blocking GPTBot opts your content out of use in future model training. Blocking OAI-SearchBot opts your pages out of ChatGPT's search results, which is the path most tied to being surfaced with a link. OpenAI documents that you can allow one while blocking the other. Lantad reports the rule that currently applies to each token so you can confirm your intent matches reality.
Why does my page look complete in a browser but empty to ChatGPT?
Because OpenAI's crawlers fetch the raw HTML your server returns and do not execute JavaScript, while your browser runs the scripts that build the visible page. If the content is injected client-side, the crawler receives a shell. Lantad's Prose Parity check shows exactly how much of the visible text is missing from the HTML a crawler sees.

See what ChatGPT can read on your site

Run a free scan: paste a URL and see exactly what AI crawlers can and cannot read, with a graded report and ranked fixes.

Scan your site free AI visibility glossary