Getting cited · 4 min read
How to get cited by ChatGPT
Four things decide whether ChatGPT can cite your page: crawler access, server-rendered content, a clear publisher identity, and a passage that answers the question in its first sentence. The first two are configuration; the last two are writing.
On this page
1. Confirm OAI-SearchBot is not blocked
OAI-SearchBot is the crawler that decides whether ChatGPT Search can cite you, so check your robots.txt for it first. The block is usually accidental: a CDN bot-blocking toggle, a security plugin, or a staging Disallow: / that shipped.
OpenAI runs three separate tokens, and only one of them governs citations:
| Token | What it does | Block it and |
|---|---|---|
| OAI-SearchBot | Indexes pages for ChatGPT Search | ChatGPT Search cannot cite you |
| GPTBot | Collects pages for model training | You opt out of training; citations are unaffected |
| ChatGPT-User | Fetches a page during a conversation | ChatGPT may not open your page when asked |
The AI crawler checker reads your file and reports each bot individually, with the exact lines to change.
2. Make sure the content exists without JavaScript
Fetch your own page with curl and search the output for a distinctive sentence from the article. If it is not there, OAI-SearchBot cannot see it either.
curl -s https://example.com/your-page | grep "a sentence from your article"The fix is server-side rendering or static generation for the main content. Prerendering for bot user-agents also works and is faster to deploy, though it needs care to avoid serving something different from what humans see.
3. Say who published it
Give the page an entity an engine can resolve: an Organization block with sameAs links to profiles that actually exist, a named author, a visible date, and a link to an About page. An anonymous claim is a claim an engine has to take on faith, and it usually will not.
The schema generator builds the JSON-LD, but the sameAs list is the field that carries the weight: a LinkedIn company page, a Wikipedia entry, a GitHub organisation.
4. Put the answer in the first sentence
This is the one that actually moves the score, and it is the one nobody wants to do, because it means editing prose rather than configuration.
Take a section headed “How long does espresso extraction take?”
| Opening | What an engine gets |
|---|---|
| “In this section we’ll look at extraction timing…” | Nothing quotable |
| “A standard double shot should extract in 25 to 30 seconds.” | A complete, attributable answer |
The second version is not shorter or simpler. It just puts the answer where an engine reads first. Everything else in the section can stay exactly as it was.
What you cannot control
You cannot make ChatGPT cite you. Citation depends on the query, the competing sources and OpenAI’s own ranking, none of which is visible from outside. What these four steps buy is eligibility: being in the set of pages that can be cited, which is strictly necessary and not remotely sufficient.
5. Check whether it worked
Two signals show it working, and neither appears in a standard analytics tag. The first is OAI-SearchBot fetching the page, which only your server sees, because crawlers do not run JavaScript. The second is visitors arriving from chatgpt.com, which many analytics setups file under referral or direct traffic.
CrawlRadar records both: its collector logs every OAI-SearchBot visit and verifies it against OpenAI’s published addresses, and its tag counts the visitors ChatGPT sends. For the manual route, see how to track ChatGPT referral traffic in GA4 and how to track AI crawlers.
Sources
- OpenAI, Overview of OpenAI crawlers: the roles of OAI-SearchBot, GPTBot and ChatGPT-User, and the 24-hour robots.txt delay.
- CrawlRadar crawler-access and answer-first checks: how each step is scored.
Frequently asked questions
Which crawler does ChatGPT use to cite pages?
OAI-SearchBot. OpenAI documents it as the crawler that surfaces websites in ChatGPT’s search results. GPTBot is a separate crawler for model training, and ChatGPT-User fetches a page when a user’s conversation needs it. OpenAI states the settings are independent, so a site can block GPTBot and still be cited through OAI-SearchBot.How long does it take to appear in ChatGPT after fixing my site?
OpenAI says a robots.txt change takes about 24 hours to reach its search systems. After that the page still has to be re-crawled by OAI-SearchBot, so expect days to weeks. There is no submission endpoint and no way to expedite it.Does ChatGPT cite pages that require JavaScript to render?
Generally no. OpenAI’s crawlers read the HTML they fetch rather than running your JavaScript, so a client-rendered page looks like an empty shell with a mount div. If your content only exists after hydration, it does not exist as far as the crawler is concerned.
Written by
Iliass El Barhoumi, Founder of CrawlRadar
CrawlRadar records the AI crawlers that visit a website and the visitors AI assistants send it. Every vendor claim in this guide links to the vendor’s own documentation.