How to Get Cited by ChatGPT (What the Docs and the Studies Say)

To be cited by ChatGPT you need two things that are usually discussed as one. First, ChatGPT's search crawler has to be allowed to fetch your pages, which is a robots.txt question and takes ten minutes. Second, when someone asks a question in your subject, the retrieval step has to reach for your domain rather than one of the several thousand others it could pick, which is a provenance question and takes months or years.

Almost all published advice is about the first thing, or about page formatting. The measurable difference between sites that get cited and sites that do not is mostly the second thing. This guide covers both, in the order that actually helps, and quotes OpenAI's own documentation rather than paraphrasing it.

What does ChatGPT use to answer a question?

ChatGPT's search features run on a search index of their own, separate from the data used to train the model. OpenAI's crawler documentation names four agents and gives each a job:

The consequence is written down plainly: "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links."

Two things follow. Blocking GPTBot to stay out of training data costs you nothing in citations, and it is a reasonable thing to want. Blocking OAI-SearchBot removes you from ChatGPT's answers entirely, and plenty of sites have done it by accident with a broad rule. The two decisions are separate, and whether to block AI crawlers goes through the whole set.

How do you make sure ChatGPT can see your site?

Five checks, in order. None of them takes long, and all of them are prerequisites rather than optimisations.

1. Fetch your own robots.txt and read it for User-agent: OAI-SearchBot and for any blanket User-agent: * disallow that would catch it. A rule that blocks every bot to save bandwidth blocks this one.
2. Check what your CDN does independently of robots.txt. Cloudflare and similar providers ship bot-management rules that block AI crawlers regardless of what your file says, and on some plans that is the default for a new domain. The robots.txt file is not the whole answer.
3. Confirm the page is indexable. A noindex on the page, or a login wall, removes it from every retrieval pipeline before any selection happens.
4. Check your server logs for OAI-SearchBot hits. If the agent has never fetched you, nothing above is working, and no amount of page-level tuning will change the outcome.
5. Ask ChatGPT a question your page should answer and see what it cites. That is the only test of the actual product. How to tell whether AI assistants are citing your site covers doing this systematically rather than once.

What makes ChatGPT choose one page over another?

The largest public dataset on this is SE Ranking's study of ChatGPT citation factors, published 24 November 2025, covering 129,000 domains and 216,524 pages across twenty niches. It is correlational, so read it as a description of what cited sites look like rather than a set of levers, but the pattern is consistent and it is the best evidence available.

The largest single factor in it is the one nobody sells a tool for. The study reports that "Sites with over 32K referring domains are 3.5x more likely to be cited by ChatGPT than those with up to 200 referring domains." Trust scores move with it, with high-trust domains earning close to four times the citations of low-trust ones. Length and freshness matter but less: articles over 2,900 words averaged 5.1 citations against 3.2 for pieces under 800, and pages updated within three months averaged 6 against 3.6 for stale ones.

The study also found that llms.txt files and FAQ schema had negligible or negative association with citations, which lines up with what the vendors document. Whether llms.txt does anything covers that specific claim in detail.

So the shape of the answer is: content quality gets you into contention on a page-by-page basis, and the domain's reference profile decides whether you are in contention at all. That ordering is the opposite of how most AI-visibility advice is written.

Which sources does ChatGPT lean on hardest?

Citations are more dispersed than the "get on these ten sites" advice suggests. An analysis of 22,295 AI answers to business-software prompts found 7,058 distinct cited domains across ChatGPT, Perplexity and Google AI Mode, with the top ten domains on each engine covering only 11 to 13 percent of citations.

Within that, one source stands out for ChatGPT specifically. Profound's analysis of citations from August 2024 to June 2025 put Wikipedia at 7.8 percent of all ChatGPT citations, the largest single domain by a wide margin. That is a useful fact in two directions: it tells you where ChatGPT anchors factual claims, and it tells you that being referenced from an editorially controlled source is worth more than being referenced from many uncontrolled ones. The platform-by-platform tables have the rest.

Using the Revised directory for this

The provenance half of the problem is the slow half, and a domain is the only part of it you can buy rather than earn. That is what our directory is for: expired domains that are still referenced by sources answer engines lean on, available to register.

Practically:

1. Open the directory and set the Basics filters to your TLD and, where the classifier has a code for it, your subject category.
2. Use the source filter, or start from a source hub directly. Expired domains with Wikipedia backlinks is the most relevant hub for ChatGPT specifically, given the 7.8 percent figure above. Domains with .edu backlinks and .gov backlinks are the two hardest reference types to obtain any other way.
3. In Quality, set a minimum Agent Citability score. Agent Citability is our published measurement of how deeply a domain is referenced by the kinds of sources answer engines cite. Third parties cannot recompute it, it ships as versioned releases, and a score can fall between releases when the pages holding the references change.
4. In History, set a minimum age and restrict to rows whose availability was checked recently, so you are not shortlisting names somebody has already taken.
5. Read the Linked by column before you reveal anything. Who references a domain tells you more than how many do, and the subject those references were about is the thing that has to match yours.

Every filter and every sort is free, signed in or not. Reveals are what a plan buys, so shortlist first.

You can also run the Agent Citability checker and the backlink checker on a domain you already own, and on whichever domains ChatGPT cited when you ran your test questions. The gap between the two numbers is the size of the job.

One condition on all of it. Google's expired domain abuse policy covers domains repurposed "primarily to manipulate search rankings by hosting content that provides little to no value to users", and the references on a mismatched domain were vouching for a subject that is not yours in any case. What Google's policy actually says has the verbatim text and the test.

What does not work

Machine-readable files aimed at AI. No major assistant documents reading one.

Blocking GPTBot in the belief that it affects search. It does not, per the table above.

Assuming a page that ranks will be cited. Retrieval for an answer and ranking for a query are related but not the same selection, and a measurement study of Google AI Overviews found that nearly 30 percent of cited domains did not appear in the first-page results for the same query at all.

Buying a domain in an unrelated subject because its score is high. The references were about something else.

FAQ

How do I get my website cited by ChatGPT? Allow OAI-SearchBot in robots.txt and at your CDN, make sure the page is indexable, then work on the reason a retrieval step would reach for your domain: references from sources that are themselves cited. The crawler part takes ten minutes. The reference part is the actual work.

Does blocking GPTBot stop ChatGPT citing me? No. OpenAI documents GPTBot as the training crawler and OAI-SearchBot as the search crawler. Only opting out of OAI-SearchBot removes you from ChatGPT search answers.

How long does it take to get cited by ChatGPT? There is no published figure and nobody can honestly give you one. A newly indexable page can in principle be cited as soon as it is retrieved. In practice the engines skew heavily toward domains with an existing reference profile, so a new domain waits as long as it takes to build one.

Does an aged domain help? An aged domain that is already referenced by sources the engines cite starts inside the pool they draw from. Age by itself does nothing, and a domain cited for an unrelated subject is a poor fit whatever its score.

Is being in ChatGPT's training data the same as being cited? No. Training and search are separate pipelines with separate crawlers. Content in the training set can inform an answer without being cited; citations come from the search index.

Can I pay to be cited? Not as a citation. OpenAI runs ads on ChatGPT, which is why OAI-AdsBot exists, but that is advertising and appears as advertising. Nobody sells placement in the cited sources.

The half of this you cannot rush is the domain, so decide it first. The Revised directory of expired domains with backlinks shows the referencing sources by name on every listing before you reveal it, with spam and trademark screening on every row, so you can find a domain that a reference source already chose to cite and check whether its subject is yours.