Get Cited by ChatGPT: A B2B AEO Playbook.

To get cited by ChatGPT, start with a fact most B2B marketing teams have never checked: ChatGPT does not run one crawler, it runs three, and they do three different jobs. One trains the model, one builds the live index ChatGPT actually cites from, and one fetches a page only when a user's own question sends it there mid-chat. Treat all three as the same thing and you can end up blocking the one that controls whether you get cited while leaving the other two wide open, or the reverse. A done-for-you AEO service exists precisely because this kind of distinction rarely gets checked before a retainer gets signed. This playbook covers what each crawler does, how often ChatGPT actually cites a source at all, and the concrete steps that follow from both.

How does ChatGPT decide which sources to cite?

ChatGPT decides what to cite through OAI-SearchBot, the crawler OpenAI's own documentation describes as building the real-time index that powers ChatGPT's search and citation features, kept separate from the data used to train the model itself (OpenAI, "Overview of OpenAI Crawlers"). A page has to be reachable by OAI-SearchBot, and the site has to allow OpenAI's published search-bot IP ranges, before it can ever show up as a cited source in a ChatGPT answer. That is a separate technical gate from whether the page was ever used to train the underlying model, which is GPTBot's job, and separate again from whether ChatGPT fetches the page live mid-conversation because a user's own question pointed it there, which is ChatGPT-User's job. Most B2B teams have configured, at most, one of the three.

Why does ChatGPT have three different crawlers, and why does that matter for citation?

It matters because disallowing the wrong one silently removes you from citation while changing nothing about training, or the reverse, and each is controlled independently in robots.txt.

A site that blocks GPTBot to keep its content out of model training, a defensible choice plenty of B2B companies make, loses nothing in ChatGPT's citation rate as long as OAI-SearchBot stays allowed, since OpenAI's own documentation confirms the two are independent settings. The opposite mistake, a stray "disallow all OpenAI agents" rule left over from a blanket AI-blocking decision, removes a company from ChatGPT citation entirely without anyone on the marketing team realising which line did it.

How often does ChatGPT cite sources at all?

Less than universally, and that rate varies enormously by engine. In SourceRank AI's own tracked run of 100 buyer questions against five answer engines, taken 17 September 2026, ChatGPT returned a source list on 83 of its 100 answers and came back with none at all on the other 17 (full source-mix research).

ChatGPT sits in the middle of the five engines SourceRank AI measures (not Perplexity, which this research does not cover): far more likely to attach sources than Gemini, which attached none in this run, but far less consistent than Copilot, which attached a source list to every one of its 100 answers. That 17% matters practically: on those answers, nothing you did on-page earned a citation link, because there was no citation link to earn. The company still either got named in the prose or did not, a separate and harder problem covered directly in our ChatGPT non-mention diagnostic.

Why does a page that ranks well in Google never get cited by ChatGPT?

The most common reason is that OAI-SearchBot was never confirmed as allowed, so the page is invisible to the one crawler that actually builds ChatGPT's citation index, regardless of how well it ranks anywhere else. Google ranking signals, Bing ranking signals and ChatGPT's own index are three separate systems built from three separate crawls, and strong performance in one tells you nothing about the other two. The second most common reason is thinner: the page is crawlable but carries weak entity corroboration, meaning no independent source outside the company's own site backs up the claim the page makes about itself, which matters even more on the 17% of ChatGPT answers that carry no citation list, because those answers are drawing on recalled, corroborated knowledge rather than a link it can point to.

What is the playbook to get cited by ChatGPT?

1. 1. Check robots.txt for OAI-SearchBot specifically, not just a generic "AI crawler" block, since a blanket disallow rule often catches it by accident.
2. 2. Confirm your hosting provider or CDN allows traffic from OpenAI's published search-bot IP ranges, the same check OpenAI's own crawler documentation recommends before assuming a robots.txt change alone is enough.
3. 3. Decide GPTBot separately and deliberately. Allowing or blocking it changes future model training, not current citation, so make that call on its own merits rather than bundling it with the OAI-SearchBot decision.
4. 4. Build independent corroboration for any factual claim you want cited: press mentions, review platforms, third-party comparison pages, and consistent facts across LinkedIn and Crunchbase, the same entity-signal work our content structure guide for AI citation covers in more depth.
5. 5. Add or tighten schema.org markup, Organization, Product and FAQPage, so OAI-SearchBot has machine-readable facts to cite directly rather than having to infer them, covered in our schema markup guide for AI citation.
6. 6. Structure your highest-priority pages with a question-style heading followed immediately by a direct, self-contained answer, since a page has to be both findable and quotable.
7. 7. Track a fixed, realistic set of buyer prompts in ChatGPT specifically on a schedule, and log whether you are named, whether a source list appears at all, and which domains get cited instead, the discipline our AI visibility measurement runs across all five engines at once.

Frequently asked questions