Read time: 2 min

How to get cited by ChatGPT, not just mentioned

Daniel RothLast updated on Aug 14, 2026Talk to our team →

Key highlights

  • Mention and citation are different outcomes an answer can name your brand while sending every click to somebody else's page.
  • The gap is real and measurable one company we measured was named in an answer that cited four of its competitors' domains and not its own.
  • ChatGPT cites what it retrieved and quoted the page that supplied the answer's facts gets the link, which is an editorial standard you can write for.
  • The fastest diagnostic is reading your category's citations run ten buyer questions, list every cited domain, and count how often yours appears.

In this article

Summarise this article with AIOpen ChatGPT →

There are two prizes in a ChatGPT answer and most visibility work only chases the first. Being mentioned means the model wrote your name. Being cited means your page is one of the linked sources, which is where a buyer actually lands when they want to know more.

The distinction sounds pedantic until you measure it. In one sweep this year we found a brand named in an answer whose citations pointed at four competitor domains and a directory, and not at the brand itself. Every click from its own mention went somewhere else. This article is about closing that gap.

Why mention and citation come apart

ChatGPT composes answers from pages it retrieves at ask time, and it links the pages that supplied the substance. A brand can be named because third-party pages describe it, while its own site contributed nothing to the answer, usually because the site said nothing liftable about the question.

That is the standard failure we see in audits: the brand's presence in the answer is entirely secondhand. The engines know of you from directories and roundups, quote those, and cite those. Your site is a bystander at your own mention.

Citation follows contribution. The cited page is the one that stated the price, the timeline, the comparison, the number, the thing the answer actually used. So the work is publishing pages that are the best available source for a specific question, not pages that describe your brand in general.

Concretely: one page per buying question, opening with a direct answer in the first sentences, carrying specifics nobody else publishes, and structured so the key claim survives being lifted out of context. If your page's best sentence would look wrong quoted alone, rewrite it.

Read the citations before writing anything

Run your ten most important buyer questions through ChatGPT with search enabled and record every cited domain. That list is your category's electorate. Some of it will be claimable, directories and review sites where you can be accurately listed. Some of it is editorial, where the play is being worth writing about. And some of it should be you, once the right pages exist.

We do this by API so the answers and citation annotations are captured with dates, but the manual version in a browser is better than nothing. What matters is doing it before content planning, because the citation list usually contradicts the keyword list.

The order of operations

First make the site retrievable: server-rendered pages, clean titles, no blocked crawlers. Then publish the question-shaped pages. Then close the third-party gaps the citation reading exposed. Then re-run the same questions monthly and track two numbers separately: answers that name you, and answers that link you.

Expect the mention number to move first and the citation number to follow. When the citation number moves, the traffic follows it, because that is the moment the buying journey starts on your page instead of someone else's.

Be the answer AI recommends.

We find the buyer questions where an engine names someone else, then publish and earn the citations that change the answer.

Book a call →