Guide

AI citations

The links an engine shows beside its answer, where they come from, and why the same question produces a different set in each engine.

Check your site for free

What a citation is, and what it is not

A citation is a link an engine shows beside or beneath an answer, naming a page it drew on while writing. That is the whole of it, and the difference from a search result is worth stating early, because it changes what is worth chasing.

A search result has a position, and second place is worth a great deal. A citation list has no positions. There are the few sources shown and there is everything else. Being the seventh best page on a subject is a decent outcome in search and no outcome at all here.

The two steps, and which one you control

Every engine does this in two moves, and separating them explains most of what looks mysterious about citations.

First it gathers. Either from a search index it already holds, or by fetching pages live at the moment somebody asks. This is the step you can influence, and it is almost entirely mechanical: can a crawler reach the page, is the content there without a script running, does the page say what it is about.

Then it writes, and attaches sources. Which of the gathered pages get named is a decision inside a model that no vendor documents. You can make a page easier to quote, by answering the question directly and keeping claims checkable. You cannot make this step choose you, and anyone selling that certainty does not have it.

The practical consequence is unglamorous. Nearly all the work that reliably changes anything happens in the first step, which is why a blocked crawler is worth more of your attention than any amount of writing advice.

Why engines disagree about the same question

Ask two engines one question and the cited sources often barely overlap. That is not a quality signal about your page. It is the gathering step differing.

An engine answering from an index inherits whatever that index already extracted about you, including markup, and inherits its staleness too. An engine fetching live reads the page as served that day, and some of those crawlers convert what they fetch to plain text before the model sees it, which throws away anything that was not in the words.

So a page carrying its meaning only in structured data can be visible to one and invisible to the other, for reasons that have nothing to do with how good it is. The defence is the same either way: put the meaning in the visible words, and treat markup as confirmation rather than as the message.

What makes a page easier to cite

Below the gathering step, where influence is real but partial, a few things recur in pages that get used as sources.

An answer that survives being lifted. If the sentence under a heading only makes sense having read the three paragraphs above it, it cannot be quoted. The paragraphs can stay; the answer has to come first.

Claims a model can check elsewhere. Specific, verifiable statements are safer for an engine to repeat than sweeping ones, because repeating a wrong claim is the failure it is most tuned against.

Being written about by others. This is the part that is not on your site and not in your control, and it is why two technically identical pages get treated differently. An engine leaning on sources about you is not reading only you.

How to actually see your citations

By asking, and only by asking. Run the questions your buyers would ask, record which sources come back, repeat on a schedule. That is what the paid tools automate and it is the whole of their method.

No engine publishes a report of how often it cited you. If a tool implies otherwise, it is sampling exactly as you could and presenting the sample as a measurement. The roundup covers which ones do it well, including where ours does not compete.

Questions about AI citations

What is an AI citation?

A link an engine shows beside or underneath an answer, naming a page it drew on. It is not a ranking position. There is no second place in a citation list, only the handful of sources shown and everything else, which is the difference that matters most when you come from search.

How do AI citations work?

In two steps. The engine first gathers candidate sources, either from a search index it already holds or by fetching pages live at the moment of the question. It then writes an answer and attaches the sources it leaned on. The first step is the one you can influence, because a page that is never gathered cannot be cited no matter what it says.

Why do two engines cite different pages for the same question?

Because they gather differently. An engine answering from an index inherits whatever that index already held about you. One that fetches live reads your page as delivered on the day, and some of those crawlers convert the page to plain text first, which discards anything that was not in the words. Same question, different evidence, different answer.

Can I see which pages AI engines cite?

Only by asking. Run the questions your buyers would ask and record which sources come back, which is what the paid tracking tools automate. There is no report from inside an engine telling you how often you were cited, and anyone offering one is sampling in the same way you could.

Do AI citations send traffic?

Less than a search result of the same prominence, because the answer is often complete without the click. Treat a citation as the thing that gets you considered rather than as a traffic source, and judge it the way you would a mention in an article rather than a position in a list.

Start with the step you control.

Whether an engine can reach your page, parse it and say what you do. Free, one page, under a minute.

Run the free check