Skip to main content
Citation rate is the share of successful runs in which a model attached at least one link to your own domain. It is successful runs citing your domain, divided by every successful run in the window — the same denominator as mention rate, so the two are directly comparable.

​
What counts as a citation?

A citation is one URL the model attached to its answer, pointing at a domain you own. Links appear as footnotes, source lists, inline references, or a sources panel depending on the model surface; Attensira normalises all of them to the same thing — a URL, attributed to a domain. As with mentions, the unit is a run. A run where the model links four of your pages counts once toward citation rate. Those four URLs still matter, but they matter to citations and sources, which counts URLs rather than runs.
Failed runs are excluded from the denominator. Runs where the model wrote an answer with no links at all still count in the denominator — a linkless answer is a real answer, and it is a real absence of a citation.

​
Why is citation rate lower than mention rate?

Almost always, yes — and the gap is the point, not a bug. Two independent things have to happen for a citation. The model has to decide your brand belongs in the answer, and then it has to decide a page on your domain is worth linking. The first is a condition of the second in practice, so citation rate is bounded from above by how often you are discussed at all. Several ordinary mechanics widen the gap:
  • Models answer from memory. A model naming you from training or internal knowledge has no URL to attach. It knows you without having just read you.
  • Not every surface links. Some model surfaces attach sources on nearly every answer; others rarely do. Your citation rate is partly a property of which models you track.
  • The link may go somewhere else. A model can recommend you and cite a review site, a marketplace listing, or a comparison article that talks about you. That is a mention with a citation to someone else’s domain.
The engine’s own framing: “Mentioned is not cited: a model can name a brand it never links.” A wide gap between your two rates is a legible, actionable signal — you are known but not read.

​
What do I do with a wide gap?

Treat mention rate and citation rate as answering different questions. Mention rate asks whether the model considers you part of the answer. Citation rate asks whether your own pages are the material it drew on. A brand with high mentions and low citations is being described through other people’s writing; the lever there is publishing pages that answer the prompts directly, and seeing which of your pages get picked up. A brand with a low mention rate does not have a citation problem yet. Fix the first number first — no amount of linkable content helps in answers where you are never named. Before reading a gap as meaningful, check n on both rates and check whether the difference cleared the significance gate described in How we collect data. Two rates built on a handful of runs can differ by a wide margin purely by chance.