What AI Overview citation rate measures
You take a fixed list of queries, check which of them return an AI Overview, and record whether your domain appears among the sources that Overview links to. The citation rate is your appearances divided by the Overviews measured. The denominator is the part people get wrong: an Overview does not appear for every query, so a rate calculated across all tracked queries and one calculated across the queries that actually triggered an Overview are different numbers with the same name.
Because the Overview is generated per request, results move with location, device, sign-in state and the model version behind it. A rate is therefore always a measurement of a sample, not a property of your site, and it should be recorded with the conditions that produced it.
Why the citation rate matters
An Overview sits above the ordinary results and often answers the question outright, so the click that used to follow a good ranking may not happen. Being cited is what is left of that visibility: your name in the source list, sometimes your phrasing in the summary. Tracking it tells you whether Google is drawing on your pages when it composes the answer for your topics.
It also gives a reason for a pattern that otherwise looks alarming in Search Console — impressions holding steady while clicks slide on informational queries. If your citation rate is healthy on those queries, you are still in the answer even though the visit did not follow, which is a very different situation from disappearing altogether. That distinction is the practical core of optimising for AI Overviews.
Where the citation rate goes wrong
The commonest error is treating a citation as a visit. It is exposure, and sometimes attribution, but it is not traffic, and reporting it as though it were will eventually get found out. The second is mixing capture conditions: mobile and desktop, Kathmandu and Sydney, signed-in and signed-out results are separate populations, and averaging them produces a figure that describes nobody.
The third is comparing your rate with a rate published elsewhere. Nobody’s query set matches yours, Overviews trigger at different frequencies in different categories, and vendors define the denominator differently. There is no shared scale here, so an external comparison is decoration rather than evidence.
How to act on it
Fix the query list around questions your customers really ask, record for each query whether an Overview appeared before recording whether you were cited, and hold location and device steady between runs. Read the rate beside clicks and impressions from Search Console rather than in its own slide, so the trade-off between exposure and visits stays visible.
Where you are absent from Overviews you should be in, look at how the cited pages are written: a direct answer near the top, a clear heading matching the question, and facts stated plainly enough to be lifted without rewriting. That is a question of extractability more than of authority, and it is usually the cheapest thing to fix. Where the queries are commercial and the Overview is thin or absent, the ordinary ranking work still applies unchanged.
