How to track AI Overviews
Track presence continuously, not at a point in time: the same query can produce an overview one week and not the next. Record trigger rate, citation rate within triggered queries, and clicks separately, because citation and traffic diverge.
Why point-in-time tracking fails here
Conventional rank tracking works because rankings are comparatively stable: a position checked on Tuesday is usually a fair representation of that week. AI Overviews are not stable in that way. Whether one appears at all varies, and when it appears the source set varies too.
So a single check tells you whether an overview existed at that moment, which is a much weaker claim than it looks in a report. A series built from weekly single checks will show swings that are sampling noise and invite decisions based on them.
The three numbers worth tracking
Trigger rate
The share of your tracked queries that produce an overview at all, measured across the window rather than on one day. This is the size of the surface available to you, and it moves independently of anything you do.
Citation rate within triggered queries
Of the queries that do produce an overview, the share where you are cited. This is the number your work actually affects, and reporting it against all queries rather than triggered ones understates your performance.
Clicks on those queries
Tracked separately and expected to fall where you win. An overview answers the question on the results page. Being cited is brand exposure with modest traffic attached.
How to build the series
- 1Fix a query set and hold it constant. Changing it mid-period starts a new series.
- 2Check each query several times across a month rather than once, and report the rate across checks rather than a snapshot.
- 3Record the cited sources every time, not just whether you were there. Source concentration is where the actionable finding usually is.
- 4Segment by query type. Informational queries trigger overviews far more often than transactional ones, and blending them hides both signals.
- 5Pull Search Console clicks and impressions for the same query set over the same window, so the traffic effect sits next to the citation effect.
- 6Annotate model and surface changes when they happen, so a step change has an explanation attached rather than a theory.
Reading the traffic impact honestly
The pattern to expect on queries where you start being cited is impressions flat or up, clicks flat or down, and position unchanged. That combination reads like a failure in a conventional SEO report and is not one.
The corresponding pattern where an overview appears and you are not cited is impressions up, clicks down, position unchanged. That is a real loss, and it is worth separating from the first case, because the two look almost identical in Search Console.
The only way to tell them apart is to know whether you were cited, which is why the tracking above exists.
Our position
We report AI Overview presence as brand exposure and never as a traffic channel. A team that budgets for it expecting sessions will conclude the work failed at exactly the moment it started working.
Related questions
How AI Overviews select their sources
Related answers
How to rank in AI Overviews
You cannot rank in an AI Overview: there are no positions.
Read the answerHow-toHow to monitor AI search visibility
Build a prompt set of 30 to 50 buyer questions and freeze it.
Read the answerDecisionHow to report AI search visibility to a board
Report four things: share of voice against a fixed competitor set, brand accuracy, source concentration, and citation rate per platform.
Read the answer