An AI Overview is not a feature you either have or lack. It appears for some queries and not others, cites a rotating cast of sources, and quietly rewrites its own sub-questions. We measure all three.
Trigger rate by intent
We sample each tracked keyword on a fixed cadence and record whether an Overview renders. Trigger rate is the share of checks where it does. Intent predicts it better than industry does: questions and definitions trigger an answer most of the time, while transactional queries mostly keep the classic layout. Intent is assigned per keyword and re-scored whenever a keyword’s behaviour drifts.
Sample snapshot, share of checks that rendered an AI Overview. Orange marks rates of 50% or more.
Citation slot turnover
Overviews cite a handful of sources, usually three to eight links, and the set rarely stays put. We track each citation slot as a seat and log who occupies it. Turnover is the share of cited URLs replaced after a given number of days. Slots are matched by canonical URL, so tracking parameters never count as turnover.
Informational answers settle: roughly three in ten citations are still there after four weeks. Commercial answers churn faster, with about one in eight surviving. A citation you earned last month is a lease, not a deed, and the retention curve tells you how long the lease runs.
Retention curve
Query fan-out stability
Before writing an answer, the system fans one query out into several narrower sub-queries and retrieves documents for each. Those hidden sub-queries decide who gets cited. We reconstruct the fan-out from rendered follow-up chips, answer sections and source clusters, then compare consecutive checks using Jaccard overlap, scored from 0 to 1.
0.71Informational overlap
0.46Commercial overlap
3–8Typical citation slots
A score near 1 means the answer is built on the same questions every time, and your content can target them. A low score means the topic is being reframed, usually after a model update, and your coverage may be answering last month’s questions.
What to do with it
Treat trigger rate as your exposure and citation retention as your defence. High exposure with weak retention means you are renting visibility from a lottery. Invest in the pages that already hold slots, refresh them before the retention curve bends, and add original data that answer engines prefer to quote. Low-exposure keywords still deserve a monthly look, because a trigger rate that jumps from five percent to thirty usually arrives without warning.
The pixel cost of an Overview
An expanded Overview is about 420 px tall on desktop and 640 px on mobile in our fixtures, so every trigger carries a displacement cost on top of any click lost to the answer itself. Telemetry joins the two: for each keyword you see the trigger rate, whether you are cited, and where Organic #1 lands. Cited and displaced is a very different problem from uncited and displaced. Model it in the simulator.
Alerts
Webhook and dashboard alerts fire when a keyword gains or loses an Overview, when your domain enters or leaves a citation slot, or when fan-out overlap falls below a threshold you set. Every alert links to the render that triggered it. Analyst plans include trigger-rate tracking. Citation and fan-out tracking start on Growth Team; see pricing.