What an Overview tracker actually records
Three facts per query, per check: did an AI Overview appear at all, was your domain among the sources it cited, and what did the generated summary say about your topic. Each fact moves independently, which is why a single presence score flattens the story.
Recorded on a schedule, those three facts become the history that answers the questions that matter: is Google showing an Overview for our queries more or less often, and when it does, are we in it.
Why spot checks fail here specifically
Overview presence is intermittent by design. The same query can show an Overview on Monday, plain results on Wednesday, and a differently sourced Overview on Friday, so any single check lands on one frame of a moving picture.
A logged-in browser makes it worse: personalization and search history steer what you see, which is how teams convince themselves they are cited everywhere or nowhere. Repeated, clean, scheduled sampling is the only read worth reporting.
Tracker and rank tracker, same query, different facts
What a rank tracker sees
Your blue-link position. It can hold perfectly steady while the Overview above it absorbs the clicks, and it says nothing about whether the Overview cites you.
What an Overview tracker sees
Whether the Overview exists, who it cites, and how its sources rotate. Citation can vanish while rank holds, and appear while rank slips; the two histories disagree often enough to need both.
Reading the history
Presence rate over repeated samples is the base metric: the share of checks where an Overview appeared and the share of those where you were cited. Movement outside normal variation is the signal; everything else is rotation noise.
AeoWatch's AI Overview tracking runs exactly this loop on your query set, with the samples kept behind every rate. The how-to guide covers the page work that improves the odds the citations go your way.