Drivers — what moved, and which questions did it
A chart that wiggles tells you nothing. Drivers answers the question you actually have — “why did my number move?” — by tracing every week-over-week change to the specific questions that flipped, the engines that flipped them, and the rivals involved. And when nothing moved beyond the margin of error, it says so, which is the part most tools can't afford to say.
How it works
Questions matched across scans
Each scan's questions are matched to the previous scan's by text, so the comparison is like-for-like even after you edit the set.
Flips, not wobbles
The drivers card reports transitions — a question you were never named on that you now win, or the reverse. Rate wobbles inside the printed margin are labeled as noise, not news.
Surfaced where you'll see it
The Overview leads with rank movement and a top-mover card; the weekly digest email carries the same sentences. No dashboard archaeology required.
In depth
“Won” and “Lost”, with receipts
Each driver names the question verbatim and the rate change — “Won ‘can I get a free tool for finding trends’ — named in 0% → 33% of runs” — and the answers behind it are one click away in the inspector.
Rank movement among rivals
The Overview's rank tile tracks your position among every brand AI names in your category — #2 of 7, up one — computed from the same share-of-voice data as the leaderboard, so the two never disagree.
Silence is a feature
Margin-gating means some weeks report nothing. That's not a bug — a tool that always finds movement is generating engagement, not measurement. When we say you moved, you moved.
What we won't do: celebrate noise. Daily-delta charts from single runs manufacture drama out of sampling variance. Weekly cadence, repeated runs, and margin-gated reporting are slower — and true.
Common questions
How do I know why my AI visibility changed?+
CiteGraph's drivers trace every change to its cause: the specific buyer questions that flipped won or lost since the last scan, with before/after rates, plus rank movement among your rivals. Every driver links to the verbatim answers behind it.
What counts as a real change?+
Movement beyond the printed margin of error, or a question flipping between never-named and named. Anything smaller is reported as within noise — visible, but not dressed up as news.
Do I get notified when something moves?+
Yes — the weekly digest email carries the same driver sentences, and in-app notifications flag the highest-leverage changes. When nothing beat the margin, the digest says that honestly instead of inventing a highlight.
See it on your own domain — first dossier in about a minute.