Old SEO dashboards ranked the world from one to a hundred. Answer engines do not produce a hundred positions. They produce a short answer with zero, one, two, or a handful of citations. If you are not cited, you do not exist in that answer.
What does “citation” actually mean inside an AI answer?
For a given query, the first thing that matters is: were you cited at all. That is a binary. Only after you clear that binary does the count of citations in an answer matter.
- Binary citation rate: % of runs where your domain is cited at least once.
- Citation share: your citations / total citations across all domains, averaged per query.
The first number tells you whether you are in the game. The second tells you how loudly your voice is in the room.
How many runs do I need before the number is stable?
A single model run is one sample from a distribution. You need five to ten runs to stabilize a binary citation rate for a given query, more if the query has high entropy (open-ended, low-consensus topics). Any dashboard reporting a single run as a “ranking” is selling a fiction.
Which comparison actually drives investment decisions?
Absolute citation rate is hard to interpret in isolation. What you actually want is: for the same query family, for the same five runs, how do your citation counts compare to your three biggest competitors? That delta is what your CFO will recognize as a scoreboard.
