Citation share is how often AI answers in your category name you versus your competitors. Run twenty buyer questions through ChatGPT, Perplexity and Google AI Overviews; if you are named in eight of the sixty answers and your rival in twenty, your citation share problem has a number. It is the AI-era successor to ranking position: the single metric that tells you whether the invisible funnel, the one your analytics cannot see, is working for you or against you.
- Citation share = your presence rate across AI answers in your category.
- It replaces ranking position as the visibility metric that matters.
- Measured with a fixed prompt set, across engines, over time.
- Trend beats snapshot: single answers vary, patterns do not lie.
Why a new metric was needed
Ranking position answered the old question: where do I stand on the list the buyer scrolls? AI answers killed the list for a growing share of decisions. When ChatGPT names two providers and the buyer contacts one, no ranking was consulted and no click was logged. Businesses kept staring at positions and sessions while the actual decision surface, the generated answer, went unmeasured. Citation share measures it. It is the difference between knowing your visibility and assuming it.
An AI answer names two or three businesses where page one listed ten. Citation share is a harsher contest than rankings ever were: most categories have a few winners and an invisible majority.
How to measure it, hand-rolled and honest
You need three fixed things: a prompt set, an engine set, and a schedule. The prompt set: fifteen to twenty five questions your buyers actually ask, phrased the way they ask them, covering hiring questions, comparisons and problem descriptions. The engine set: at minimum ChatGPT, Perplexity and Google AI Overviews, because their retrieval differs and so does your presence in each. The schedule: monthly, same prompts, fresh sessions so your own history does not bias the answers.
Log per answer: were you named, in what position among the named, with what framing, and who else appeared. Your citation share is the percentage of answers naming you. Your competitors' shares come free from the same log, and the gap between yours and theirs is your roadmap, expressed as a number.
Reading the number like an operator
Zero share with competitors present: a legibility or trust gap, diagnose with the competitor analysis. Share in one engine but not another: retrieval difference, usually an indexing or crawler-access issue on the missing engine, check who can read you. Named but with wrong facts: a consistency problem, which counts as damaged share, not share. Flickering presence month to month: normal at the margin, models sample, which is exactly why the trend across months is the signal and any single answer is noise.
The advanced layer: share of model
Citation share counts appearances. Its deeper cousin, share of model, asks how favorably and in what role you appear: are you the recommended pick, the alternative, the caveat? Two businesses with equal citation counts can have wildly different value if one is always "the specialist to hire" and the other "a cheaper option some mention". When you log citations, capture the framing sentence too. Moving your framing is GEO working at its highest level, and it is the part of measurement no tool automates honestly yet, which is why I still do it by hand for clients.
Common questions
How many prompts do I need to measure citation share?
Enough to make the number stable: fifteen to twenty five buyer questions is the practical range. Fewer and single-answer randomness dominates; more and the monthly audit stops happening because it became a chore. Fix the set and resist editing it, because a changing ruler measures nothing.
Is there a tool that tracks citation share automatically?
Tools exist and are maturing, some run prompt sets across engines on a schedule. The honest caveats: engine coverage varies, prompt realism matters more than volume, and framing, the role you play in the answer, still needs human reading. Automation helps the counting; the judgment layer remains manual for now.
What is a good citation share?
Category-dependent, and the useful benchmark is internal: your trend versus your competitors' trend on the same prompt set. In an uncontested niche, majority share is achievable fast. In a contested one, moving from zero to consistently-named is the first victory, and framing improvements are the second.
Want your citation share measured properly?
Scan your site free to check the foundation, then send me your category. I will run the prompt set and give you the number, and the gap.
Run my visibility check