AI visibility metrics that matter

Four metrics cover AI visibility: Share of AI Answer (the headline — % of tracked questions where you’re cited), per-engine citation rate (where you’re weak), competitor win rate (who beats you on the questions you lose), and answer-source mix (which domains engines cite in your category — your to-do list).

Last updated 2026-07-24

The four that matter

• Share of AI Answer — cited answers ÷ total answers across your prompt set. Track the 30-day trend, not the daily number.
• Per-engine citation rate — the same score split by engine. A 50% ChatGPT / 10% Gemini split is an action plan, not a curiosity.
• Competitor win rate — on prompts where you are NOT cited, who is? The domains that repeatedly win your lost prompts are your real competitors in AI answers.
• Answer-source mix — the third-party surfaces (directories, communities, publications) engines cite for your category. Presence there is usually cheaper than outranking anyone.

Metrics to skip

• "AI mentions" without a denominator — 12 mentions of what? Without a fixed prompt set there is no rate and no trend.
• One-off screenshots — answers vary run to run; a single sample proves nothing in either direction.
• Sentiment scores on tiny samples — with a handful of answers per week, sentiment classification noise swamps any real signal.

Reading the trend honestly

Grounded answers are stochastic: the same prompt can cite you today and not tomorrow. Treat daily runs as samples, read weekly averages, and only act on movements that persist. A disciplined prompt set held constant for months is worth more than a sophisticated dashboard over a changing one.