Share of voice in AI search, sometimes called answer share, is the percentage of all category answers that name your brand across a basket of priority buyer questions. It is the cleanest single rollup of AEO performance, the AI equivalent of organic share-of-voice in classic SEO, and the metric most often reported to leadership when an AEO program reaches steady state.
How to compute it
Pick a basket of 25 to 50 priority buyer questions in your category. Run each one across each of the major answer engines (ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews) 25 to 50 times per engine to wash out stochastic variance. Count, across all those runs, what fraction of answers named your brand. That is your AI share of voice. The same denominator applied to each competitor produces the competitive landscape.
Why share of voice beats single-question metrics
Mention rate on a single buyer question is noisy and easily distorted by one viral comparison article. Share of voice across 25-plus questions is much harder to game and reflects real category authority. It also smooths out the per-engine variance: a brand that is strong on Perplexity but weak on ChatGPT may show flat single-question metrics but a clear share-of-voice gap. Reporting the rollup makes the trend interpretable.
Reading week-over-week changes
Share of voice tends to move in 2 to 5 point weekly swings under normal conditions. A 10-point move in a single week almost always corresponds to either a real change in the category corpus (a new comparison article going viral, a competitor launching, a major review site updating) or a measurement artifact (a question rephrased mid-week, samples too small). When a move is real, you can usually identify the source within an hour by sampling the citations that changed.
Weighting by question importance
Equal-weighted share of voice is a useful default but misleading once your priority questions vary in commercial value. A weighted version uses each question's estimated buyer volume (or revenue per buyer) as the weight, so a 5-point gain on a low-volume awareness question counts less than a 1-point gain on a high-intent purchase question. AskRanker's report supports both views and we recommend the weighted version for executive reporting.
What share of voice does not capture
Share of voice is binary at the answer level: you are either named or not. It misses position (were you first or fifth in the answer), sentiment (were you described positively, neutrally, or as a fallback), and citation (was your URL the source). Pair share of voice with position rate and citation rate for a more complete picture, especially if you are competing against entrenched incumbents whose mention rate is hard to dislodge but whose position you can leapfrog.