C
Citations.io
Start free
Metric · Learn

Share of voice in AI search.

The single most useful competitive metric for the generative era — how often your brand appears in AI answers versus the competitors you actually care about.

TL;DR
AI share of voice is the share of sampled answers, across a fixed prompt set and engine list, that mention your brand vs. tracked competitors. It's measured weekly, expressed as a percentage, and labelled with a confidence tier based on sample size.

The definition

For a given prompt set P, engine list E, and competitor set C (you plus competitors), share of voice for brand b is:

SoV(b) = mentions(b) / sum(mentions(c) for c in C)

Mentions are counted per analysed answer with a deterministic mention parser (exact + alias + entity match). One brand can be mentioned multiple times in one answer but counts as one mention per answer.

Why it's the right metric

  • Bounded — sums to 100% across the tracked set, so it's directly comparable week over week.
  • Engine-aware — you can compute SoV per engine or pooled across engines.
  • Competitive — it explicitly answers "are we winning the answer over X?".
  • Actionable — gaps map cleanly to specific prompts, citations and source pages.

How to grow it

Three reliable levers:

  1. Earn comparison-page coverage — third-party "best X" and "X vs Y" pages are the single biggest driver of mention rate in synthesised answers.
  2. Publish evidence-led claims — pages with specific numbers, dates, named customers and structured comparisons get extracted more often than vague brand copy.
  3. Close prompt-level gaps — find the prompts where a competitor outranks you, then ship content and PR targeted at those specific intents.

Pair this with the tracking playbook and the improvement playbook for an end-to-end program.

Is AI share of voice the same as SEO share of voice?+
No. SEO share of voice is weighted by SERP position and search volume. AI share of voice is computed from sampled answers — the share of answers across a fixed prompt set that mention your brand vs. tracked competitors.
How big does my prompt set need to be?+
At least 50 prompts for directional reads; 200+ for moderate confidence; 500+ for strong reads. Confidence depends on the count of analysed answers, not just prompts — see our methodology.
Can share of voice be over 100%?+
No. It's bounded 0–100% per prompt set, per engine. If multiple competitors share one answer, each gets a partial credit; share-of-voice sums to 100% across the full tracked set.