Share of voice (AI search)

Share of voice is the competitive view of AI visibility. For a defined set of buyer prompts, you count the answers that mention or cite each brand in the category and express each brand’s count as a share of the total. The term comes from advertising, where it meant a brand’s portion of all the advertising in its category, and the logic carries over: presence only means something relative to everyone else competing for the same attention.

Tools differ in what they count. Some count any mention, some count only citations, and some weight a mention by where in the answer it appears or by the search demand behind the prompt. Two share-of-voice figures from different tools are rarely comparable, so the definition matters as much as the number.

A raw mention rate has no context. Being named in a third of answers is strong in a crowded category and weak where there are only two serious vendors; share of voice tells you which. It also names who is taking the answers you are missing, which is usually the best place to start the content and PR work.

Advertising research has long linked holding more share of voice than your share of market with growth, which is why marketers already track it. In AI search the same instinct applies, with one caution: generated answers vary run to run, so share of voice has to be measured across many runs, not read off one.

• Semrush gives the basic formula as AI share of voice = (your AI mentions ÷ total AI mentions across all brands in your category) × 100, and says its Brand Performance report weights both how often a brand is mentioned and how high it appears.Semrush, How to measure AI share of voice, 17 July 2026, 15 Sept 2026
• Ahrefs defines AI share of voice in Brand Radar as the percentage of AI responses within your topic set that mention or cite your brand versus competitors.Ahrefs Brand Radar product page, 15 Sept 2026
• An IPA article on excess share of voice states that every 10ppts of ESOV would be expected to result in about a 0.5ppt increase in market share, depending on the category.IPA, ESOV: an excessive focus on the wrong thing?, 13 November 2023, 15 Sept 2026
• <1%SparkToro found that repeated runs of the same recommendation prompt returned the same list of brands less than 1% of the time.SparkToro research on AI recommendation consistency, reported by Search Engine Journal, 30 January 2026, 15 Sept 2026

1. 1

   Fix the competitor set first: the brands a buyer would genuinely shortlist, not everyone in the category. Adding or dropping a competitor changes every share, so keep the set stable between reports.

2. 2

   Decide and write down what counts: mentions, citations, or both, and whether position in the answer is weighted. Report mention share and citation share separately if you can.

3. 3

   Run each prompt on each engine several times per period and total across runs before calculating shares, so one unusual answer cannot swing the figure.

4. 4

   Break the result down by prompt stage and by engine. A healthy overall share can hide a competitor owning every comparison-stage question.

5. 5

   For the prompts where a competitor leads, read which pages are cited and what they contain. That list is your content and outreach brief.

• Comparing share-of-voice numbers across tools that define it differently.
• Changing the prompt set or competitor list between periods and reading the difference as a trend.
• Letting a few high-volume branded prompts dominate the total.
• Chasing share of voice on engines your buyers do not use.

AI visibility

AI visibility is your absolute presence in answers. Share of voice is that presence as a share of all brands appearing for the same prompts.

Share of voice in advertising

The advertising measure counts paid media weight. The AI search measure counts organic appearances in generated answers, which you cannot buy directly.

1. 1Semrush, How to measure AI share of voice, 17 July 2026 read 15 Sept 2026
2. 2Ahrefs Brand Radar product page read 15 Sept 2026
3. 3IPA, ESOV: an excessive focus on the wrong thing?, 13 November 2023 read 15 Sept 2026
4. 4SparkToro research on AI recommendation consistency, reported by Search Engine Journal, 30 January 2026 read 15 Sept 2026