AI share of voice (AI SoV) is a competitive visibility metric: the percentage of brand mentions in AI-generated answers that belong to one brand, measured against competitors across a defined set of prompts, engines and period. It adapts the classic share-of-voice idea from advertising and search to answer engines such as ChatGPT, Perplexity, Gemini and Google AI Overviews, and shows how much of a category's AI conversation a brand holds.
Practical example
A Belgian HR software vendor tracks 60 prompts such as "best payroll software for SMEs in Belgium" across ChatGPT, Perplexity and Gemini, run several times a week. In a month, the answers contain 900 brand mentions: 270 for the vendor, 360 for competitor A, 180 for competitor B and 90 for others. The vendor's AI share of voice is 30%. Competitor A leads at 40%, so the gap is clear.
Why is AI share of voice important for AEO?
Visibility alone says whether a brand shows up; share of voice says whether it shows up more or less than the brands it competes with. Because answer engines often present a shortlist rather than a single result, the category conversation is finite, and every mention a competitor gains is relative ground lost.
It also makes AEO work measurable: when content or third-party coverage shifts the balance across the same prompts, the metric moves. It is only as good as the prompt set, competitor list and sampling behind it, so it should always be reported with that method attached.
How is AI share of voice calculated?
Share of voice is a ratio, so every figure depends on what goes into the numerator and the denominator. The usual method has five steps:
- Define the scope. Choose a prompt set that reflects real buyer questions (category, comparison and "alternatives to" queries rather than branded ones), a named competitor set, the engines to track, the country and language, and the reporting period.
- Sample the answers. Run every prompt on every engine several times across the period. Answer engines do not return stable results, so one run is a snapshot and the metric lives in the aggregate.
- Count the mentions. Record each tracked brand, including product names and spelling variants, once per answer. Put brands outside the competitor set into an "others" bucket so the denominator stays complete.
- Calculate. AI share of voice = your brand's mentions ÷ mentions of all tracked brands × 100. With the "others" bucket included, the shares add up to 100% across the set.
- Segment. Report the figure per engine and per prompt cluster. A brand can lead on Perplexity and trail on ChatGPT, or dominate "alternatives to X" prompts while missing from "best for SMEs". Position-weighted variants, where a first mention counts more than a fifth, exist but belong next to the raw share, not in place of it.
Two methods, one name. Some tools and articles instead divide the answers that mention a brand by the total number of answers. That is a mention rate, or AI visibility, not a share of voice: the percentages across brands can add up to far more than 100%, because one answer usually names several brands. A brand can appear in 65 of 100 answers (65% visibility) and still hold only 25% share of voice if competitors appear just as often. Check which calculation a tool uses before comparing numbers.
Related concepts
- AI visibility (mention rate): the share of relevant answers in which a brand appears at all. It is absolute; share of voice is relative. Both can move independently, which is why they are tracked together.
- AI citation share: the share of cited sources in AI answers that point to a brand's own domain. Citations and mentions are counted separately: an engine can name a brand without linking to it, or link to a page without naming the brand.
- AI recommendation rate: how often an engine actively recommends the brand rather than listing it. A high share of voice built on passing mentions is weaker than a lower share built on recommendations.
- AI answer sentiment: the tone of those mentions. Share of voice counts them; sentiment qualifies them.
Frequently asked questions
There is no universal benchmark. The figure depends on how many competitors are tracked: with five brands in the set, 20% is parity. HubSpot's guidance for share of voice generally is that 20 to 30% signals a strong position in a competitive category, and that a share above actual market share tends to precede growth. The useful reading is the trend against your own baseline and the gap to the category leader.
Because the inputs differ: prompt sets, engines, countries, sampling frequency, whether the tool counts share of mentions or share of answers, whether citations are included, and whether positions are weighted. Each number is valid under its own definition, but none is comparable with another tool's. Pick one method, document it, and compare the figure only against itself over time.
Usually not. Most tools report mention-based share of voice and a separate citation share, because the two measure different things: mentions show whether an engine knows and names the brand, citations show whether it uses the brand's content as a source. Reporting them together, with recommendation rate and sentiment, gives a fuller picture than any single percentage.