AI share of voice: the slice of citations your brand holds
AI share of voice compares a brand's citations with those of every brand cited in the same AI answers. The measurement uses a fixed set of target questions and complements citation rate, which measures presence, with a measure of relative position.
What is AI share of voice?
AI share of voice is the slice of citations a brand holds across a fixed set of target questions, compared with every brand cited in the same answers. It measures how much of the recommendation space went to the brand, by engine and by period.
The question set is what gives the number meaning. It is defined in the diagnostic, from the questions the company's customers ask assistants, and frozen before the first measurement. Change the set and the number changes, so any change is disclosed in the report.
Each engine produces its own share of voice. ChatGPT, Gemini, Claude, Perplexity and Copilot retrieve sources from different indexes, and the same brand can hold a large slice in one engine and almost none in another. The consolidated figure adds up the runs from all of them.
How is it calculated?
Divide the brand's citations by the sum of the brand's citations and competitor mentions, over the same set of questions and the same period. With 20 brand citations and 380 competitor mentions, share of voice is 20 ÷ 400, or 5%.
| Scenario (100 runs) | Brand citations | Competitor mentions | Share of voice |
|---|---|---|---|
| A | 20 | 380 | 20 ÷ 400 = 5% |
| B | 20 | 20 | 20 ÷ 40 = 50% |
Source: ReBo AI Visibility baseline protocol. Numerical example, no client data.
Both scenarios have the same citation rate, 20 citations in 100 runs, or 20%. In scenario A, the brand appears surrounded by 19 competitor mentions for each of its citations. In scenario B, it appears next to one competitor mention per citation.
The math only works if every competitor is recorded in each answer. A missed competitor inflates the brand's share of voice, so each run's entry carries the full list of names. The recording protocol is in the AI citation test.
How does it differ from citation rate?
Citation rate measures in how many runs the brand appears: citations divided by runs. Share of voice measures relative position: how much of all brand citations went to it. The two metrics move independently and have to be read together.
A company can hold its citation rate steady and lose share of voice, because competitors started being cited more often in the same answers. The reverse also happens: the whole category becomes more visible and the rate rises while relative position stays put.
In a growing market, share of voice can fall while citation rate rises. The brand appears more, and so do its competitors. Read alone, the drop looks like failure; read next to the rate, it shows a category that is growing.
What share of voice does not prove: management quality, investor preference or purchase intent. It measures comparative presence within a question set defined by the measurement.
Related terms
It complements citation rate and citation gap, and is measured against the AI Visibility baseline. How both metrics evolve helps place the company in the 4-level framework.
Frequently asked questions
What is a good AI share of voice?
There is no universal number. The value depends on how many competitors the sector has and how many brands fit in one answer. What matters is the time series against the baseline and the comparison with direct competitors on the questions with the most commercial value.
Does share of voice change by engine?
It changes a lot, because each engine retrieves from different indexes and weighs sources differently. That is why the metric is always reported by engine, alongside the consolidated figure for the period.
Can you measure share of voice without naming competitors?
Yes. In ReBo's public material, competitors are shown in aggregate. In the client report the list is named, because that list guides production in the next cycle.