AI share of voice puts a number on a simple question: How does your brand compare with its competitors in AI answers? For example, out of 100 answers to the same questions, 30 name your brand, while competitor A appears in 50, competitor B in 40, and competitor C in 20. Together, the four brands receive 140 mentions; 30 of them go to your brand, a share of about 21%.

This sets AI share of voice apart from mention rate, which looks at a brand on its own: it states what percentage of answers name the brand—in the example, 30 of 100, or 30%. Only the comparison shows whether 30% is a lot or a little in a given competitive field.

Where the term comes from

Share of voice is an advertising term. There, it describes a brand’s share of all advertising in its product category, usually measured by spend. In a 1990 Harvard Business Review article, for example, advertising scholar John Philip Jones of Syracuse University proposed setting advertising budgets by comparing a brand’s market share with its share of voice, which he defined as the value of the brand’s main advertising placements relative to all such placements in its product category.

In GEO, the idea is applied to AI answers. Instead of paid advertising, what counts is how often a brand is named in the answers that AI systems such as ChatGPT, Gemini, Claude, and Perplexity write themselves, compared with its competitors.

Mentions or citations

AI share of voice can be measured in two ways. One form counts mentions—whether an answer names a brand in its text. A variant, also called citation share, counts source references instead: What share of the AI citations in the answers points to your website? This share is measured either against all cited sources or only against the citations that go to your website and your competitors’ websites. The share of mentions and the share of citations can differ widely. A brand can be named often while the answers mostly cite competitors’ pages, industry portals, or comparison sites as sources.

In June 2026, Microsoft began rolling out such a share, called Citation Share, as a preview in Bing Webmaster Tools, its tool for website owners. There, the AI Performance report gives a website’s percentage of all citations shown for a specific query—one of the phrases the AI used to retrieve content for its answers. Because the report does not reveal which other sites hold the rest, the value cannot be narrowed to a group of competitors you have chosen. Microsoft explicitly describes it as an observational metric rather than a competitive scoreboard.

Why values from different tools differ

There is no standard formula for AI share of voice. Like every metric derived from AI answers, it depends on which questions the prompt set contains and on which AI systems, with which settings, it is measured. Share of voice adds two decisions that each tool makes for itself:

  • Competitors: The list of competitors determines the total that each brand’s mentions are divided by. If a brand that appears in the answers is added to or removed from the list, every share changes, even if the answers stay the same. A share of all brands that appear in the answers at all, in turn, differs from a share within a fixed list.
  • Counting rules: Whether a brand counts at most once per answer or every time it is named changes the result. It also makes a difference whether all mentions are added up or a share is first calculated for each answer and then averaged. If one answer names your brand and one competitor, and a second names your brand and nine competitors, your brand gets 2 of 12 mentions when they are added up, or about 17%. Averaging the two answers, at 50% and 10%, gives 30% instead.

Two tools can therefore report clearly different values for the same brand without either one being wrong. AI share of voice is most meaningful as a trend within one tool: with the same competitor list, the same prompts, and the same AI systems, measured at regular intervals such as once a month. If the competitor list changes, that belongs in the report, so that a jump in the trend is not mistaken for a success or a setback.

What AI share of voice shows and what it doesn’t

As a single figure, AI share of voice works well in reports, including reports to management: it shows at a glance how present a brand is in AI answers relative to its competitors. In GEO monitoring, it complements metrics such as mention rate and citation rate.

What it doesn’t show is the tone in which a brand is named, whether it is recommended or merely listed, and whether the statements about it are correct. The last of these is measured by brand accuracy. A high share of mentions is therefore only a good sign if the portrayal is accurate as well.