What Is Share of Voice (AI)?

How AI Share of Voice Works

Share of voice is a decades-old marketing metric: your slice of the total advertising, mentions, or attention in a category. AI Share of Voice ports that idea into generative search. A generative engine answers a question by retrieving a handful of documents and synthesizing them into one response, and along the way it names some brands and cites some sources. Share of voice asks: across a set of those answers, how much of the brand visibility is yours versus everyone else’s?

The measurement has three moving parts. First, a prompt set — the questions you choose to run, ideally the ones your buyers actually type. Second, a competitor set — the rivals you are measuring against, since share is meaningless without a denominator. Third, a counting rule — what you tally in each answer. Most implementations count both brand mentions and citations, and many weight them by prominence: being the first recommended tool counts for more than a name buried in a caveat at the end.

Two properties make it behave unlike a search ranking. It is prompt-set dependent, so the same brand scores differently on “best CRM for startups” than on “cheapest CRM.” And it is engine-specific: the same query pulls different grounding sources and names different brands on ChatGPT, Perplexity, and Google AI Overviews, so a share computed on one engine does not transfer to another. Reported together across engines, share of voice rolls up into overall AI visibility.

Formula

There are two common ways to express AI Share of Voice, and they answer slightly different questions.

The presence-rate form measures how often you show up at all:

Presence SoV (%) = (Prompts where your brand appears ÷ Total prompts measured) × 100

The competitive-share form measures your slice relative to rivals, and is the truer analogue of classic share of voice:

Competitive SoV (%) = (Your weighted mentions ÷ All brands’ weighted mentions) × 100

The pieces that make either number meaningful:

• Weighting — a prominence multiplier. A common scheme scores the top recommendation highest, a mid-answer mention lower, and a passing reference lowest, so “recommended first” and “mentioned once” are not treated as equal.
• Scope — the prompt set. The figure is only comparable against the same list of questions over time.
• Competitor set — the denominator in the competitive form. Add or drop a rival and your share shifts even though your own visibility did not change.
• Engine — computed and reported separately per platform, because brand visibility does not carry over between them.

Example of AI Share of Voice

A documented worked example comes from a January 2026 study by Amit Sharma, who tested 112 startups from Product Hunt with 2,240 queries split into two types: direct questions (“What is ProductX?”) and discovery questions (“What are the best tools for X?”). Because every brand was run against the same fixed prompt set on the same two engines, the discovery-query results are a clean presence-rate share-of-voice measurement.

The gap between recognition and share was stark. On direct questions, the engines knew the products almost every time — Perplexity recognized them 94.3% of the time and ChatGPT roughly 99%. But on discovery questions, where share of voice actually lives, the brands surfaced far less: Perplexity recommended them in only about 8.3% of category prompts and ChatGPT in only about 3.3%. Same brands, same questions, but Perplexity handed out more than double the share of voice ChatGPT did — the platform-specific behavior the metric is built to expose.

The study also probed what moved share of voice. The strongest correlate of Perplexity visibility was Reddit mentions (about +0.40), followed by referring domains (about +0.32) — off-site authority signals, not on-page tweaks. That single-study finding is not the last word, but it lines up with the broader pattern that engines lean on sources they already trust when deciding which brands to name.

For a second reference point, AthenaHQ’s State of AI Search 2026 report put the average brand mention rate at 17.2% across the queries it tracked, meaning a typical brand appears in fewer than one in five relevant answers. Read against that baseline, the Sharma discovery rates are a reminder that most brands hold a thin share of voice in generative search, and that the leaders are the few who show up on the questions that matter.

Share of Voice vs Citation Share

Share of voice and citation share are close cousins, and teams routinely conflate them. The clean way to hold them apart: share of voice is the wider measure of visibility, citation share is the narrower measure of attribution nested inside it.

The relationship is a hierarchy. Every citation is a form of visibility, so it feeds share of voice — but not every mention is a citation. A brand an AI names as the recommended option without linking anyone’s page has earned share of voice with zero citation share; a brand whose page is cited as the source has earned both. That is why the two numbers diverge: share of voice tells you whether the machine is talking about you, and citation share tells you whether it is sourcing you. Track share of voice to see if you are in the conversation at all, then read citation share to see whether you are the authority the answer was actually built on.

Frequently Asked Questions

The Bottom Line

AI Share of Voice is the old marketing question — what fraction of the conversation is about us — rebuilt for answers a machine synthesizes instead of a market you buy into. It rewards brands that show up across the prompts buyers actually ask, weighted toward being the recommended pick rather than a footnote. Scope it to real questions and real rivals, read it per engine, and treat citation share as the stricter sibling that tells you whether the machine is willing to source you, not just name you.