AI Share of Voice: Measure It on ChatGPT, Gemini, Perplexity

The short version:

• AI share of voice is the percentage of brand mentions that go to your brand when ChatGPT, Gemini and Perplexity answer your buyers’ questions.
• The formula: your brand’s mentions / total mentions of all brands x 100, on the same set of questions.
• It is a relative metric, built for comparison: a 20% share of voice dominates a market where fifteen brands split the mentions, and means little against a duopoly.
• The raw rate hides three traps: the position of each mention, run-to-run variance, and cross-engine averages.
• You can count it by hand in one to two hours, or see the ranking of brands actually mentioned on your questions in one minute with the free audit.

What is AI share of voice?

AI share of voice is the percentage of all brand mentions that go to your brand in ChatGPT, Gemini and Perplexity answers, on the questions buyers ask about your market. It answers the question visibility alone leaves open: who do the engines actually recommend, you or your competitors?

The formula fits on one line:

Share of voice = your brand’s mentions / total mentions of all brands x 100, computed on the same questions for every brand.

An example: you ask an engine 10 buyer questions, and its answers contain 55 brand mentions in total, 12 of them yours. Your share of voice is 12 out of 55, or 22%. This is exactly the formula Pythie applies in its Competitors tab.

Don’t confuse share of voice with mention rate, the share of answers where your brand appears. A brand can appear in 8 answers out of 10 and still be marginal if every answer names a dozen competitors ahead of it. Mention rate measures your presence; share of voice measures your weight.

Why this metric matters in 2026

Because the recommendation now forms inside the answer, before any click. The Pew Research Center measured, across 68,879 Google searches from a panel of 900 US adults in March 2025, that when an AI summary appears, only 8% of users click a classic result, versus 15% without one. The split of mentions decides who captures that intent.

Shopping behavior has already shifted. According to Adobe’s survey of more than 5,000 US consumers, 39% have already used generative AI for their online shopping and 53% planned to do so in 2025. A growing share of purchase research now starts inside an AI conversation, where a handful of named brands make the shortlist.

The clicks that remain are worth more. According to Semrush’s July 2025 study, a visitor coming from AI search is worth on average 4.4 times a visitor from classic organic search, based on conversion rate. Fewer clicks, but far more decided ones: being the recommended brand has never weighed so much.

How do you measure AI share of voice manually?

The protocol fits in four rules: brand-free questions, several runs per question, a tally of every brand mentioned (not just yours), and the same denominator for everyone. Plan one to two hours for three engines, to be redone at every measurement.

1. Reuse the measurement hygiene from our audit protocol: 10 to 15 brand-free buyer questions, fresh conversations, web search enabled, at least two runs per question, each engine separately.
2. Write down every brand named in every answer, with its position (1st, 2nd, 3rd...): this exhaustive tally is what sets share of voice apart from simply tracking your own brand.
3. Add up the mentions per brand and divide each total by the grand total: every brand gets its share, on the same denominator.

Your tally sheet looks like this, for one engine:

The reading is immediate: your brand exists (22%), but competitor A carries nearly twice your weight in the answers. This is the tally the free audit automates: 10 questions, 3 engines, the ranking of the brands actually mentioned on your questions and the real answer excerpts, in one minute, no account needed.

The three traps of the raw rate

A raw share of voice treats a mention in 8th position like a mention at the top, a single run like a stable truth, and three engines like one. These three shortcuts produce wrong conclusions. Here is how to neutralize them, whether you count by hand or with a tool.

Position in the answer matters

Being named 8th in a list is not worth being named first: buyers read the first recommendation, rarely the eighth. That is why Pythie’s visibility score weights every mention by its position: full credit in 1st position, 10% less credit per rank, with a floor at 50% from the 6th position. By hand, note at least each brand’s average position next to its mention count: two equal shares of voice can hide two opposite realities.

Run-to-run variance

The same question asked twice on the same engine can return two different brand lists. A share of voice computed on a single run measures one draw, not a trend. The fix: several runs per question, aggregated. Pythie’s weekly scan applies this principle: 15 questions x 3 engines x 2 runs, or 90 answers analyzed every week to smooth that variability.

The single cross-engine average

A share of voice averaged “across all AIs” hides opposite realities: we measured the same brand at 100% visibility on Gemini and 30% on ChatGPT, a gap detailed in our AI visibility article. Compute one share of voice per engine, never a global average.

Share of voice or visibility score: which one to track?

Both, because they don’t answer the same question. The visibility score measures your absolute presence: the share of answers where your brand appears, weighted by the position of each mention. Share of voice measures your competitive position: your relative weight against every brand the engines name on the same questions.

The two metrics can diverge, and that divergence is precisely their value. Your score can climb while your share of voice drops: you are improving, but a competitor is improving faster. Conversely, a comfortable share of voice can coexist with a mediocre score, when the engines name very few brands on your questions: you dominate a debate that barely happens.

In practice, the score tells you whether you exist in the answers; share of voice tells you whether you win the comparison. Track the first as a health indicator and the second as a battle indicator: it names the competitor to catch and the gap to close.

To see both at once, the free audit asks ChatGPT, Gemini and Perplexity 10 questions from your market, no account or credit card, and shows the ranking of the brands actually mentioned. The subscription then recomputes that ranking every week, engine by engine, in the Competitors tab.

Frequently asked questions

What is a good AI share of voice?

There is no universal benchmark: it depends on how many brands the engines mention on your questions. In a market where three brands concentrate the mentions, 25% is weak; where fifteen brands split the answers, 25% dominates. The useful reference is relative: your gap to the leader, engine by engine, and its trend over time.

How often should you measure AI share of voice?

Weekly, for real tracking. Rankings move as the models and their sources are updated: a one-off measurement goes stale within weeks. A weekly rhythm with several runs per question separates a real trend from measurement noise; a one-off audit remains a good starting point.

Does AI share of voice replace SEO?

No, it complements it. According to Similarweb, AI platforms drove more than 1.13 billion referred visits to the world’s top 1,000 websites in June 2025 (up 357% year over year), against 191 billion from Google Search over the same period. SEO still dominates the volume; AI answers concentrate the growth. Track both.