AI share of voice

AI share of voice (AI SOV) is a competitive metric that measures a brand's share of all brand mentions in AI-generated answers to a set of prompts, compared with its competitors. It adapts advertising's share of voice to assistants such as ChatGPT, Gemini, Microsoft Copilot and Google AI Mode.

Also called: Share of model

Why AI share of voice matters

AI share of voice matters because an AI answer is a short list. When a buyer asks an assistant for the best tool in your category, it names a handful of brands, and the buyer may never look further. The question is no longer only whether you are named, but how much of the conversation you get compared with your rivals.

Marketers already trust that logic in advertising. LinkedIn's B2B Institute states the classic rule this way: "Brands that set their share of voice (SOV) above their share of market (SOM) tend to grow (all other factors being equal), and those that set SOV below SOM tend to shrink." AI share of voice borrows the idea for answers. One caution: that rule was measured on advertising spend, not on AI answers, so treat AI share of voice as a snapshot of your competitive position, not a sales forecast.

Share of voice in classic marketing

Share of voice started as a media-spend metric. Nielsen defines it as a brand's media spending "expressed as a percentage of all media expenditures in the category, in that market, on that channel and at that particular point in time." It measures whether a campaign has the means to compete, not whether it worked.

The metric earned its reputation through the gap between share of voice and market share, called extra (or excess) share of voice. Nielsen reports that Binet and Field studied 171 campaigns between 1980 and 2010 and found that market share typically gained 0.5% for every 10 points of excess share of voice. LinkedIn's B2B Institute, working with the same researchers, measured 0.7% of market share growth a year per 10 points in B2B, against 0.6% in B2C.

PR and social listening teams later stretched the term to mean a brand's share of mentions or conversations. AI share of voice comes from that second meaning: it counts mentions in organic answers, not money spent.

How AI share of voice works

AI share of voice is calculated from a sample of answers, in five steps:

Fix the competitive set. List the brands you compare: yours and your main competitors. This list is the denominator.
Choose unbranded prompts. Use the questions buyers ask about the category, such as "best CRM for a small agency", not questions that already name your brand.
Run them repeatedly on each engine. Answers change from one run to the next, so one sample proves little.
Count the mentions. For each answer, record which tracked brands it names (each one is a brand mention).
Divide. Each brand's mentions over the total mentions of all tracked brands.

There is no standard. Some tools count every occurrence of a name, some weight mentions by position, some add citations. The Harvard Business Review article on "Share of Model" assesses it through three lenses: "mention rate, human-AI awareness gap, and brand and category sentiment."

Formula

A simple, common version counts one mention per answer that names the brand:

AI share of voice = mentions of your brand ÷ mentions of all tracked brands × 100

Worked example (fictional values). You track 50 answers about project management software for agencies, against three competitors. Your brand is named in 20 answers, Brand B in 35, Brand C in 25 and Brand D in 10, for 90 mentions in total.

Your share of voice: 20 ÷ 90 = 22.2%
Brand B: 35 ÷ 90 = 38.9%
Brand C: 25 ÷ 90 = 27.8%
Brand D: 10 ÷ 90 = 11.1%

Now add a fifth brand, named in 30 of the same answers. The total becomes 120 mentions and your share falls to 16.7%, although not a single answer changed. Your mention rate stays at 40% (20 answers out of 50). This is the main trap of AI share of voice: it moves with the competitor list, so compare figures only when the list is identical.

AI share of voice vs mention rate

AI share of voice and mention rate come from the same answers but divide by different things.

	AI share of voice	Mention rate
Calculation	Your mentions ÷ all tracked brands' mentions	Answers naming you ÷ all answers
Sum across brands	Always 100%	Can exceed 100%
Adding a competitor	Lowers your share	Changes nothing
Best for	Comparing your slice of the conversation	Following your own presence over time

How to measure and improve AI share of voice

Measure AI share of voice on a stable base: the same prompts, the same engines and the same competitor list from one period to the next. Run each prompt many times. SparkToro found that the length of AI recommendation lists varies a lot: "sometimes the AI gave as few as 2-3 recommendations, and equally often 10 or more". Longer lists add mentions for everyone and shift the denominator.

Hawkscan tracks your share of voice against your competitors on the same prompts and engines: every tracked brand, yours included, is ranked by visibility score, mention rate or average position, and a positioning map places each one by mention rate and average position, so you see who takes the space you are missing.

To grow your share, start with the prompts where competitors are named and you are not. Read those answers, find the pages the engines cite (see AI citation), and work to be covered on them. Then follow the trend over several weeks rather than day to day.

Common questions

What is share of model?

Share of model is another name for AI share of voice. Harvard Business Review describes it as "a new concept that helps marketers gauge how prominently their brand appears before consumers in LLM searches." Its version is broader than a mention count, since it also looks at sentiment and at the gap between how well people and AI models know a brand.

What is a good AI share of voice?

A good AI share of voice depends on how many brands share your category's answers. With two competitors, 30% is below an even split; in a crowded market with dozens of names, 10% can lead. A useful reference point is your market share: if your share of the answers sits well below it, assistants under-represent you.

Should you track AI share of voice per engine?

Yes. The same prompts can split very differently from one engine to another. In Chen et al.'s test of unbranded cola prompts, major brands took 56.3% of the brand mentions in ChatGPT's answers and 67.9% in Perplexity's. One blended figure can hide a rival that dominates a single engine, so compare engines separately, on the same prompts and competitor list.