Glossary
AI share of voice: your slice of the answers, measured on a sample you define
AI share of voice is your share of the brand mentions or citations across a defined set of AI answers, compared with competitors. How it is calculated, why two reports on the same brand disagree, and what a trustworthy figure has to state.
Want to know where your own site stands? Run a free Growth Scan
What AI share of voice measures
AI share of voice is the share of brand mentions, or citations, that go to your business across a defined set of AI answers, compared with the competitors you track. It borrows the name from advertising, where share of voice meant your share of a category’s advertising, and from SEO tools, which use it for your share of estimated visibility across tracked keywords.
In AI search there is no fixed results page to count. Answers are generated, vary between runs and depend on where and how the question was asked. So AI share of voice is always a sample: a chosen set of prompts, run on chosen platforms, at chosen times.
How the figure is built
Fix a prompt set
Write the questions real buyers ask about your category, such as the best provider in a city or how two options compare, drawn from search data, sales calls and support questions rather than from what flatters the brand.
You get: A versioned list of prompts that stays fixed between reports
Choose platforms and locations
Decide which assistants and AI search features to sample, and from which country or city, because answers differ by both.
Run every prompt several times
One run is an anecdote. Repeating each prompt shows how stable the answer is and turns a yes or no into a rate.
Record who is mentioned and cited
For each answer, note which tracked brands are named and which pages are linked as sources. Keep mentions and citations separate.
Divide, and say what you divided by
Your mentions divided by all tracked brands’ mentions gives share of voice among competitors. Answers that mention you divided by all answers gives a mention rate. State which one the headline figure is.
One set of answers, two valid numbers
An illustration, with invented numbers for the arithmetic only. Forty prompts, each run three times, give 120 answers. Across them, your business is mentioned 24 times, competitor A 48 times and competitor B 28 times.
Share of voice among the three tracked brands is 24 out of 100 mentions: 24%. The share of answers that mention you at all is 24 out of 120: 20%. Both describe the same answers correctly, and a report that quotes one without saying which is inviting a wrong comparison.
Add a fourth competitor to the tracked list and the first figure falls, although not a single answer changed. That is why the list of brands tracked belongs in the method too.
Why two reports on the same brand disagree
- Different prompt sets
- A prompt set built from branded questions will always flatter the brand. One built from open category questions is harder and more useful. Neither is comparable with the other.
- Different denominators
- Share among tracked brands and share of all answers can differ widely on the same data, as the illustration above shows.
- One run against several
- A single run per prompt captures one draw from a variable answer. Repeated runs give a rate that moves less between reports.
- Different locations and settings
- Where a prompt is run from, and whether the assistant searched the web or answered from its training, changes which businesses appear.
How BOOSTD is built to report it
AI visibility is one of the eleven categories in the BOOSTD Growth Score, and the methodology marks it as sampled: every figure is reported with its prompt set, platforms, dates and the counting rule used, and mentions and citations are kept apart.
BOOSTD is in early access. Prompt sets, answer collection and citation extraction are built and tested against simulated answers; no assistant has yet been queried live for a customer. The free scan does not measure AI visibility, because each sampled answer costs money to collect, and it says so rather than guessing.
Questions about AI share of voice
Is a higher AI share of voice always better?
Usually, but look at the answers behind it. A mention that describes your business inaccurately or recommends a competitor over you counts the same as a good one in a simple share. Read a sample of the answers, not only the percentage.
How many prompts are enough?
Enough to cover the real questions buyers ask about your category, run several times each. A handful of prompts produces a number that swings with every run. The right set is the one you can keep fixed, so that change over time means something.
References
Sources
The primary documents and published research this page relies on. Platform rules change, so check the source before acting on a detail.
Last updated · Published by Zubair Afzal (responsible editor), on owner authorisation · Reviewed quarterly — next review: