Share of voice in AI answers: definition, calculation, how to read it
Benjamin Libor10 min read
Share of voice in AI answers is your company's share of all the brand mentions in the answers to your prompt set. If the answers to twenty prompts name brands 36 times in total and nine of those are you, your share of voice is 25 per cent. It puts you and your competitors on one scale, and it is read best alongside visibility (how many answers name you at all) and position (where in the answer you appear).
In short
- Share of voice is your mentions divided by all brand mentions across the answers to your prompt set, with each brand counted once per answer.
- Visibility is the share of answers that name you; position is where in the answer you appear. The three move for different reasons and answer different questions.
- Compute it per engine and per market as well as overall, because an average across engines hides the one that changed.
- It is noisy: few prompts, few runs and one engine's update all move it. Steady it with repeats, a fixed prompt set and rolling four-week windows.
- The target is a gain against named competitors on the prompts you chose, not an absolute number.
The definition
Share of voice (AI answers): of all the brand mentions in the answers to a fixed set of prompts over a period, the share that are your brand. Each brand counts once per answer, whether it is named once or five times.
Two choices in that definition matter. First, which brands count: every brand named in the answers, or a fixed list of you plus named competitors. The fixed list is more stable and easier to explain; the open one is more honest about new entrants. Pick one and keep it. Second, once per answer. An answer that mentions a competitor six times in a long comparison is not six times the signal; it is one answer that names them. Counting once per answer keeps a single verbose answer from swinging the month.
The calculation, worked
Fernwick, a scheduling tool, tracks a prompt set and four named competitors. One week, twenty answers come back. The team records which brands each answer names, once per brand per answer, and totals them.
| Brand | Answers that name it (of 20) | Share of voice |
|---|---|---|
| Slotwise | 14 | 38.9% |
| Fernwick | 9 | 25.0% |
| Bookline | 7 | 19.4% |
| Calmeet | 4 | 11.1% |
| Tidyslot | 2 | 5.6% |
| Total mentions | 36 | 100% |
Share of voice for each brand is its mentions divided by the total: Fernwick is 9 ÷ 36, or 25 per cent. The shares add to 100, which is the point: this is a measure of the room, and when someone gains, someone else loses.
The same table gives visibility: Fernwick was named in 9 of 20 answers, so its visibility is 45 per cent, and Slotwise's is 70. Visibility is a share of answers; share of voice is a share of mentions. They look alike in a small table and behave differently over time.
Share of voice, visibility and position
Three numbers, each from the same answers, each answering a different question.
| Measure | Question it answers | Denominator | Moves when |
|---|---|---|---|
| Visibility | How often do the answers name us at all? | Answers | We are added to, or dropped from, answers |
| Share of voice | Of the brands named, how much of the room is ours? | Brand mentions | We or any competitor gain or lose mentions |
| Position | When we are named, how early? | Our mentions | The order inside answers changes |
They move differently because their denominators do.
Visibility can rise while share of voice falls. Say the assistants start answering "best scheduling tool" with longer lists. Fernwick now appears in 12 of 20 answers instead of 9, but every answer names six brands instead of three. Fernwick's visibility rose from 45 to 60 per cent; its share of voice fell, because the total mentions grew faster than Fernwick's. Visibility says you are in more answers. Share of voice says the answers got more crowded.
Share of voice can rise while nothing happened to you. A competitor drops out of the answers (a stale list was updated, a product was discontinued) and the remaining brands' shares all rise. Your visibility is unchanged. Share of voice told you the room changed; visibility told you that you did not.
Position moves on its own. You can be named in the same number of answers, with the same share of voice, and slip from first to fourth in each. For "best X" prompts, where the reader stops after two or three names, that is the change that costs you, and neither of the other two shows it (what is AI visibility covers position).
Read all three together. Share of voice up with flat visibility is a competitor's loss, not your gain. Visibility up with flat share of voice is a more crowded answer. Both up, with position holding, is the one you want.
Compute it per engine and per market
An overall share of voice is for the board. For the work, split it.
Per engine. ChatGPT, Perplexity, Gemini, Claude and Google's AI Overviews search differently, cite different pages and name different brands. A brand can hold a strong share on Perplexity and a weak one on AI Overviews because the pages each engine leans on differ, and an overall figure averages that away. A change in one engine, which is the usual kind, shows only in that engine's line.
Per market. If you sell in more than one country or language, run the prompts in each and keep the shares apart. The competitors differ, the cited pages differ, and a share of voice that mixes the German and the US answers describes neither.
Per prompt group. "Best X", "X vs Y" and "how do I" prompts behave differently, and the competitors named in each differ. A share per group points at which pages to work on; how to choose the prompts to track explains grouping.
What makes it noisy, and how to steady it
Share of voice from a small prompt set in one week is a sample, and behaves like one.
- Sampling. The same prompt to the same engine returns different answers on different days, naming different brands. One run per prompt per week is a coin toss on the margins.
- Few prompts. With ten prompts, one answer swinging moves a brand's share by several points.
- One engine changing. A model update or a change in how an engine searches can shift every brand's share in that engine overnight, and an overall figure moves with it.
Four habits make it readable.
- Repeat each prompt. Ask every prompt several times per engine per week and count each answer. More answers, less swing.
- Fix the prompt set. Change it rarely, and when you do, note the date and treat the series as new from there. A share of voice from a prompt set that drifts is not a series.
- Use rolling four-week windows. Report the share over the last four weeks, moved forward a week at a time. The week-to-week noise flattens and the trend stays.
- Keep the denominator fixed. Name the competitors in it and leave them there for the quarter, with new ones added at the quarter's start.
With those in place, a move of a few points over a month is usually real; the same move in a week usually is not.
How to read a change
When share of voice moves, go to the answers and the citations before the explanation. The usual causes are findable.
- A competitor launched a comparison page. Their "Slotwise vs Fernwick" page gets cited for your "vs" prompts and their mentions rise. You will see it as a new citation on your competitor-owned list and a rise in their share on the "vs" group. The response is your own comparison page (how AI uses "best X" lists and comparison pages).
- A listicle was updated. The "12 best scheduling tools" page that gets cited across nine prompts added a brand, dropped one, or reordered. One page, and the shares of every brand on it move at once. Find it by sorting citations by prompts covered.
- A model or search update. Every brand's share in one engine shifts on the same day, and the citations for that engine change character. Nothing on your side caused it, and nothing on your side fixes it; note the date and compare from there.
- Your own change landed. A page you published or a listing you completed starts being cited, and your mentions rise in the prompts it covers. This is the case you want to prove, so record the date the page went live and watch those prompts specifically.
The habit: the change, then the engine and prompt group it came from, then the citation that explains it, then the action. Echo's AI visibility page walks that path in that order; it is the same path in a spreadsheet.
What a sensible target is
Not an absolute number. A share of voice of 25 per cent is excellent in a category with ten active brands and poor in a category with two. The targets that work are relative and scoped.
Gain against the named competitors, on the prompts you chose, over a quarter.
For example: "Pass Bookline on the 'best X for agencies' prompt group across all engines by the end of the quarter, and hold position two or better on Perplexity." That names the competitor, the prompt group, the engine and the time, and each can be checked. "Reach 30 per cent" names none of them and can be hit by a competitor's collapse.
Start with the prompt groups where you are closest to a named competitor and the cited pages are ones you can act on.
Share of voice in the monthly report
One chart and one sentence.
The chart: a line per named brand, share of voice on the vertical axis, the last six months on the horizontal, rolling four-week windows, overall or for the engine that matters most. Five lines at most. If engines diverge enough to matter, a small table under the chart: one row per engine, one column per brand, for the current window.
The sentence: the change, the cause and the response. "Fernwick's share of voice rose from 21 to 25 per cent over the quarter, mostly on Perplexity, after the review site listing was completed and two listicles added us; Slotwise remains first at 39 per cent, and their new comparison page is the next gap to close."
That sentence takes the reader from the number to the next action, which is what the number is for. Keep visibility and position in the appendix, and keep the method note (prompt set size, engines, window, denominator) the same every month so the series is trusted. An AI visibility routine for a two-person team fits this report into a monthly rhythm.
Questions people ask
What is share of voice in AI search?
Share of voice in AI search is your brand's share of all the brand mentions in the answers AI assistants give to a fixed set of prompts over a period. If the answers name brands 36 times and nine of those are you, your share of voice is 25 per cent. Each brand is counted once per answer, and the shares across brands add to 100 per cent.
How is share of voice different from AI visibility?
Visibility is the share of answers that name you at all; share of voice is your share of all brand mentions in those answers. Visibility can rise while share of voice falls if answers start naming more brands, and share of voice can rise with no change to you if a competitor drops out. Read them together.
What is a good share of voice in AI answers?
There is no absolute good number; it depends on how many brands the answers name in your category. A sensible target is relative: gain against a named competitor on a chosen prompt group, in a chosen engine, over a quarter, and hold your position in the answers while doing it.
Written by Benjamin Libor, founder, echo.
Something wrong or out of date? Write to hello@echo-aeo.com and we'll fix it.