Share of Voice is your brand's mentions as a percentage of all the brand mentions across the prompts you track. Koalr does not run a separate set of queries for each rival: it asks your questions once, and reads every brand named in each answer out of that same answer. The AI Visibility Leaderboard then ranks you against them prompt by prompt, so a loss is a specific question rather than a position.
Your AI Visibility Score in under 5 minutes. No card required.
Every brand in the answer, read from the same answer, on all seven engines.
The comparison is like for like because there is nothing to line up. It is one measurement, read several ways.
Your prompts belong to your site, not to a brand. A run puts each one to each engine, and every answer that comes back is stored once, with the full text and the list of every brand named in it. Your numbers and your competitors' numbers are then derived from that identical text. Nobody was asked a slightly different question, on a different day, from a different country.
Because every name in an answer is extracted, Koalr also finds the brands you are competing with and did not know to list, and surfaces them for you to add. Adding one gives it a row of its own on the leaderboard and its own numbers to watch. In practice this is where the useful entries on a competitor list come from — the ones you already knew about rarely tell you anything.
For each answer naming you, Koalr records whether your name falls in the opening tenth of it, and tracks that share as a trend beside the others. A buyer reading a recommendation reads the top of it, so moving from fifth to first inside answers you were already winning is real progress that a share percentage alone would not show.
One prompt — "best accounting software for a small UK agency" — run on four engines in a single run. You track your own brand and two rivals.
You and both rivals. Three mentions go into the total, one of them yours.
Knowing you hold 18% tells you how you are doing. It does not tell you what to do on Monday.
It ranks you and your tracked competitors across the prompts you share. The individual questions behind a rank are in Prompt Explorer, answer by answer, with what each engine named and what it cited when it named them. That is where the work is: not "we are third", but "on these nine questions a rival is named and we are not".
A prompt you lose is a specific question with a specific page currently winning it, which makes it something Actions can rank and the Content Studio can draft against. The route from the leaderboard to a draft is the point of measuring this at all.
The same question put to seven engines does not produce seven versions of one answer. A rival who dominates one engine can be absent from another, usually because the two are reading different sources. The leaderboard ranks across the six daily engines together, so the per-engine view on the dashboard, Claude included, is where that difference shows — and the difference is the actionable part.
Nothing here counts customers, revenue or traffic. It counts how often, of all the times an engine named a brand in answer to your questions, the brand was you. A category leader with a quiet website can hold a small share of voice, and that gap is the thing worth acting on rather than a measurement error.
The total it is measured against is the brands named in the answers, one slot per brand per answer. When engines begin naming more brands per answer — six where there were three — your slice shrinks even though you are named exactly as often. That is a real change in the answers your buyers read, and the AI Visibility Score beside it is what tells you whether you also lost ground yourself.
If an engine explains a category without recommending anyone, there is no naming to take a share of, so the answer drops out of the calculation entirely. It still counts against your AI Visibility Score. A high share sitting on a low score means the engines are rarely recommending anyone — and that is an opening, not a loss.
There is no minimum sample. One answer naming one brand is a 100% share for that scope, arithmetically correct and worth nothing. Give a new site a few runs before reading the share as a trend, and use the prompt-level view in the meantime, where a single answer is a legitimate thing to look at.
In the dashboard's per-engine matrix, a competitor matched only by a partial name, an acronym or a generic domain is counted only when its domain is also cited in that answer. That rule stops a common word being read as a competitor, and the cost is that some rivals are under-counted there. The headline share and the leaderboard use the full list of brands read out of each answer, so if a competitor looks quieter in the matrix than on the leaderboard, that is usually why.
The leaderboard orders brands by how often each is named, not by share. Two brands on the same rate keep their list order and still get consecutive rank numbers, so the row above is not beating the row below.
Share of Voice and the AI Visibility Leaderboard are on every plan, including Starter. Share is read per engine on all seven; the headline figure and the leaderboard are built from the six that refresh daily. What changes is how wide the field is: 5 tracked competitors per site on Starter, 15 per site on Growth and 20 per site on Business, alongside 25, 40 and 50 tracked prompts per site. Both allowances are counted per site rather than pooled across your account. Full detail is on the pricing page.
Share of Voice is the percentage of brand mentions across your tracked prompts that were you, each brand counted once per answer. If the AI engines filled a hundred brand slots across the answers to your questions and eighteen of those were you, your Share of Voice is 18%. It answers a different question from your AI Visibility Score: the score is how often you were named at all, the share is how much of the naming was yours.
Koalr does not run a separate set of queries for each rival. It puts your tracked prompts to the engines once, and reads every brand named in each answer out of that same answer. Your numbers and your competitors' numbers come from identical text collected at the same moment on the same engine, so the comparison is like for like by construction rather than two reports lined up next to each other.
5 competitors per site on Starter, 15 per site on Growth and 20 per site on Business. The allowance is counted per site rather than drawn from one account-wide pool, so adding a second brand does not take competitor slots away from the first. Koalr also surfaces brands it finds in your answers that you never added, which is usually how the list gets its most useful entries.
The AI Visibility Leaderboard ranks you and your tracked competitors by how often each is named across the prompts you share, built from the six engines that refresh daily; Claude, which refreshes weekly, is shown per engine on the dashboard instead. The individual questions behind a rank are in Prompt Explorer, answer by answer, with what each engine named and cited. An aggregate tells you that you are third; a prompt tells you which question to answer better, and the prompts you lose are what feed Actions.
Usually because the engines started naming more brands per answer. Share of Voice divides the answers that named you by every brand slot across all the answers to your prompts, so when an engine that used to recommend three products begins recommending six, your slice of the total shrinks even if you are named exactly as often as before. Nothing about you changed; the field did. Read the AI Visibility Score alongside it — if that held steady, you did not lose ground, the answers got more crowded.
Yes, and it is a common and important pattern. Share of Voice only counts answers that named somebody. If the engines answer most of your questions without recommending any brand at all, those answers are invisible to the share and still count against your AI Visibility Score. A high share on top of a low score means that when engines do recommend someone it is often you, but they mostly are not recommending anyone.
Almost, and where it differs the difference is deliberate. A brand named plainly in an answer counts the same whoever it is. The gap is in the weak matches, and it applies to the dashboard's per-engine matrix: there, a competitor matched only by a partial name, an acronym or a generic domain is counted only when its domain is also cited in the same answer. The rule stops an ordinary word being read as a brand, and the cost is that some rivals are counted conservatively there. The headline Share of Voice, the AI Visibility Leaderboard and the prompt view use the full list of brands read out of each answer, so a rival can look quieter in the matrix than on the leaderboard.
Scan your site free and get your AI Visibility Score, plus your first ranked fix list, in minutes. No card, no demo call.
One scan, 25 real prompts, 7 AI engines, under five minutes, zero setup.
Chase
%22%20d%3D%22M11.04%2019.32Q12%2021.51%2012%2024q0-2.49.93-4.68.96-2.19%202.58-3.81t3.81-2.55Q21.51%2012%2024%2012q-2.49%200-4.68-.93a12.3%2012.3%200%200%201-3.81-2.58%2012.3%2012.3%200%200%201-2.58-3.81Q12%202.49%2012%200q0%202.49-.96%204.68-.93%202.19-2.55%203.81a12.3%2012.3%200%200%201-3.81%202.58Q2.49%2012%200%2012q2.49%200%204.68.96%202.19.93%203.81%202.55t2.55%203.81%22%2F%3E%3C%2Fsvg%3E)
GB
