What is share of voice in AI answers?
Share of voice is your portion of all the business mentions across a set of AI answers: your mentions measured against every other business named, over a stated window, and published with the counts behind it.
- Share of voice
- Your mentions set against all the business mentions in the same set of answers, over a stated window.
- Mention set
- Every business named across the answers in the window, not a competitor list you supplied. The engines decide who is in the room.
- Evidence window
- The stretch of checks the figure is calculated over. Apex shows no share-of-voice figure at all until the window behind it is long enough that the number holds still.
Why the number stays hidden at first
Early on, share of voice is a ratio of two small numbers, and small numbers move violently. One assistant phrasing a sentence differently swings your share hard in either direction, for no reason connected to your business. Show that to an agency and they will either celebrate noise or panic at it, and either way it ends up in a client report.
So Apex does not show it. Until enough checks stand behind the figure for it to hold still, the share-of-voice panel says exactly that, in those words, instead of showing a number. Appearance rate is visible from the first result, because it carries its own denominator and stays honest at any size.
How it differs from appearance rate
Appearance rate, the number Apex publishes as your AI Visibility Score, measures you against the questions: of the checks we ran, how many named you. Share of voice measures you against everyone else: of all the names the engines produced, how many were yours. They answer different questions and they move independently.
You can hold a strong appearance rate and a weak share of voice. That is what it looks like when the engines name you and a crowd of competitors in the same breath, every single time. You are always in the answer and never the answer.
Who ends up in the set
The comparison set is whoever the engines actually name, which is regularly not the list of competitors you would have written down. Local operators find national aggregators sitting in their set. Agencies find that a directory page is being treated as a better answer than their client. That surprise is most of the value of the metric.
What moves it
- Appearance rate. Being named more often is the input; share of voice is the consequence.
- The answers where engines currently list many businesses: those are the ones diluting your share, and they are where a shorter list would help you most.
- Corroboration on the sources the engines lean on for your area, which is what gets you into a shortlist rather than a long one.
- Time. This is a trend metric: it settles as the window behind it lengthens, and it is worth nothing at all read off a single morning.
Questions people ask
Why can I not see my share of voice yet?
Because there are not yet enough checks behind it for the figure to hold still. It exists internally and is not stable enough to report; making you wait is a better product decision than handing you a number that will halve tomorrow.
Is share of voice better than appearance rate?
Neither is better. They answer different questions. Appearance rate tells you whether you are in the answer. Share of voice tells you how crowded the answer is when you are.
Can share of voice fall while appearance rate rises?
Yes, and it is common. If the engines start naming more businesses per answer, everyone's share falls even as the raw number of mentions rises. That is exactly why the two are reported side by side.
What window should I judge it over?
Long enough that one unusual morning cannot move it, and longer still before you act on the direction it is pointing. Anything shorter is a snapshot of a system that is non-deterministic by design.