Ask the assistants your buyers use the questions your buyers ask, and count how often your brand appears in the answers. That fraction - across a stable prompt set, across models, over time - is your share of voice in AI answers. It is the cleanest single number for a channel most teams still measure with anecdotes.
The metric
Presence rate per model, then a weighted blend across models by your audience's usage. Track alongside it: how you are described (the models' one-line positioning of you), who else is named (your real answer-set competitors), and which sources get linked (where authority is flowing from).
How to measure it honestly
- Fix the prompt set and version it; changing questions changes the number.
- Sample daily or weekly; single checks are lottery tickets.
- Vary phrasings per intent to smooth model volatility.
- Keep raw answers so any month's number can be re-derived and audited.
What actually moves it
Three levers, in effort order: citability (crawlable, extractable, definition-first pages), specificity (numbers, comparisons and claims only you can make), and corroboration (presence in the review sites, lists and threads the models keep citing for your category). Move all three and the trend follows within weeks; the AstroFabric GEO autopilot exists to run exactly that loop on schedule.
Every playbook on this blog ships as a runnable mission.
Open a free workspace and the playbook library is waiting - describe the outcome and the agents carry it end to end, with 1,000 trial credits on us.