What is AI share of voice (AI SOV)?

What AI SOV measures

Traditional share of voice measures your presence in a medium you bought: your impressions as a proportion of category impressions. AI SOV measures presence in answers you did not buy — how often an assistant names you when someone asks a question in your category.

The mechanics are straightforward. Define a prompt set that represents how buyers actually ask about the category. Run it repeatedly across the assistants that matter. Record whether the brand is named, in what position, and in what framing. The proportion of answers naming you is your AI SOV.

How to build a prompt set that means something

• Use real buyer language, taken from sales calls and support tickets, not from keyword tools.
• Cover the decision, not just the category. "Best X for Y" behaves very differently from "what is X".
• Include competitor-framed prompts, because the answer to "alternatives to [competitor]" is where displacement happens.
• Fix the set and re-run it on a schedule. A prompt set that changes each time measures nothing over time.
• Sample repeatedly per prompt. Model outputs vary between runs, so a single response is a data point, not a measurement.

What it does and does not predict

AI SOV is a reasonable leading indicator of consideration: brands that are named are the brands that get shortlisted, and in a conversation the shortlist is often three names rather than ten links.

It is a poor indicator of anything downstream on its own. Being mentioned is not being recommended, and being recommended is not being chosen. Framing matters as much as presence — appearing as the budget option in a premium comparison is a mention that hurts. Any serious measurement records sentiment and position alongside the count.

It is also volatile in ways that have nothing to do with you. Model updates move it, retrieval sources move it, and a change in the prompt wording moves it. Trend over a fixed set is meaningful; a single reading is not.

How to move it

Paid and organic both work, and they work differently. Paid inclusion — citation bidding, sponsored placement — moves presence immediately and stops when spend stops. Organic presence is a function of the trust graph: what authoritative sources say about you, how consistently, and how machine-readable it is. That is the discipline of GEO / AEO, and it compounds.