Share of Voice in AI Search: How to Calculate It and What Good Looks Like

Share of Voice is the brand's mentions divided by the total mentions of every tracked entity — the brand plus its configured competitors — across the same prompt set, providers and date range. A brand with 40% SoV in a category is mentioned four times for every ten mentions any tracked brand receives.

SoV is not the same as market share, and it should never be reported as one. It reflects what a set of AI systems generated for a specific list of prompts on a specific day. Change the prompt wording, the region or the competitor list, and the number moves — which is exactly why the method has to stay fixed between comparisons.

Define the prompt set that represents the category — not just branded queries, but the discovery, comparison and problem-solving questions a buyer actually asks. Run every prompt across the chosen AI systems and record every entity mentioned in each answer, not only the brand being tracked.

Count mentions per entity across the full run, sum them into a total, then divide each entity's count by that total. Repeat the exact same prompt set and provider mix on the next run so the two numbers are comparable. A single-day SoV number is a snapshot; a SoV trend across several runs is evidence.

AI answers are not static index lookups. A model update, a provider's retrieval change, a competitor's new publication or simple response variance can shift SoV between two runs even if your own pages were untouched. Treat a single-run swing as noise until it repeats.

The fix is the same discipline used in any monitoring system: hold the prompt set and provider mix constant, run on a schedule, and only act on a change that persists across multiple runs and, ideally, multiple providers. AI Control keeps runs, dates and filters attached to the same project so a trend line stays comparable over time.

SoV alone can hide an important distinction: two brands can have the same SoV while one is consistently recommended first and the other is buried in a list of alternatives. Always read SoV together with average position — presence without prominence is a weaker result than the raw percentage suggests.

Citations add a third layer. A brand can hold strong SoV while the cited sources belong entirely to third parties — reviews, directories, comparison sites — rather than its own domain. That combination points to a content and authority gap even when the visibility number itself looks healthy. How LLM answers, positions and citations differ goes deeper into reading these three metrics together.

There is no universal healthy percentage — it depends on how many competitors are tracked and how concentrated the category is. A category with three tracked brands and one dominant incumbent will naturally show a higher ceiling than a fragmented market with fifteen alternatives.

The more useful benchmark is the brand's own trend and its position relative to its closest tracked rivals, not an external target. A rising SoV alongside a stable or improving average position is the clearest sign that AI visibility work is compounding rather than just moving noise around.