What is AI Share of Voice?
Share of Voice (SoV) measures how large a portion of the total brand-mention landscape your brand occupies across AI-generated answers. In concrete terms: when all the brands mentioned in a set of AI responses are counted together, what percentage of those mentions belongs to you?
A brand with 40 mentions out of a pool of 200 total brand-mentions holds a 20 % Share of Voice.
How Mentova computes it
- Run prompts across models. Each campaign sends a set of prompts to every selected AI model and stores the raw responses.
- Extract brand mentions. A multi-strategy extraction pipeline (pattern matching, NER heuristics, optional LLM fallback) identifies every brand name in every response, then normalizes and deduplicates them (e.g. "Emakina" and "Emakina Group" are merged).
- Count total mentions. Mentova sums the mention counts for every meaningful entity found across all completed prompt runs within the selected scope.
- Divide. Your brand's total mention count is divided by the sum of all entities' mention counts and expressed as a percentage.
Entities that appear in only one prompt run are treated as likely hallucinations and excluded from the pool, keeping the signal clean. Your own brand is always included even when it appears rarely.
How to read it
- High SoV (> 30 %) - AI models cite your brand very frequently relative to competitors in this topic area.
- Low SoV (< 10 %) - you are underrepresented; competitors dominate the AI narrative on this topic.
- SoV by model - you can break the figure down per AI model to spot where you are strong or absent (e.g. mentioned by Gemini but invisible to Claude).
SoV is calculated over the 90-day rolling data window.
Common pitfall
Do not confuse Share of Voice with Mention Rate. SoV tells you how you compare to the competition; Mention Rate tells you how often you appear at all. You can have a high SoV on a topic where even the market leader is barely mentioned - that topic may simply not trigger many brand references.