// Concept
What is Share of Voice in AI search?
Share of voice (SoV) in AI search is the percentage of tracked prompts in your category where an AI engine recommends or cites your brand vs. competitors.
// Replacement
Why SoV replaces rank.
Keyword rank measures one position on one page for one query. That worked when users clicked. In an answer-engine world, rank can stay flat while citations vanish: the engine simply paraphrases your competitor and never names you. SoV measures the thing that now drives buyer awareness — whether the machine mentions you when somebody asks it for a recommendation. It rolls many prompts into one number, normalizes across engines, and compares directly against the brands you actually lose deals to. Rank tells you whether your page is reachable. SoV tells you whether your brand is being chosen.
// Measurement
How SoV is measured.
- Per engine. ChatGPT, Perplexity, Google AI Overviews, Bing Copilot, Claude, Gemini each get an independent score. They diverge more than people expect.
- Per prompt set. A curated list of 50–500 prompts that represent how real buyers ask about your category. Quality of the prompt set is the whole game.
- Mention vs citation. A mention is your brand named in the answer text; a citation is a linked source card. Most useful dashboards track both.
- Normalized share. Your mention rate divided by the total brand mentions on the prompt set. That’s your percentage of the conversation.
- Position weighting. First-mentioned brands compound; being named eighth in a list is worth less than being named first. Weighted SoV captures that.
// Reading
Reading a SoV chart.
A useful SoV view shows three things at once: your trendline, your top three competitors’ trendlines, and the engine breakdown. Watch the relative shape, not just the absolute number. A flat 22% while two competitors are climbing means you are quietly losing share even though the headline didn’t change. A spike on one engine and a flat line on the others usually means a single high-traffic source updated and the rest of the engines haven’t recrawled yet — informative, but not yet structural. The signals that matter are sustained, multi-engine moves.
// Levers
Lifting SoV: the four levers.
- Schema. Structured data the engines can extract without ambiguity. The cheapest lever and usually the fastest to move the needle.
- Citation-worthy content. Specific claims, primary data, original frameworks. Engines cite what stands out, not what restates the consensus.
- Entity completeness. One canonical brand identity across your site, Wikipedia, Wikidata, Crunchbase, LinkedIn, G2, and the analyst graphs. Disagreement confuses the engine.
- PR and earned mentions. Third-party citations on authoritative sites still feed the engines disproportionately. Press hits are still GEO infrastructure.
Vizelo measures all four and ships the changes that lift the score. GEO is the program; SoV is the scoreboard.
// Related
Keep reading.
FAQ
Common questions.
What is share of voice in AI search?
Share of voice is the percentage of prompts in a defined category where an AI engine mentions or cites your brand. It is measured per engine, per prompt set, and normalized so it can be compared across categories and competitors.
How is share of voice different from keyword rank?
Rank measures position on one results page for one query. Share of voice measures presence inside answers across many prompts and many engines. Rank can stay #1 while SoV drops to zero if the engine summarizes around you instead of citing you.
What is a good share of voice score?
It depends on category concentration. In a category with three real players, 25–35% is healthy and 40%+ is leadership. In a fragmented category, 10–15% can already mean you are top three. The score matters less than the trendline against your direct competitors.
How often should you check share of voice?
Weekly for normal monitoring, daily during launches or campaigns, hourly when you ship a major content change and want to see how fast each engine reflects it. Engines reindex on different cadences, so any single check can mislead.
Can you fake share of voice?
No. AI engines synthesize across many sources; you cannot bid your way in. You can earn share of voice through schema, entity completeness, citation-worthy content, and PR, but you cannot pay for it directly the way you can pay for an ad.