Rankings tell you where you sit on a page. Share of voice tells you whether AI mentions you at all. Here is how to measure it.
A keyword ranking answers one question: where does my page sit for this term. AI search needs a different question. When someone asks an AI to recommend a tool like yours, are you in the answer, and how often. That is share of voice.
Three metrics make it concrete.
Sentiment sits alongside these. Being mentioned negatively is its own problem, and worth tracking separately.
The method is the same whether you do it by hand or with a tool.
1. Pick the prompts that matter. Not keywords, prompts. The actual questions a buyer asks: "best X for small teams," "X vs Y," "is X worth it." Twenty to a hundred good prompts beats a thousand vague ones.
2. Run them across the models your buyers use. ChatGPT, Claude, Perplexity, Gemini, and increasingly Grok and DeepSeek. As the citation data shows, they answer differently, so one model is not a proxy for the rest.
3. Record who gets mentioned and cited, including competitors. Your share of voice only means something next to theirs.
4. Repeat weekly. A single reading is a snapshot. The trend is the insight.
Good is not a number you hit once. It is a line that climbs. Rising share of voice next to rising branded demand is the healthy signal, the one executives actually ask for: more people are hearing about you, from a source they trust, and coming to find you.
Doing this by hand across six models every week is tedious, which is exactly why it gets skipped. The work that does not get skipped is the work a tool runs for you. Either way, the unit you are tracking is the prompt, and choosing the right ones is its own skill.
