Share of Model

How it's measured

Take a curated list of high-intent buyer questions for a category (typically 30–50 spread across the journey). Run each question against every relevant AI provider — ChatGPT, Claude, Gemini, Perplexity, Tavily, Exa — via direct API calls, with several temperature samples per query. Count how many answers mention the brand. Compute the rate per provider, then average across providers so a chatty model can't dominate the score.

Why it matters

When buyers ask AI assistants instead of search engines, the answer is a paragraph naming two or three brands — not ten blue links. Share of Voice measured attention in the channel buyers used to live in. Share of Model measures presence in the channel buyers are moving to.

What it replaces

Share of Voice (SOV) — the legacy metric for share of paid impressions or organic results — assumed attention correlated with consideration. Generative AI breaks that chain by collapsing the consideration set to a handful of named brands per query. SoM measures the new gating step directly.