What is share of model?

Share of model borrows the logic of share of voice and applies it to generative engines. Instead of counting your brand's slice of ad impressions or search results, you count how often it shows up in AI answers: ask ChatGPT or Claude a category question such as best project management tool for agencies, and record which brands appear. Run enough of those prompts and the mention rate becomes a benchmark you can track against competitors month over month.

There is no console that reports how often a model mentions you, so measurement is a sampling exercise. You build a prompt panel: twenty to fifty questions real buyers ask in your category, phrased in different ways. Each prompt runs several times per engine because outputs are non-deterministic; the same question can name you on one run and skip you on the next. Your share of model for a category is mentions divided by total runs, per engine, tracked on a schedule.

The metric has limits. A raw mention count hides quality: being named third with a caveat is worth less than being the single direct answer, so position and sentiment need scoring alongside frequency. Results also shift with prompt phrasing, and a model update can move your score without any change on your side. Our July 2026 study found 27 of 30 brands materially misdescribed by an AI assistant answering from memory, so a high mention rate can still carry wrong claims.

In practice, treat share of model as a trend line rather than a precise figure. Pick a stable prompt panel and hold it constant, because direction over quarters matters more than any single reading, much as share of voice was read in media planning. Tooling is early: Discoverable runs this style of check with a free AI visibility snapshot that needs no login, with Claude checks live today and other engines rolling out, and its Growth plan adds weekly automatic re-checks so the trend line keeps updating without manual runs.