Glossary · Measurement
Share of Model
A brand's share of an AI model's static, parametric knowledge about a category — tested with prompts that block live search or browsing — as distinct from its share of voice across live, retrieval-augmented answers.
How it's tested
Share of model is measured by running prompts in a mode where the AI can't browse the web or use live retrieval, forcing it to answer purely from what it learned during training. Whatever the model says about a category in that mode reflects its baked-in knowledge, not what it can look up in the moment.
How it differs from share of voice
Share of voice is measured from live, often search-augmented answers, which can shift day to day as new content gets published and indexed. Share of model reflects the training snapshot itself, which only changes when the underlying model is retrained — making it a slower-moving, more foundational baseline than live share of voice.
Why it matters
A brand with low share of model but improving share of voice is winning through fresh content and citations even though the model's core knowledge hasn't caught up yet. That gap is useful to know, since it points to a ceiling that content alone can't fully overcome until the next training cycle.
See more terms in the full glossary, or read the AI Visibility 101 guide for the full picture.