A measure of how deeply a brand is woven into a model's memory, from how often and how early the model recalls it across many runs.
Associative embeddedness is our measure of how deeply a brand is woven into a language model's memory — not just whether the model knows it, but how central it is to the model's web of associations. It's the metric behind our AI Brand Authority Index, which ranked 2.9 million brands by their standing in Gemini's memory.
It's built from recall behaviour. Across 200,000 runs asking the model to name brands at random, each brand gets a frequency (how many runs it appeared in) and an average rank (how early it was recalled). Those combine into a seed weight — a brand recalled in every run and named first scores near 1.0; one recalled once at position 98 scores near zero — which then feeds a Personalized PageRank walk over a directed association graph.
The result quantifies brand authority in a way traditional rankings can't, and it sits alongside frequency and share of voice as a core AI visibility measure.
