# Answer variance

> The tendency of an answer engine to give different answers to the same question.

## What it means in practice

Sampling, retrieval, and personalization all move between runs, so a single answer is one draw from a distribution rather than a reading of a fixed rank. A brand present in one answer and absent from the next has not necessarily changed. The practical consequence is that any AI visibility number built on one observation per question is not measuring the brand, and that comparisons need repeated runs over a fixed prompt set.

## Category

Concept in How AI answers work.
This definition is editorial category vocabulary, not a metric refd computes.