GEO Research
The limits of measuring AI search
Variance, personalisation and opaque systems: what any AI visibility number can and cannot tell you.
Any AI visibility figure is an estimate drawn from a sample. This note sets out the sources of uncertainty our methodology accounts for.
Run-to-run variance
The same query can produce different answers minutes apart. Single observations are anecdotes; shares across repeated runs are measurements.
Personalisation and context
Signed-in history, location and conversation context change answers. Reports must state the conditions sampled.
System changes
Models and retrieval pipelines are updated without notice. A change in visibility may reflect the system, not the organisation's work.
Implication for reporting
Reports should show the method, the sample size and the date range alongside every number, and should avoid attributing changes to specific actions without evidence.
Related reading
- AI VisibilityHow to measure AI visibilityA repeatable method: define the query set, sample systematically, and report measures separately.8 min read
- AI SearchWhat is Generative Engine Optimization?A plain definition of GEO, what it changes about search work, and what it cannot promise.6 min read
- SEO + GEOSEO and GEO: how they relateGEO does not replace SEO. Here is where the two disciplines overlap and where they diverge.5 min read
Terms used here are defined in the glossary.