AI visibility research, with the method in view.

How we measure AI visibility, published in full. We publish findings only when the data supports them.

MethodologyUpdated 22 October 20261 min read

How we log AI visibility: 10 questions × 4 engines, weekly

Starting September 2026, we log 10 buyer questions across 4 AI engines every week, 3 runs per question per engine — 120 sampled answers per week — and publish the method here.

Read the method
Principles

A method you can inspect.

We publish the method before we publish findings, and findings only when the data supports them. Each post states its sample, its dates and its limitations, so you can judge it for yourself.

Repeat observations

AI answers vary across runs, prompts, locations and settings, so we ask each question several times.

Record provenance

We distinguish app checks from API outputs, mentions from linked citations, and the date of collection.

Publish limitations

Small samples are directional. They are not population-wide proof of visibility.

New to the terms? Start with our guide to AEO, GEO and SEO.

See where AI cites your business.