AI Visibility Testing

I ran 25 Google AI Mode searches to explore how AI recommendation share of voice can be measured manually. The results show why overall visibility percentages mean little without knowing which prompts, customer needs and counting rules produced them.

A simple AI search can reveal whether your website is being described as you intended. This practical sense check shows how to test AI’s interpretation of your brand, identify factual misunderstandings and monitor how that perception changes over time.

Google AI Mode mentioned my new website but incorrectly described it as an automated monitoring platform. This experiment shows why AI visibility should measure not only whether a brand appears, but whether it is understood and represented accurately.

Repeating the same Google AI Mode product search 20 times revealed changing shortlists but a surprisingly clear core of favoured recommendations. Here is what the experiment showed about repeated testing, recommendation patterns and establishing a baseline for future monitoring.

Using ten repeated Google AI Mode searches, this guide shows how to calculate recommendation coverage, share of voice and position—and why AI visibility is a changing pattern rather than a fixed ranking.

One AI Mode response contained three recommended products but 24 different linked pages. This experiment examines the difference between being recommended, mentioned, cited or used only as contextual support.

A practical guide to running repeatable manual AI visibility tests, from setting consistent conditions and recording results to establishing a baseline and monitoring how visibility changes over time.

I ran the same product search through Google AI Mode ten times to see whether its recommendations remained consistent. Six different products appeared, four occupied first place, and only one was recommended every time.