What we learn from running brands through five AI assistants and Google AI Overviews every week — and what actually moves the answers. Written from our own run logs, with sample sizes stated.
The category defined without a sales pitch: the five metrics, the denominator problem that hides a zero, why citation ownership matters, and four questions to ask any vendor — including us.
Free checkers measure recall, not discovery: one branded question, one answer, no ownership check. Useful as a first signal, misleading as a score. What to run instead, in twenty minutes.
It's not a ranking problem — it's a missing-evidence problem. What we measured across the 4 core answer surfaces, why more pages on your own site rarely help, and the 6 fixes that do.
Asking ChatGPT once is an anecdote, not a measurement — the same prompt can change answers the same day. Here is the 20-minute setup: three things to fix, five columns to record, and three rules for telling signal from noise.
Our tool said 87/100. Google said nothing. The gap was our own measurement — we were putting the brand name in every prompt and reporting it as if buyers would see it.
Two audits can report 6% and 60% for the same week and both be arithmetically correct. Here is what an audit actually records, what the price tiers buy, and five questions that separate a measurement from a marketing PDF.
The word came from a 2024 paper, not from a vendor. Here is the definition it gave, the condition attached to its headline number, the three things GEO is not, and how to measure it without fooling yourself.
Same category, same name, different measurement choices. What each tool covers, where each is the better buy, and how to tell which BrandScope an AI answer is talking about — by domain, engine list, or price.
Most AI visibility dashboards show numbers that cannot change a decision. Here is the filter: what each metric answers, where it lies to you, and the four you should actually track.