How it works

What is measured, how, and by whom.

Most “AI visibility” tools hand you a number. This page tells you exactly how ours is produced, so you can argue with it.

Step 1

The questions

We ask each assistant three questions phrased the way a customer types them — never your brand name, because the point is to find out whether you come up when nobody is looking for you specifically.

Local questions use your city and state; national questions use your country. You can also write your own questions if your customers ask something particular.

Local · default set
“What are the best {category} in {city}, {state}?”
“Can you recommend a {category} near {city}, {state}?”
“Who are the top-rated {category} businesses in {city}?”
National · default set
“What are the best {category} companies in {country}?”
“Which {category} providers would you recommend?”
“Who are the top-rated {category} businesses I should consider?”
Step 2

Named, linked, or absent

Each answer is checked two ways. Named means the assistant used your business name (aliases and short forms count). Linked means it included your domain — either in the text or in the assistant’s source list, which Perplexity and Gemini return explicitly.

Linked is worth more than named. A name is a recommendation; a link is a recommendation the customer can act on.

Step 3

The score, with the arithmetic showing

AI visibility is the average across every engine × question: a link scores 1, a name-only 0.6, absent 0. Site readiness is a weighted pass/warn/fail across the 19 checks below — a blocked crawler weighs nine points, a missing Open Graph tag weighs one.

Overall = 60% visibility + 40% readiness. Letter grades: A 80+, B 65+, C 45+, D 25+, F below.

Step 4

The 19 website checks

Every one of these is deterministic: we fetch your homepage, robots.txt, sitemap and llms.txt and look. Each finding in the report says what we saw, why the assistants care, and — in the purchased report — what to do about it.

Technical

Structured data

On-page & content

Local entity & trust

What we don't do

Honest limits.

Answers vary

Why two runs can differ.

Assistants are probabilistic. Ask the same question twice and you may get a slightly different list. That’s why the report keeps every run and shows the trend rather than a single snapshot, and why three questions beat one. Treat one report as a reading; treat six months of them as evidence.

Now you know how it's measured. Measure yourself.

Free readiness scan. Full report $99.