Including the one that cannot honestly be scored — and where we stand on each. Use this list on us before you use it on anyone else.
There are two different products in this market, and they are frequently confused.
A monitoring platform continuously measures how visible you are. It answers "how are we doing?" and it answers it every day.
An audit investigates why you are not appearing, what AI currently believes about you, which sources are shaping that belief, and exactly what to change. It answers "what is wrong and what do we do?"
They are not substitutes. A monitoring subscription gives you a number and leaves you to work out what to do with it. We are the second kind, and if what you actually need is daily tracking across a large brand portfolio, we will tell you that and point you at a platform.
Ask any provider — us included — to answer all six. A provider who cannot show you where they fall short on any of them is telling you something.
We measure it. We ask the questions your buyers actually ask, across five AI engines, in fresh signed-out sessions, and count how often you are named. We report a rate with a range, never a bare percentage, because AI does not answer identically twice.
We capture it. We refuse to score it — and this is the most important line on this page.
SparkToro and Gumshoe ran 2,961 tests with 600 volunteers across ChatGPT, Claude, and Google AI. The same prompt returned the same list of brands fewer than once in 100 runs, and the same order fewer than once in 1,000. In their own summary: any tool that gives you a "ranking position in AI" is, quote, "full of baloney."
So we record where you appeared, and we never turn it into a score or a trend line. If a competing provider shows you an AI ranking position, ask them how it is reproducible. That question does not have a good answer.
We measure it. Every source each engine cites while answering your buyers' questions, grouped and counted, with your own pages marked present or absent. This is usually the most actionable section in the report, because the sources shaping your category are finite and mostly reachable.
We measure it, and we report the errors rather than filtering them out. We separate two different problems: things AI invented about you outright, and things that are really about you but stated wrongly. The source of truth is what you confirm, never another website. We treat this as a brand-safety finding, because a buyer acts on what AI tells them and will never call to say they were told something false.
We measure it. Share of the category conversation, who is named alongside you, which sources carry them and not you, and where their pages answer a question yours does not.
This is the deliverable, not an appendix. A prioritized 90-day plan with owners and checkboxes: entity and identity fixes first, then trust and corroboration, then reach. Every recommendation is labeled durable (holds for a year or more) or signal-dependent (reflects where the engines are looking right now, and gets re-checked at the re-measure), so you always know which is which.
"You have a 32% AI visibility score" is not useful. It tells you nothing about cause and nothing about what to do next.
What a finding should sound like:
"You are absent from most of the high-value commercial questions your buyers ask, because the engines are not finding independent evidence connecting your brand to the things you sell. Here are the specific sources currently shaping your category, here is where your competitors appear in them and you do not, and here is the order to fix it in."
That is the standard we hold ourselves to, and the standard we would apply to anyone else you are considering.
In August 2026 the Interactive Advertising Bureau (IAB) — the American trade body that has set advertising measurement standards since 1996 — published Measuring Visibility in the AI Era, the first common standard for measuring how brands appear in AI answers. It sets out nine quality criteria and eleven disclosures any provider should be able to answer.
We build every audit on it, and we publish how we score on all nine criteria — including the ones we do not fully meet. Our reports state their measurement tier on the first page.
The framework's own line on evaluating providers is worth keeping in mind: "where a provider cannot or will not disclose against a required item, that absence should itself be treated as a signal."
