How we work, and what we will not claim.
Most of this industry sells confidence. We would rather publish the method, grade every finding, and tell you which of your questions the evidence cannot settle.
Every finding carries a grade
A report is only useful if you can tell the difference between something we watched happen and something we think is probably true. So each finding is labelled, and the label travels with it.
That is why the report is longer than a dashboard and worth more than a score.
- ObservedWe ran the conversation and preserved it. Dated, quoted, repeatable.
- VerifiedChecked against a source that is not us and not you.
- Our interpretationOur read of what the evidence means. Reasonable, and still a judgement.
- HypothesisWorth testing. Not yet tested. Never charged for as a finding.
- UnknownThe evidence cannot settle it. Said out loud rather than papered over.
How the work is done
- We agree what the report is forTwenty minutes. What decision it should inform, which markets and languages matter, what you already suspect is wrong.
- We check you can be found at allCrawler access, whether your pages survive without JavaScript, duplicated language versions, product data. Unglamorous, and nothing below it matters until it is right.
- We run the prompts your customers useBy a person, not scraped. Real questions in your markets and languages, across the AI search platforms your customers actually use, each one recorded with its date and conditions.
- We ask the same thing more than onceThese systems do not answer identically twice. A finding that appeared once is reported as one occurrence, never as a pattern.
- We trace what we find back to its sourceWhen AI search says something wrong about you, it got it somewhere. We go and find where.
- You get the report and a sessionWe walk you through it and prioritise together. That session is part of the work, not an upsell.
What we will not claim
We can't make AI search recommend you, and there is no ranking dial to turn. Anyone selling you a guarantee is selling something else.
We will not give you a single visibility score. Being mentioned, being cited, being recommended and being described accurately are different things, and averaging them hides the one that is costing you money.
We will not tell you a competitor's private numbers, forecast a curve on platforms that change this often, or present a finding from one conversation as a trend.
And where the honest answer is that we do not know, the report says so, with the reason.
Ask us how we would test your business.
Twenty minutes, no access to anything, and you will know whether this is worth doing.