How it works
Let's walk through a scan.
Step by step, here's what happens from the moment you type your search to the report. We don't promise timings; we show you the real status of every step.
01 / 05
Clarifying the search
You confirm whether what you typed is a brand or a keyword. For a brand, we set the category, domain and aliases; for both types, the target market and answer language.
02 / 05
Building the question set
We build a set of category-specific buying, comparison, alternative and trust questions. In brand searches, none of the questions contain your brand name; otherwise the result would be biased. The set is versioned and frozen.
03 / 05
Asking the providers
The questions are sent to the configured AI providers' APIs with web search enabled. Each call's status is tracked separately: succeeded, refused, timed out or errored.
04 / 05
Extracting the evidence
Brands mentioned in the answers are matched against their aliases; ambiguous matches are set aside for review. The source links the provider returned are recorded exactly.
05 / 05
Building the result
Only valid answers enter the denominator. The result is presented with the denominator, a provider breakdown, observed and not-observed examples, and sources.
Why is one search just “one” search?
A single brand or keyword analysis can trigger dozens of AI requests behind the scenes. Those aren't counted individually: the analysis you start is one search. A scan that fails completely doesn't use up your allowance.