How to measure AI visibility
Choose questions, repeat tests, keep conditions separate and avoid treating a citation as a recommendation.
Takusen / Guides & Research
Practical guidance for designing a useful benchmark and making public content accessible. These guides explain methods; they do not claim proprietary studies or measured performance gains.
Choose questions, repeat tests, keep conditions separate and avoid treating a citation as a recommendation.
Understand robots policies, served HTML, response codes and the limits of user-agent diagnostics.
Definitions, records, denominators and exclusions for an agreed audit panel.
A dated internal example that separates source observations from public HTTP evidence.