Blog

Proven, not assumed

Every article here is built on its own measurement. The rule is simple: no figure, no article.

18 September 2026 · 6 min

llms.txt: useful file or empty shell? What the data shows in 2026

Plugins generate llms.txt with one click, but no major AI assistant demonstrably uses it. Here is what the data says and what actually works measurably.

Case study: one brand, three assistants, three realities

180 measurements on the same questions: 36% in Claude, 5% in ChatGPT, 1% in Gemini. The full case study, including what happened next.

Choosing an AI visibility tool in 2026: what to look for

Prompts or measurements, with or without an interval, white-label or not: the five questions that decide which tool suits you, including when you are better off with a competitor.

Is my company in ChatGPT? Here is how to find out (and how not to)

The question every entrepreneur should ask themselves, and why asking it yourself is the worst measuring instrument there is.

We asked ChatGPT, Gemini and Claude the same question 60 times

Same brand, same questions, three assistants: 36%, 5% and 1%. Why AI assistants fundamentally disagree, with all the figures.

A site that scores 90 technically and appears in no AI answer at all

Why your AI visibility is not decided on your own site but on the sources assistants cite from, illustrated by one painful measurement.