White papers
Reference documents, freely available, where every claim is backed by a verifiable public source.
We publish these documents for a simple reason: on most technical subjects that touch B2B marketing, the verifiable and the invented read exactly alike. Telling them apart means tracing every claim back to its primary source, or establishing that it has none.
These documents recommend nothing. They set out what is known, and how it can be known. They are published in the open, with no form to fill in, and may be cited with attribution.
August 2026
Counting bots, not lies: measuring AI traffic with verified identity
A bot is whatever it declares itself to be, and almost every published figure on AI crawling sums those declarations. Thirteen days of logs, two sites, every request verified: of 1,129 requests presenting themselves as an AI assistant, 876 were proven impersonations, and they are what manufactures the error rate blamed on engines.
Read moreAugust 2026
The recoverable message: what answer engines read of your site
An answer engine can say what your company does, not what sets it apart. Twenty edge cases submitted to three extractors show why: the narrative survives, the proof dies. This document measures the share of positioning that survives a read without JavaScript, and what French tech loses there, across 94 audits.
Read moreJuly 2026
LinkedIn Algorithm 2026: what is proven, what is measured, what is invented
On 12 February 2026, twenty-four LinkedIn engineers published the model that ranks the feed for 1.2 billion members. Almost none of the articles explaining "the new algorithm" cite it. This document compares what LinkedIn publishes, what independent studies measure, and what everyone repeats.
Read more