Skip to main content

How this site is written

Editorial policy

The rules this site holds itself to: who writes what, how AI is used and disclosed, what gets verified, and what happens when we get something wrong.

Production and review

The six commitments

A named human owns every claim

Every article carries a real byline. Lars Nyman is the author of record for the corpus: the positions are his, the experience cited is his, and he is accountable for every factual claim. There are no ghost personas and no invented experts on this site.

AI-assisted drafting, disclosed

Much of the library is drafted with AI tooling, the same category of tooling the firm advises clients on, and then edited and fact-owned by the named author. We think disclosure beats pretense: a firm selling AI-search expertise that hid its own AI use would deserve the skepticism.

Nothing invented, ever

No fabricated case studies, no invented client names, no manufactured statistics, no fake reviews. When we have a number, it traces to the benchmark dataset or a named external source. When we do not have a number, the article says so or does without.

Data claims trace to public methodology

Figures about AI visibility come from our own audit engine's aggregate data, published with a downloadable dataset and a public scoring rubric covering every check, weight, and limitation, including the sample's self-selection bias.

Style is enforced by machines, not vibes

An automated gate blocks deploys that violate the house style: budgeted phrase overuse, banned openings, em dashes, hard avoid-list marketing words, drifting founder-experience claims, and self-serving review markup are all build failures, not suggestions.

Dates mean something

Published and updated dates change when the substance changes, not to game freshness. A weekly job reports stale content so it gets a real revision or an honest old date.

When we get it wrong

Corrections

If you find a factual error anywhere on this site, including in the press citations, the benchmark figures, or any claim about how an AI engine behaves, send a message with the URL and the issue. Material errors get corrected in the page itself with the updated date bumped; disputed scoring questions get checked against the public rubric. We respond within one business day.

The full internal style specification this page summarizes lives in the site repository and travels with every content change; the enforcement gate runs on every build.