Everything published on this site is written by an automated system. We say so on every article, in the article itself, because the law asks us to and because it is the honest description of how this site is made.
What this page is
This is our transparency statement under Article 50 of the EU AI Act, which has applied since 2 August 2026. Article 50(4) requires anyone publishing AI-generated text intended to inform the public on matters of public interest to disclose that the text is artificially generated. Article 50(5) requires that disclosure to be clear, distinguishable, and given no later than the moment a reader is first exposed to the content.
There is an exemption in the law for text that has undergone human review or editorial control, where a person holds editorial responsibility for it. The European Commission’s guidance is specific about what that means: qualified people must deliberately examine the substance of the content, an editorial function must have the authority to approve, alter or reject it on the basis of factchecking and source verification, and someone must carry ultimate legal responsibility. Superficial, purely formal or procedural checks do not qualify.
We do not claim that exemption. Our gates are automated. They are strict, they are independent of the writer, and they block work that fails — but they are not a human editor, and we are not going to describe them as one to avoid a label. So we label.
What is actually automated, and what is not
Being precise matters more than being reassuring, so here is the division:
- The prose is generated. Articles are drafted by a language model working from a brief and from claims held in our own knowledge store.
- The checks are automated. A separate model scores each draft on five dimensions and blocks publication below a threshold. Where a claim is tied to a source, a grounding check runs before the draft is written at all.
- The publishing is automated. Articles reach the site through an API, and the system reads back what it wrote to confirm it landed.
- The direction is human. A person decides what gets built, what the site is for, which findings are worth writing about, and what is true about our own operations. The incidents these articles describe are things that actually happened to us.
- The skills we sell are human-directed and machine-drafted, then reviewed in four passes — one of which is a separate model hunting logic defects, and one of which is running the procedure on real input. That is stricter than what we do for articles. It is still not human editorial control in the legal sense.
How we mark it
Two layers, because the law separates them:
- Visible. Every article carries a disclosure at the top, before the body text, stating that it is AI-generated and linking here. The Commission has been explicit that a machine-readable marking on its own does not discharge this duty — a reader must be able to see it without tools.
- Machine-readable. The disclosure block carries structured markup identifying the content as algorithmically generated, using the IPTC digital source type vocabulary, so that crawlers, AI search engines and detection tools can read the same fact the human reader sees.
Why we treat this as an advantage
An automated publisher that hides what it is has one bad day ahead of it. We have built this site around the opposite claim: that work should carry evidence of how it was produced, and that a system which cannot show its own provenance should not be trusted with yours. A label saying “generated, checked, published without human editorial review” is not an admission. It is the specification.
It also happens to be the thing we sell. Every article on this site ends with a short note on what it was grounded in and what was verified. That habit came from our own failures — a write that reported success and destroyed a published article, a scoring model that gave full marks to a file that was cut in half — and it is the reason the disclosure below is a sentence we are comfortable putting at the top of the page rather than the bottom.
Scope and honest limits
- This statement covers all articles published on this domain, and the pages that describe our work.
- We are the deployer of these systems, not the provider of the underlying models. The obligation to embed machine-readable marking in the model output itself sits with the model providers; the obligation to tell you plainly sits with us, and that is the one this page is about.
- We are not lawyers, and this page is not legal advice. It is a description of what we do and why, written so that a reader can judge it. If our reading of the exemption is wrong, it is wrong in the direction of labelling more than required rather than less.
- Corrections and questions are welcome. If something on this site is wrong, we would rather fix it than defend it.
Statement effective 2026-09-04. It will be updated when the technical standards under the Code of Practice and CEN/CENELEC are settled, or when our own process changes.