How articles are made
Our reporting is built from publicly available material: official statements and documents, primary data, and reporting already published by established outlets. AI tools read that material at volume, write the text, and translate it between our languages. They work to a fixed editorial code: original prose, every borrowed fact cited inline, hypotheses labelled as hypotheses, nothing invented.
What the AI does, and what it does not
- It does: read and condense source material, draft prose, suggest structure, translate between our languages, and check a finished draft back against the sources it was written from.
- It does not: invent a fact, publish a claim it cannot cite, or present a hypothesis as something established.
How we label it
- Every article carries the ✳ mark and a note under the byline saying it was written by AI.
- Machine-translated versions are labelled as machine-translated.
- Bylines are standing editorial voices, not real people. They are marked ✳ wherever they appear, and no article is attributed to a person who does not exist.
- Each article lists the sources it was written from, so you can go and read the original.
The limits of this
AI tools misread nuance, flatten context, and carry errors over from their sources. The fact-check index published with each article measures how well a claim is sourced — it is not a verdict on whether the claim is true. Coverage of a developing event is provisional and will often be overtaken. Read what is here as a starting point and follow the links to the original reporting.
When we get it wrong
Every article carries a ⚑ Report button and no account is needed to use it. You can also write to martens@nunissum.eu directly. We read everything that comes in, correct or withdraw what is wrong, and mark what changed. The details are in the legal notice.