Editorial Policy

AI disclosure

Every article on AInformed is generated by a large language model. No article is written by a human journalist. We do not publish AI-generated text as though it were human reporting, and we label the provenance of every story on the story itself.

We do not do original reporting. Every story is a summary of reporting published elsewhere, and every article links to the original source. If a story matters to you, read the source — it is the primary record, and our summary is not a substitute for it.

What we source from

Candidate stories come from RSS feeds and public pages of established AI research labs, technology publications, and open-source project announcements. We only summarize material that is already public. We do not paraphrase paywalled articles, and we do not republish source text verbatim.

Every published article names its source outlet and links to the original with attribution.

How stories are selected

Candidates are scored 0–10 on five criteria, and the highest composite scores are published:

• Relevance — is it genuinely about AI, machine learning, or language models?
• Novelty — is it new, or a restatement of something already covered?
• Significance — does it matter to researchers, developers, or the wider industry?
• Readability — is the framing precise rather than clickbait?
• Accessibility — can it be explained clearly to someone who does not follow AI closely?

Stories covering the same topic as a recent article are penalized, and near-duplicates of the same underlying event are removed so we publish one story per event rather than several.

Accuracy review

Each draft is passed through a second, separate model acting as an editor. That pass checks the draft against the source material for claims the source does not support, exaggerations, and missing context, and rewrites where necessary.

This is an automated check, not human fact-checking. It catches a meaningful share of errors. It does not catch all of them.

Known limitations

We would rather state these plainly than let you discover them:

• Language models can state things confidently and incorrectly. Names, figures, dates, and version numbers are the most common failure points.
• A summary compresses. Nuance, caveats, and dissenting views present in the source may not survive into our version.
• If a source is wrong, we will usually reproduce the error.
• Our articles are short by design. They are a starting point for orientation, not a substitute for reading primary sources or research papers.

Treat AInformed as a daily index of what happened, and follow the source link before relying on any specific claim.

Corrections

If you find an error, email ainformed@proton.me with the article URL and what is wrong. We aim to respond within two business days.

How we handle them:

• Factual errors are corrected in place. The article records the date it was updated.
• Stories that should not have run — where the underlying premise is wrong — are removed rather than patched, and the URL redirects.
• Source-side corrections are followed. If a source retracts or amends, we update or remove our summary to match.

Rights holders who want a summary of their reporting removed can use the same address and we will action it without argument.

Funding and independence

AInformed does not accept paid placements, sponsored stories, or affiliate arrangements, and no third party can pay to have a story included, excluded, or framed a particular way. Story selection is driven only by the scoring criteria above.

Contact

More about who runs this site and how it is built is on the About page. For anything else, see Contact.