Editorial Policy
PageNews does not employ journalists and produces no original reporting. This page sets out the standards that do apply: how sources are chosen, what we change, and what we never change.
Choosing sources
A publisher is added when it meets practical criteria: it operates a public RSS or Atom feed intended for syndication, it publishes regularly, it identifies its own authors and corrections process, and it is an established outlet rather than an aggregator republishing others. Most publishers absent from this site are absent because they offer no usable public feed, not because of a judgement about their reporting.
We do not describe our sources as “trusted” or “verified”. We have no editorial apparatus that would make such a claim meaningful. The full list is on the sources page so you can judge for yourself.
What we never alter
- Headlines appear exactly as the publisher wrote them.
- Attribution is never removed, obscured or replaced.
- We never present another publisher's reporting as our own.
What we do alter
- Summaries are truncated to a consistent length, and HTML markup and feed artefacts are stripped.
- Tracking parameters are removed from links; the destination is unchanged.
- Timestamps are converted to UTC and displayed relative to now.
Automation and its limits
Categorisation, story grouping, trending topics and ordering are entirely automatic. No person reviews an individual story before it appears. This is fast and consistent, and it is also fallible: stories land in the wrong section, related events are occasionally merged, and topic labels can read awkwardly. We would rather state this plainly than imply an editorial review that does not happen.
No summaries we invented
Every summary shown on this site comes from the publisher's own feed. We do not generate summaries, and we do not use language models to rewrite or extend what a publisher wrote. No sentence of anyone's reporting is reworded and presented as ours.
Coverage analysis
Where several publishers cover the same event, PageNews measures and compares the language each of them used — how far a report leaned on the vocabulary of conflict or of agreement, whether it foregrounded a named individual or an institution, how it framed cost or consequence. Those measurements are ours, they are labelled as ours wherever they appear, and they are the reason an event page here is worth reading alongside the reports it links to.
This is measurement, not writing. The analysis is rule-based rather than a language model: it counts terms from curated lists and shows the terms it counted, so any figure can be checked against the report it came from. Nothing is generated, and the promise above is unchanged — no publisher's text is rewritten into prose of ours.
Every score is shown with a separate confidence figure and with what was actually read, which is a headline and, where the publisher supplies one, their own short summary. It is never the full article, because we do not hold the full article.
What the analysis does not do: judge whether a report is accurate or fair, and describe a publisher's politics. It measures word choice in one report on one event. It can be wrong, and where it matters you should read the original reporting — every report is linked. The full method, including its limitations, is on the methodology page.
Commercial independence
No publisher pays for inclusion, placement or ranking. There is no advertising on this site and no commercial relationship affects the order of anything.
Errors
See Corrections.