Editorial policy

Articles on this site are produced by an automated pipeline and written with machine assistance. No human reporter attends events or conducts interviews for this publication. We say so plainly because you are entitled to know what you are reading.

The pipeline monitors public trend data, collects reporting from multiple independent outlets, extracts the factual claims those outlets have published, and writes an original summary and analysis from those claims.

• We do not copy or republish source articles. Only headlines and publisher-written summaries are ever ingested; full article text is never fetched or stored.
• We do not fabricate quotes, statistics or events. Every claim must originate in a cited source, and the writing stage is given no other material to work from.
• We do not invent authors. Articles are attributed to the publication, not to fictional journalists with invented biographies.
• We respect robots.txt and API terms. Where a publisher disallows automated access, we do not access it — even when that costs us a source.

Each extracted claim is scored by how many independent publishers support it, weighted by their reliability. Counting URLs would be misleading, because one wire story syndicated to ten sites is one source, not ten.

A story must draw on at least 2 independent publishers before it is written at all. Claims that fall below that bar are attributed in the text (“according to…”) rather than stated as established fact. Stories where sources actively disagree are labelled, and the disagreement is described rather than resolved.

Articles that are still developing carry a visible notice. Health, financial and safety topics are held to a higher sourcing bar than general news, because the cost of being wrong is higher.

Rather than repeating a citation list under every article, we describe our sourcing once, here. Each article states how many independent outlets it drew on, and the full source list for every piece remains embedded in that page’s machine-readable metadata.

Every story is compiled from a combination of the following:

• News publishers’ own syndication feeds — including the BBC, NPR, The Guardian, Reuters-carrying outlets, CBS, NBC, CNBC, Der Spiegel, Le Monde, NHK, The Hindu, Indian Express and others, market by market.
• Bing News search, for topic-specific coverage across a wider set of outlets than any single feed carries.
• GDELT, an open index of worldwide news coverage, used to find reporting we would otherwise miss.
• Wikipedia, for background and to confirm which entity a story is actually about. Never used as evidence for breaking news.
• Google Trends and Wikipedia pageview data, to identify which topics people are actively searching for.

We take only headlines and publishers’ own summaries. We do not fetch, store or reproduce the body of any article. Where reporting is disputed or thin, we would rather publish less than fill the gap with speculation.

Published articles are re-checked against their sources on a schedule. When the underlying facts change materially, the article is rewritten and its modification date updated. When nothing has changed, we deliberately leave the date alone rather than refresh it to appear current.

If you find an error, write to corrections@example.com. Substantive corrections are applied to the article and noted.

Photographs come from Wikimedia Commons, Openverse and comparable free libraries. We do not use AI-generated imagery. Where no appropriately licensed image exists, the article runs without one.

Each image carries a visible credit naming its creator and licence.