The sources
This desk currently reads 42 active outlets of 47 in the registry. It has taken in 13,259 reports. Every story names the outlets it rests on, and every claim names the ones that carried it.
How an outlet gets read
We read published feeds, at a deliberate pace, one request at a time per outlet. robots.txt is fetched and honoured before anything else; where a feed asks not to be read, it is switched off and the reason recorded. We send conditional requests, so an outlet that has not published since our last visit serves us nothing and pays nothing.
Languages
The registry is deliberately not English-only. It currently covers English, Spanish, French, German, Italian, Portuguese, Arabic. A story assembled from outlets in three languages is stronger evidence than three outlets in one, because they are less likely to be repeating each other.
When an outlet stops working
Feeds break, move and disappear. Every fetch is logged — successes as well as failures — because a log that only keeps failures cannot tell “this outlet is broken” from “nobody has tried it since Tuesday”. An outlet failing repeatedly is flagged for a person, and one that is switched off must carry a recorded reason.
What we do not take
We do not store or republish full articles. We read what an outlet publishes for readers, write our own explanation, and link back. If an outlet would rather we did not read it at all, that is their call to make and we will honour it.
Corroboration is not agreement
Several outlets carrying the same wire copy is one report wearing several mastheads. Where we can tell, we treat it as such — and where we cannot, a status marked confirmed still shows you exactly which outlets it rests on, so you can judge for yourself. What each status is claiming is set out on the statuses page.