Documentation · Updated 13 November 2026
Review and publication
Our collector records far more changes than appear on the site. This page explains what is published, what is held back, and why.
Why there is a review step
Reading the story out of a web page is imperfect. Sites redesign their pages, serve different layouts to different visitors, and put promotional text in the middle of articles. Any of those can look like an edit when no editor touched the story. Publishing a false "this outlet changed its story" would be worse than publishing nothing, so a person checks each change before it becomes public.
What happens to each change
When a change is recorded it is sorted by rule into one of three groups.
Sent to review. Headline rewrites, added corrections, changed figures or quotations, removed paragraphs, other rewording, and removed articles. A reviewer compares the two saved versions and either approves the change, sets it aside, or marks it as a collection error.
Kept but not listed. Small fixes of a few characters, stories that only gained new paragraphs, byline changes, and changes to the summary alone. These are saved in the archive in full. They aren't listed because there are a great many of them and they are routine.
Marked as our error. A difference caused by our own software, for example after we improve how story text is read from a page. These are logged as ours and never shown as an edit by the outlet.
What a reviewer checks
- That the change is in the story itself and not in something around it, such as a caption, a related-links box or a subscription prompt.
- That both versions are complete. A page that loaded only half its text can look like a mass deletion.
- That the excerpt shown is a fair picture of the change.
A reviewer does not decide whether an edit was justified, and does not approve or hold changes according to the outlet or the subject. The test is only whether the change is real and correctly shown.
What publication means
An approved change appears on the Changes page and on the story's record, usually within about five minutes, and is added to the RSS feed. It shows short excerpts of the changed passages and links to the original article.
If we get one wrong
If a published change turns out to be a collection error, we take it down and mark it as our error in the archive. If you think a record is wrong, here is how to tell us.
What we can't see
- An edit made in the first minutes after publication, before our first capture.
- Anything behind a paywall or login, and any outlet that blocks automated access. We don't work around either.
- Changes to images, video, captions or page layout. We record text.
- Edits made after a story is 30 days old, when we stop checking it.
- Two edits made between the same pair of checks. We see the combined result as one change.