Methodology
How NewsFacts collects articles, decides what counts as a change, and what it deliberately leaves out.
What we collect
We read each outlet's public RSS feed or news sitemap, the same lists outlets publish for search engines and feed readers. When a new story appears, we fetch the article page and save its headline, social headline, summary, byline, the story text and any correction or editor's notes.
We store the text only, not the page design, ads or images. Each saved version carries the time we captured it and a content fingerprint, so any version can be checked later.
How often we check
- Feeds: every 2 minutes.
- Each story: every 10 minutes for its first hour, every 30 minutes up to 6 hours, every 2 hours up to a day, every 6 hours up to 3 days, daily up to a week, then every 3 days until it is 30 days old.
- After any change, the next check comes within 10 minutes, because edits tend to come in bursts.
What counts as a change
A change is a difference in the words of the story, its headline, summary or byline between two saved versions. We compare paragraph by paragraph and then word by word, and record whether the headline moved, whether figures or quotations changed, and whether paragraphs were added or removed.
We ignore differences that are not editorial: spacing, curly versus straight quote marks, dash styles, "Updated at" timestamps, and site furniture such as subscription prompts or puzzle promotions that appear inside article pages.
If a story that we saved disappears (two "not found" responses in a row) or starts redirecting to a section page, we record that as an event too.
Our own mistakes
Extracting the story text from thousands of differently built web pages is imperfect. When we improve that process, any difference it causes is logged as our change, not the outlet's, and is never shown as an edit. If you believe a recorded change is wrong, tell us and we will check it against the saved versions.
Same rules for every outlet
Every outlet is collected, compared and displayed under the same rules. We list outlets alphabetically, and we do not publish bias ratings. A change is reported as a fact ("this sentence changed from X to Y at this time"), never as a claim about anyone's intent. Updating a developing story is ordinary, responsible journalism, and the record treats it that way.
What we do not do
- We do not get around paywalls, logins or blocks. Outlets that block automated access are simply not collected.
- We obey each site's robots.txt and identify our crawler honestly. More about our bot.
- We do not republish full articles. Public pages show the changed passages as short excerpts and link to the original.
Known limits
- A story edited in the first minute after publication, before we first see it, can't be caught.
- Live blogs, videos, galleries and podcasts are skipped. They change by design or carry little text.
- Some outlets serve different versions of a page to different visitors. When we detect that, we pause the outlet until we can collect it reliably.