About and method

One story, read in every language we can find.

When something happens, the world doesn’t read the same news. A vote in Asunción is a resignation in Jakarta and a removal in London. One court speaks with two official voices. A count of 372 becomes “377 problems”. cotext follows one story at a time through the coverage, in every language we can read, and shows you where the accounts split: who says what, with which numbers, in which words.

How a story is made

1. Collect the coverage

We gather headlines and links from Google News in 20 to 40 country and language editions, and add wire services and official sources: court and central-bank statements, parliamentary records, company repositories. A story usually ends up with between 1,400 and 7,300 articles in 15 to 24 languages. Every one of them is listed on the story page and links to the original.

2. Translate, then check the translation

Every headline gets an English translation. Machine translation gets numbers and names wrong surprisingly often, so headlines in non-Latin scripts, and any whose numbers don’t match the original, are re-translated and checked. On the Paraguay story that replaced 151 translations that were materially wrong. Featured headlines are checked by hand.

3. Count what changes

We count words and numbers in the original-language text, not in the translation: which verb, which number, which noun, in which country, on which day. Each finding says exactly what was counted, for example “1,640 headlines with a status verb”, so you can see the size of the sample behind a percentage.

4. Check the claims

Every claim in our video, and the most repeated claims in the coverage, gets a plain verdict: Holds up Unclear Didn’t hold, with the evidence and links. When sources disagree, we say so and mark the point as contested.

5. Keep the page open

Stories move after the video is posted. Pages marked Developing are updated as things change. Every update and correction is dated and stays visible at the bottom of the page.

What the numbers can and can’t tell you

  • It is a sample, not everything. Google News doesn’t list every outlet, and caps how many results it returns. Some languages and regions are better covered than others.
  • Times are when a story was first listed, not when it was published. Many items carry only a date. Where timing matters we say how precise it is.
  • “Country” means the outlet’s country, not where its readers are.
  • Most counts are about headlines. A headline can say “resigned” while the article explains the vote. That gap is part of what we measure, but it is not the whole article.
  • Reading figures come from Wikipedia page views, a rough signal of what people looked up, not of what they read or believed.

Corrections

If we get something wrong, we fix it on the page, say what changed and when, and keep the note there. Corrections that change a number in a video are noted on the story page, and the video is updated where the platform allows it.

Spotted a mistake, or have a source we missed? Write to hello@cotext.io.

How we use AI

Our research is AI-assisted. We use AI tools to collect and sort coverage, to translate headlines, to code words and numbers, and to help draft. Every figure on a page comes from data we keep, the method behind it is stated next to it, and the sources are linked so you can check them.

Fairness

We describe what was published, not who is right about the underlying dispute. People who are charged are described as charged, not guilty. We avoid naming or showing private individuals, and we don’t repeat unverified claims except to check them.

Sources and credits

Headlines and links via Google News, quoted for analysis; they belong to their publishers. Page-view data from the Wikimedia Foundation (CC0). Maps from Natural Earth (public domain). Flags from flag-icons (MIT licence). Each story page lists its own sources.

Privacy

This site uses no cookies, no ads and no tracking. Fonts and scripts are served from cotext.io itself.

Contact

hello@cotext.io

Each story is also a short video on TikTok: @cotextnews. The links are plain links: nothing from TikTok loads on this site.