Bex Labs
Different News Views
How the story is picked, how countries are compared, and what the comparison cannot tell you.
Updated 7 Oct 2026, 21:37 UTC. It updates itself every day.
What it shows
Once a day, the page picks the biggest story in the world's news. For the countries that covered it most, it shows how much they wrote, which angles they took, how positive or negative the tone was, and a few of their headlines.
Step 1. The news
The news comes from GDELT, a project that reads news websites around the world every 15 minutes. GDELT reads news in English and in 65 other languages, which it translates by machine so that all articles can be compared in the same way.
The page reads 24 hours of this news, using one 15-minute batch from every hour, in English and in translation. That is a sample of about a quarter of everything GDELT reads in a day, spread evenly over the day. For each article GDELT lists the people, places and organisations it names, the themes it covers, and a tone score.
Step 2. The country of each article
Each article is given the country of the outlet that published it, using GDELT's own list of news websites by country. So "Germany" means outlets based in Germany, not articles about Germany. Articles from websites that are not on the list are left out.
Step 3. Picking the story
A computer cannot read the news and say what the big story is, so the page uses a simple rule instead. It counts how many articles named each person, place and organisation. It keeps those counts for the last seven days. The story of the day is the name that jumped most compared with that week:
- It must appear in at least 60 articles in the sample.
- At least 4 countries must have 12 or more articles naming it, so there is something to compare.
- Its share of the news must be at least twice its share in the previous week.
- Among the names that pass, the winner has the highest size times jump. The jump counts for at most eight times, so a small story that came from nowhere does not beat a big one that grew fourfold.
Names that are in the news every day, like Washington or Moscow, rarely pass this test. A name only wins when something happened. Platforms and news agencies, such as Instagram or Reuters, are never counted as names, because they appear in articles for reasons that have nothing to do with the story.
The rule picks a name, not a story. Sometimes that name is only part of it: a football match between Argentina and Benin may come up as "Benin" when the story is really about a famous player. The names it was most often mentioned with, and the headlines, show what the story is.
Step 4. Comparing countries
Every article that names the story is part of the comparison. The page shows the eight countries with the most of these articles, as long as they have at least 12.
Angle. GDELT tags each article with themes from a list of thousands. The page groups them into seven angles: security and conflict, economy and money, politics and government, diplomacy and alliances, human rights and people, health, and environment and climate. For each country it shows the share of articles that touch each angle. A country "leans on" an angle when its share is at least 5 points above the share across all countries.
Tone. GDELT scores each article by counting positive and negative words. The score is usually between minus 10 and plus 10, and most news sits a little below zero. The page shows each country's average.
Headlines. Up to three per country, preferring articles in the original language of that country's press, in their original words.
What the comparison cannot tell you
- Machine translation and tagging. Themes and tone for non-English articles come from a machine translation. Tone scoring counts words. It misses irony and context.
- Themes are broad. A sports story can pick up "politics" because a president attended the match. The angles are a rough guide, not a reading of each article.
- Not every outlet. GDELT reads a very large set of websites, but not all of them, and some countries are better covered than others. A country with few outlets in GDELT may not appear even when its press covered the story.
- A sample of the day. One batch per hour keeps the daily job fast. A story that broke and faded within an hour can be missed.
How it updates
Every morning a job on GitHub Actions reads the last 24 hours of GDELT news, picks the story, and publishes the site again. Nobody chooses the story or edits the results by hand.