The Tuesday series

The Verified Number of the Week

Every week, we take the most widely repeated AI number, go back to the primary source — the study, the press release, the regulatory filing — and publish the gap between what the source says and what the web made of it. The number is almost always accurate. The story, rarely.

Next issue: Tuesday.

This week · No. 5
0,57 pt sur 7

What the Blind Test on AI-Written Fiction Actually Shows

1.54 versus 0.97, on a scale that runs from minus three to plus three. Both scores are positive, and nobody could reliably tell which story came from a machine.

Our method is public. The number is accurate until proven otherwise: what we check is the story. And if the source confirms the coverage, there is no issue.

Read the method →