Why AI-generated text is easy to spot: lcamtuf shows 150 kids' books with the same title

The 100k Whys of AI

lcamtuf searched Amazon for “100000 whys” and found ~150 children's books with near-identical covers: same title, roaring dinosaur in the top-left corner, recurring red-and-white rockets, golden retrievers, and lions. The issue isn't that a single model sounds inhuman—it's that different users with similar prompts get ~80% identical output. This quasi-determinism is the real fingerprint of AI text. He argues you can trust your gut when reading casual content, because producing slop now costs far less than consuming it. The post does not name specific models or generation dates.

Why it matters: lcamtuf runs a sharp empirical demo with 150 AI-generated kids' books on Amazon, proposing 'quasi-determinism' as a detection method that's more convincing than statistical debates. The piece has concrete examples, a method, and a clear stance—hits all three HKR axes. Score he...