Readers Prefer AI Short Stories Blindly, But Scores Drop Once Revealed

The Decoder · rss · 2026-08-08

A new study reveals that people cannot effectively distinguish ChatGPT-generated short stories from human-written ones. In a test involving over 2,500 participants, accuracy in identifying AI text was no better than chance.

Interestingly, participants actually rated the quality of AI-generated texts higher when unaware of the authorship. However, scores dropped significantly the moment participants learned the stories were written by a machine.

Original post →

More from AGI Musings

AGI Musings channel →