ChatGPT Stories Rated Higher Than Human Fiction — Until Readers Learn the Author

A Cambridge study with 2,500 participants found that short stories written by ChatGPT 4.0 were rated higher than work from top literary journals — 1.54 vs. 0.97 for quality and 1.42 vs. 1.00 for immersion. Participants were unable to tell AI-written text from human-written: their accuracy was at chance level.

Once authorship was revealed, ratings for the very same texts dropped. This means that the bias against AI-generated content rests on the label, not the actual quality — and as that label fades, the line between literary fiction and generative text will blur faster than it might seem.