Study Finds AI Short Stories Beat Human Ones in 1,682-Reader Test
Updated
Updated · The Guardian · Aug 4
Study Finds AI Short Stories Beat Human Ones in 1,682-Reader Test
3 articles · Updated · The Guardian · Aug 4
Summary
1,682 adults rated ChatGPT-written short stories as more absorbing and higher quality than human-written versions in a study published in Judgment and Decision Making.
Three AI stories were matched by theme with three human stories, and readers were sometimes misled about authorship; stories labeled human still received higher scores, suggesting authenticity bias.
Two follow-up experiments with 905 adults found readers generally could not tell the difference: one group identified the human author only 40% of the time, another 52%.
Researchers said positive attitudes toward AI lifted ratings for stories presented as AI-written, while greater AI expertise—not fiction expertise—improved detection of AI’s stylistic patterns.
The findings point to AI as a credible fiction generator even as the study's authors defended human creativity and outside critics argued the technology carries broader ethical and environmental costs.