Research reveals no 'AI penalty' in short fiction, but effort perception shifts drastically

A new study by Michael Todasco and Joselyn Cesare, titled 'Know Your Author: Does the AI Penalty Hold in Short Fiction?' (arXiv:2606.00006), challenges the widespread assumption that labeling content as AI-generated automatically lowers its perceived quality. In a pre-registered experiment with 254 participants, each person read a ~200-word vignette randomly labeled as 'Human-written,' 'AI-written,' or presented with no author line. The results showed no reliable main effects on creativity, enjoyment, recommendation, or originality—effect sizes were uniformly small. This suggests that simply telling readers a story was written by AI does not, on average, cause a negative bias in how the story is evaluated.

However, the labels had a dramatic impact on perceived effort: participants estimated that human-labeled stories took an average of 148 minutes to create, compared to just 6 minutes for AI-labeled stories. Across all conditions, higher inferred effort predicted greater enjoyment, and this relationship held even within the AI-labeled group. Additionally, readers' prior attitudes toward AI moderated their recommendation ratings: those with more positive AI views gave higher recommendations to AI-labeled stories, but not to human-labeled ones. These findings imply that while AI labels don't inherently devalue short fiction, they powerfully shape perceptions of author effort and interact with individual beliefs to influence downstream judgments.