AI-Generated Short Stories Outscore Human Writing in New Study, but the ‘Human’ Label Still Pays
A study published in the journal Judgment and Decision Making challenges the common assumption that AI cannot write well. In an experiment involving 1,682 adult readers, short stories generated by ChatGPT were rated more absorbing and of higher quality than stories written by humans.
- A study in Judgment and Decision Making found ChatGPT-written short stories rated more absorbing and higher in quality than human-written ones by 1,682 readers
- Stories labeled as human-written scored higher regardless of their actual author, showing a premium on the human label
- Lead researcher Deena Skolnick Weisberg says views of AI’s abilities should be updated, but human writing’s value as self-expression remains
How the Experiment Worked
Researchers prepared six short stories on matched themes, three written by human authors and three by ChatGPT. Participants were told the authorship of each story, but the information was not always correct. Two findings stand out. Readers who actually read AI-written stories gave them higher ratings for absorption and quality than those who read human-written stories. But stories labeled as human-written scored higher regardless of who really wrote them, meaning the human-author label itself carries a premium.
Deena Skolnick Weisberg, a senior author of the study at Villanova University, said AI systems can already generate short stories seen as at least as good as, if not better than, human-written ones, and that views of AI’s abilities should be updated accordingly. She added, however, that this does not mean writing should be left to AI. Human writing is a form of creative self-expression, a way to challenge ourselves and make sense of experience, and AI producing human-like stories does not change that.
TechApple’s Take: The Empirical Counterpoint to ‘AI Ruins Writing’
The study works as an empirical counterpoint to recent opinion pieces arguing that AI destroys writing, such as Bret Stephens’ New York Times column “I’m Begging You: Never Write With A.I.” On raw content quality, AI short fiction has caught up with or surpassed human output. Yet the human label still commands a premium, suggesting readers value the creative process and author identity more than the text itself.
For the content industry, the lesson is straightforward. Using AI to raise output and efficiency is the practical reality, but brand value, author identity and creative context remain the differentiators that cannot be replaced. The experiment has limits: short stories are not novels, and a rating exercise does not capture real reading experience. It should not be read as proof that writers are about to be replaced.
Source: The Guardian
