AI-written short stories rated higher quality than human ones in new study
A study of 1,682 adults found that stories written by ChatGPT scored better on quality than comparable human-authored work. Researchers think simpler, cleaner writing may be driving the result.
Key points
- A peer-reviewed study published in Judgment and Decision Making asked 1,682 adults to rate six short stories, three written by humans and three by ChatGPT.
- Readers consistently rated the AI-written stories as higher quality than the human-written stories on similar themes.
- Researchers believe ChatGPT's plainer, more direct prose may make its stories easier to follow, which readers interpret as better writing.
- The finding does not mean human authors are obsolete; researchers note that simplicity and literary quality are not the same thing.
From Homer to Agatha Christie, humans have always owned the story. A new study suggests that may be getting complicated.
Researchers recruited 1,682 adults and gave each person one of six short stories to read. Three came from human writers. Three came from ChatGPT, the AI chatbot made by OpenAI, with each AI story matched in theme to one of the human pieces. Participants then rated what they had read.
The AI stories won.
Readers rated the ChatGPT stories as higher quality than their human-written counterparts. The study, reported by The Guardian, appeared in the peer-reviewed journal Judgment and Decision Making.
Why would simpler writing score higher?
Researchers think the gap comes down to readability. ChatGPT tends to produce clean, direct sentences with few stumbling blocks. Human literary fiction often uses complex imagery, ambiguous phrasing, and structural risk-taking, choices that reward close reading but can feel harder on first pass.
A reader skimming a short story in a study setting may simply find the smoother text more satisfying. That is not the same as finding it more meaningful, moving, or original.
The researchers were careful to say the result does not make human writers redundant. Ease of reading and literary depth pull in different directions, and a study measuring one is not measuring the other.
What does this mean for everyday readers?
For most people, nothing changes today. The study used short, controlled pieces, not novels, screenplays, or the kind of long-form work that builds a readership over years. What it does suggest is that AI text can clear a basic quality bar that readers notice.
That matters most in areas where writing is functional: product descriptions, news summaries, instructional content. In those spaces, "good enough to read easily" is often the whole job.
For readers who love literary fiction, the nuance, the surprise, the author's singular voice, the study offers little threat. Those qualities are exactly what the simpler AI prose tends to flatten out.
Common questions
Does this mean AI will replace novelists?
Not based on this evidence. The study tested short, theme-matched stories on a single readability-style measure, not the sustained originality or emotional depth that defines literary careers.
Should I trust AI-written stories I encounter online?
It is worth knowing they exist. AI text can read fluently and still be factually thin or creatively generic. Reading a second piece by the same source, human or AI, usually tells you more than a first impression does.
