Villanova Study Finds AI-Generated Short Stories Rated Higher Than Human Ones

Villanova University AI-generated stories human vs AI story quality study ChatGPT

A study conducted by researchers at Villanova University, including senior author Deena Skolnick Weisberg, examined how U.S. adults evaluate short fiction written by humans versus artificial‑intelligence systems such as ChatGPT. In the primary experiment, 1,682 participants read six short stories—three human‑written and three AI‑generated—and rated each for quality and engagement. The AI stories received higher scores, about 6 % better on quality and 8 % higher on engagement. When participants were told the stories were authored by humans, the AI pieces were rated an additional 3 % higher, indicating a bias toward perceived human authorship.

Two follow‑up experiments tested participants’ ability to identify the true origin of the stories. Only 39 % to 52 % of respondents correctly distinguished AI‑generated from human‑written texts, showing that most readers cannot reliably tell the difference. Familiarity with AI systems modestly improved recognition, while self‑identified literary expertise did not. The findings were published in the journal *Judgment and Decision Making* and suggest that readers may prefer the clearer, more direct style of AI‑generated narratives, even when they believe the work is human.

The researchers note that the results highlight a gap between public assumptions about creative writing and the actual capabilities of modern language models, and they propose that increased AI literacy could help people navigate an increasingly AI‑enabled media environment.

Entities

ChatGPT · Deena Skolnick Weisberg · Deena Weisberg · Judgement and Decision Making · Judgment and Decision Making · Sydney Sears · Villanova University

Claims

What the coverage asserts, and how many sources carry each claim.

[● 2 SOURCES] AI‑generated stories were rated about 6% higher in quality than human‑written stories. time.com · ipaddisti.it
[● 2 SOURCES] Participants correctly identified AI‑generated stories 39% of the time in the first experiment and 52% in the second. time.com · www.fayerwayer.com
[○ 1 SOURCE] AI‑generated stories were rated about 8% more engaging than human‑written stories. time.com
[○ 1 SOURCE] When participants were told the stories were written by humans, AI stories received an additional 3% higher rating. time.com
[○ 1 SOURCE] AI‑generated stories received an average engagement score of 1.42, higher than human stories. ipaddisti.it
[○ 1 SOURCE] AI‑generated stories received an average quality score of 1.54 on a –3 to 3 scale, versus 0.97 for human stories. ipaddisti.it
[● 3 SOURCES] The study involved 1,682 adult participants. time.com · ipaddisti.it · www.fayerwayer.com
[● 3 SOURCES] The study was led by Deena Weisberg and Sydney Sears at Villanova University and published in the journal Judgement and Decision Making. time.com · ipaddisti.it · www.fayerwayer.com