Most people can't tell AI-generated fiction from human-written stories - and when they can, they often rate the AI version higher. That's the finding of a new study from Cambridge University Press and researchers at Villanova University, which asked 1,682 participants aged 18 to 81 to read short stories and judge both their authorship and quality.
The study ran three experiments. In the first, participants were told a story was either AI-generated or human-written before rating it - and the AI stories scored higher, especially when falsely labeled as human work. In the next two experiments, participants received one human story and one ChatGPT story with no disclosure. Over 60% in one group and over 48% in the other failed to identify which was AI-generated.
Why AI writing wins points
Deena Weisberg, senior author of the study and professor of psychological and brain sciences at Villanova, said the results reveal a bias toward "narratives written by real people." She noted that people assume creative writing requires "emotional understanding and lived experience," which leads them to underestimate AI's capabilities.
Weisberg attributes the preference to style. AI-written stories tend to be more direct, explicit, and simple, while human stories leave room for interpretation and rely on subtler literary devices. One AI passage from the study reads: "The koi continue their dance beneath the water's surface, oblivious to the change they have witnessed in my life." A human passage, by contrast, meanders through metaphor: "We swim in uneasy circles around each other, and all the while the golden sun continues to rise and set, keeping its own mysterious time."
That difference isn't accidental. AI systems are trained on writing that humans have already enjoyed for decades. "It's more that these technologies amplify existing tendencies, rather than creating them," Weisberg said. For AI for Writers, the takeaway is that the tools are tapping into familiar narrative patterns, not inventing new ones.
What the research means for writers
This study aligns with earlier findings. A 2023 MIT-led study by Renee Richardson Gosline found that AI-generated and AI-assisted content was rated higher in quality than work by human experts. The pattern holds across short stories, essays, and poetry.
For writers, the implication isn't that AI writes better - it's that readers reward clarity and directness. Human writing often carries complexity that requires more effort to parse. AI, trained on what people already like, delivers the stylistic equivalent of comfort food.
Why this matters for writers
If readers can't reliably distinguish AI text and often prefer it, the value of human authorship shifts. The craft advantage isn't in producing simpler prose - that's now commoditized. It's in the lived experience, emotional depth, and interpretive ambiguity that AI can imitate but not originate. Writers who lean into those human qualities, rather than competing on polished surface style, keep the differentiation that matters.
That's a useful lens for AI for Creatives as well: the goal isn't to mimic AI output, but to understand what readers respond to and where human voice still carries weight.
Your membership also unlocks:
