Humans have been spinning yarns since the days of Homer. Agatha Christie didn’t use algorithms. Yet a new study suggests a strange twist in the tale: readers actually judge AI-written stories as better quality than those penned by people.
It feels counterintuitive. You buy a book for the human hand, the sweat, the soul, right? But the data says otherwise.
Researchers published their findings in Judgment and Decision Making. They gathered 1,682 adult participants. Each person read one of six short stories. Half the stories came from ChatGPT. The other half were human-authored. The themes matched up—one human story mirrored one AI story in tone and subject.
Then, the researchers got tricky.
They told participants whether the story was human or AI-made. Sometimes they were right. Sometimes they lied. Then came the rating. Participants scored the stories on absorption, engagement, and overall quality.
The results were stark. When participants believed a story was AI-generated, they rated it higher. It was more absorbing. Better crafted. The text was cleaner. The plot moved faster.
But flip the script. When told the same text was written by a human, the ratings dropped.
Dr. Deena Skolnick Weisberg from Villanova University led the study. She is also a creative writer herself. She notes the disconnect clearly.
“AI systems can already generate short stories seen as being at least as good if not better than human-written stories.”
This isn’t just about better grammar. It’s about readability. AI models are optimized to be smooth. To be clear. To avoid the messy tangents humans often include when they are thinking out loud. We prefer the polish.
Yet, Weisberg argues, this doesn’t mean humans should step aside.
We write for reasons AI cannot replicate. Creative self-expression. Challenging limits. Making sense of chaotic experiences. The process matters as much as the product. If we only care about the final page, we lose what makes writing human.
The study also revealed how bias plays a role. Participants with positive attitudes toward AI rated those stories higher. Those skeptical of technology? Less so. Expectations shape reality.
Can You Spot the Difference?
The researchers didn’t stop at ratings. They tested discernment.
905 new participants read two stories. Same theme. One human. One AI. They had to guess which was which.
In one test, about 40% guessed correctly. That’s worse than random chance. In another, 52% succeeded. Barely better than flipping a coin.
There is no reliable general evidence that readers can distinguish human-written text from AI-generated fiction.
Why?
“AI writing has a particular style we can learn to recognize with practice.”
Weisberg suggests we can train our eyes to see the AI signature, much like we spot a favorite author’s quirks. But right now? The lines are blurred. The AI stories are often tighter. More structured. Less prone to error.
The preference for human stories when labeled as such ties to a desire for authenticity. We want to know someone real felt something. But when we strip away the label, we prefer the easier read. The AI text is digestible. Efficient.
The Ethical Weight Behind the Words
Not everyone sees this as progress.
Luke Kennard, a film and creative writing professor at the University of Birmingham, calls the AI debate highly contentious. He pushes back against the notion that AI is just a better pen.
“We tend to personify AI. Imagine it as some glowing robot with a benign face. It’s not. It’s predictive code based on a massive database of stolen actual writing.”
Kennard highlights the physical cost. Data centers. Energy consumption. Environmental destruction. Communities displaced by tech infrastructure. These aren’t abstract concerns. They are real.
He argues against calling AI a tool. Spellcheck is a tool. A thesaurus is a tool. AI is something else. An existential threat to the concept of authorship itself.
Sure, you can prompt it into spitting out a coherent poem. Of course it can. It was trained on thousands of years of human creativity. It would be embarrassing if it couldn’t mimic us.
But should it?
Kennard frames the question simply: Is it worth it?
His answer is categorical. No.
The study shows AI can mimic the surface of storytelling. It can produce text that is rated higher for quality and engagement. But it misses the context. The labor. The cost. The intent.
We are left with a strange paradox. We like the result, but we distrust the source. We want the efficiency, but we crave the authenticity.
The stories are good. Maybe too good.
Does that make them better? Or just different in a way we aren’t ready to accept?
The data says we prefer the AI tale. Our bias says we trust the human. Which one wins?
Probably neither. We just keep reading. And wondering.
