Study finds a third of ChatGPT prompts involve fiction or role-play

University of Washington researchers analysed 500,000 ChatGPT prompts and found a small group of power users driving heavy demand for AI-generated stories and erotica.

Key points

  • A study of over 500,000 ChatGPT prompts found more than a third involve fiction, role-play or erotica.
  • Researchers Melanie Walsh and Neel Gupta estimate 7 percent of users engage in fiction generation, far above OpenAI's 1.4 percent figure.
  • The researchers argue a committed group of readers is content with AI-written fiction, which cuts against assumptions behind recent publishing scandals.

A study of more than 500,000 ChatGPT prompts has found that more than a third involve some form of fiction generation, from prose and poetry to fan fiction, erotica and role-play, according to Wired. The researchers argue that a committed subsection of readers is now generating custom, personal fiction.

The study was coauthored by Melanie Walsh, an assistant professor at the University of Washington Information School, and Neel Gupta, a PhD candidate there, Wired reported. They analyzed prompts drawn from WildChat, a public dataset of ChatGPT interactions, and found that most fiction-related activity came from a small group of "power users" who generate and revise stories obsessively. In one case, Wired reported, a single user prompted ChatGPT to produce versions of the same fan-fiction narrative, set in the world of the anime video game Doki Doki Literature Club!, thousands of times over several months.

Even controlling for that kind of heavy use, the researchers estimate that 7 percent of chatbot users engage in fiction generation, Wired reported. That figure is far higher than one produced by a separate 2025 study from OpenAI and the National Bureau of Economic Research, which put fiction-related prompts at 1.4 percent of all ChatGPT activity. Wired noted that OpenAI's figure excludes role-play, and the company could not confirm whether it counts erotica.

OpenAI told Wired that "WildChat is not a representative sample of all ChatGPT users." The dataset's prompts are anonymized, so no demographic data is available, and participants had to agree to make their prompts public, which may skew who is represented.

Readers who don't mind the machine

Wired's reporting centers on people already living this shift. Petra Ferraz de Novaes, a 36-year-old software engineer from Brazil, told Wired she began asking AI models to build out her story ideas in 2024, after finding every decision that goes into writing, from style and syntax to character development, too much to manage alone. She now uses a program called SillyTavern to run open-source models on her own computer, sketching a premise and letting the model fill in the rest.

Novaes told Wired she moves between multiple fantasy worlds in a single session, some original and some borrowed from Star Wars, Pokémon or Mass Effect, sometimes producing self-contained stories the length of a small novel. "It can be never-ending. You find a quest, you complete it, you go for the next one," she told Wired. "The limit is your imagination."

Walsh told Wired that readers in the study generally know the fiction they're consuming was machine-made and don't mind. "They know [the stories are] produced by an AI model, and they don't care," Walsh said. "In fact, that seems to be a big part of the attraction."

For erotica in particular, Wired reported, AI lets readers insert themselves directly into invented scenarios. Razrien, a 40-year-old man from Indiana who spoke to Wired under his online username, described building fantasy characters for role-play, including a "big angry rat wife" and an "eight-foot tall werewolf mommy with a bad temper." He told Wired he has learned which AI models suit which kinks, currently favoring Google's Gemma 4. "It's a digital choose-your-own-adventure book but with infinite swipes," he said.

What this means for authors

Wired placed the study against a backdrop of recent publishing scandals involving suspected AI use by professional authors, including a canceled $2 million book deal, a novel withdrawn from shelves, and reader allegations that a prize-winning short story was AI-generated. Those incidents, Wired noted, carry an implicit assumption that readers don't want machine-written fiction. The study's authors argue that assumption overlooks what a segment of readers are already doing on their own.

Walsh told Wired that AI models have qualities that differ from those of human writers and that appeal to at least some users. "These models are judgment-free; they can produce stories instantaneously; they are nearly inexhaustible," she said.

Gupta told Wired that widespread AI fiction use complicates assumptions about why people read at all. "AI use helps to reveal that maybe for a lot of readers, this is a much more simple economic exchange. This is about getting some entertainment—getting what they want," Gupta said.

Wired reported that current AI models still struggle with ornate prose, complex characters and coherent narrative at book length, performing best in genres built on familiar tropes. The researchers told Wired that a tiered publishing system could emerge as models improve: a large volume of AI-generated fiction serving mass entertainment, alongside a smaller field of human-authored work aimed at readers seeking formal experimentation or ambiguity. "English majors and professors, and some people in publishing, have always had kind of a niche interest," Walsh told Wired.

Novaes told Wired that AI-assisted fiction has drawn her toward reading rather than away from it. She now seeks out other people's AI-generated stories online. "I was never much of a book reader," she said. "But now I do read other people's stories. These are the only novels I'm reading."

Sources

  1. ‘I Like My Big Rat Wife’: Meet the People Using Chatbots to Write Custom Fiction — Wired,

Every quotation and figure in this article is checked against the source reporting linked above before publication. How we report