Study: Over a Third of WildChat Prompts Are Fiction

The study arrives amid a run of publishing scandals over the suspected use of AI by professional authors: one publisher canceled a $2 million book deal, another withdrew a novel from the shelves, and readers alleged a prize-winning short story was AI-generated. Those episodes reinforced the assumption that machine-written fiction has no audience. Walsh and her colleagues argue that the commotion has obscured what readers are actually doing on their own.

The researchers' evidence comes from the WildChat dataset, a public collection of prompts whose anonymized nature makes demographic analysis impossible, and whose participants had to agree to make their prompts public. OpenAI told WIRED the set is not a representative sample of all ChatGPT users. The gap between the study's estimate and OpenAI's 1.4 percent figure—which does not include role-play—likely hinges on how each side defines fiction and on the dataset's volunteer skew.

The stakes, then, may rest less on whether the true share is 7 percent or 1.4 percent than on what a committed minority of readers does with the tools. The researchers sketch a possible tiered publishing system in which a large quantity of AI-generated and self-generated fiction satisfies the median reader while a smaller rung of human-authored work serves those seeking experimentation in form. For publishers, the test is whether they treat that pattern as a threat or a market.