Chatbot Fiction Is Booming, and Readers Don’t Seem to Care

People are using chatbots to write custom fiction at scale, and a new study suggests the audience is larger and more committed than many in publishing assume. Researchers found that a substantial share of chatbot prompts involve storytelling, role-play, fan fiction and erotica, pointing to a growing appetite for AI-generated narratives in 2026 and beyond.

The findings matter because they challenge one of the culture war assumptions surrounding artificial intelligence: that machine-made fiction is unwanted, disposable or inherently inferior. For a notable slice of users, the appeal is not despite the AI element, but because of it.

Petra Ferraz de Novaes, a 36-year-old software engineer in Brazil, is one of them. When she is not working, she often turns to AI-assisted storytelling tools to build fantasy worlds, improvising alongside the model and following its unexpected turns. What began as an experiment in 2024 has become a regular creative routine.

Novaes says the appeal lies in unpredictability. She likes that the model does not always obey her exact instructions, because the odd detours create narrative possibilities she might not have imagined on her own. In her view, the model’s tendency to “hallucinate” is not a bug in this context — it is part of the draw.

Her experience reflects a broader shift in how people are using large language models. Beyond work tasks and quick answers, a growing number of users are turning chatbots into collaborative fiction engines, personalized role-play partners and on-demand improvisation machines. That use case is still far from the mainstream image of AI, but the data now suggests it is more common than many people realize.

What the new research found

Researchers analyzing more than 500,000 prompts from a public ChatGPT dataset called WildChat found that fiction-related requests made up a large slice of chatbot use. Their study identified prose, poetry, fan fiction, erotica and role-play among the most common creative categories.

The biggest takeaway is not just that people experiment with fiction once or twice. It is that a comparatively small group of heavy users appears to drive much of the activity, repeatedly refining stories, revisiting characters and iterating on the same prompts many times over.

In one example cited by the researchers, a single user returned to ChatGPT thousands of times over several months to generate variants of the same fan-fiction premise set in the universe of Doki Doki Literature Club! That kind of sustained engagement suggests these tools are not simply novelty gadgets. For some users, they have become a steady creative habit.

Even after accounting for the fact that power users skew the totals, the researchers estimate that around 7 percent of chatbot users engage in fiction generation in some form.

Why are people turning to AI for fiction?

People are turning to AI for fiction because it gives them speed, control, privacy and endless revision without judgment. That combination is difficult for traditional writers, publishers or fan communities to match.

The researchers argue that those qualities make chatbot fiction appealing for reasons that go beyond cost or convenience. A user can ask for a story immediately, revise it without friction, change the tone on command and keep going indefinitely. For many users, the experience feels less like commissioning literature and more like steering a custom entertainment engine.

Melanie Walsh, an assistant professor at the University of Washington Information School and one of the study’s coauthors, says the key attraction is not that users think AI stories are brilliant literature. It is that they know what they are getting and are fine with it.

Walsh says users are often fully aware that the stories are machine-generated and do not seem to care. In her view, that indifference is itself revealing, because it suggests a meaningful audience is happy to trade literary prestige for direct personal satisfaction.

The study’s authors say this challenges the assumption that people want fiction to come only from human imagination. Instead, many users appear to value a story that adapts to them, rather than one shaped by an author’s independent artistic judgment.

How does AI fiction differ from conventional writing?

AI fiction differs from conventional writing because it is interactive, immediate and highly customizable. Instead of consuming a finished work from an author, users can co-create the plot, alter characters on the fly and keep regenerating scenes until they get the experience they want.

That flexibility makes the format especially useful for people who want stories built around specific premises that may never be published commercially. It can also blur the line between writing, gaming and role-play.

• Stories can be generated instantly.
• Characters and settings can be revised repeatedly.
• Users can direct the narrative direction in real time.
• Outputs can be private, personalized and highly specific.

The erotica and role-play market inside chatbot fiction

Erotica and role-play appear to be especially strong drivers of chatbot fiction use. For some users, AI enables a form of intimate, personalized fantasy that is difficult to replicate in conventional pornography or self-published erotica.

One user interviewed for the story, who goes by the username Razrien, described using AI to create dominant fantasy characters and pursue highly specific sexual role-play scenarios. He said the tools let him explore desires that would otherwise be hard to satisfy through ordinary media.

Razrien said the experience feels like an endless choose-your-own-adventure story, except the story can be rerun, redirected and intensified without limit. He added that he has tested different models to find those that best handle his preferred fantasies.

According to Razrien, one model from Google’s Gemma family is particularly useful for the kinds of scenes he wants to generate. His comments highlight a broader point: users are not only choosing AI over human writing, they are also learning which models do certain genres better than others.

That kind of model shopping may become more common as the ecosystem grows. Users who want speculative fiction, fandom-based stories or erotica may gravitate toward systems that better sustain tone and character behavior, even if the output is still imperfect.

What makes erotic role-play so popular?

Erotic role-play is popular because it is interactive, responsive and deeply customizable. Users can create characters, insert themselves into scenes and control the pace and boundaries of the interaction in ways that static media cannot easily provide.

For some, that is about sexual fantasy. For others, it is about emotional companionship, self-expression or experimenting safely with identity and desire.

• It offers highly specific personalization.
• It allows users to repeat and refine scenes.
• It can include conversation, not just text description.
• It avoids many of the limits of traditional adult content.

How reliable is the data?

The data is informative, but it is not definitive. The WildChat dataset only captures prompts from people who agreed to make their interactions public, which means it may not represent the full population of ChatGPT users.

It also cannot show who those users are, because the prompts are anonymized. That makes it impossible to say whether fiction generation is more common among younger users, women, specific regions or particular hobby communities.

OpenAI told WIRED that WildChat is not a representative sample of all ChatGPT users. The company also has its own numbers, and they are significantly lower. A 2025 study by OpenAI and the National Bureau of Economic Research found that only 1.4 percent of ChatGPT prompts were fiction-related.

The gap between the two estimates does not necessarily mean one side is wrong. It may reflect differences in sampling, classification or what exactly is counted as fiction. For example, the researchers behind the new study included role-play in their fiction tally, while OpenAI’s figure did not. The company also could not confirm whether it counts erotica in the same category.

Still, even the lower estimate would point to a meaningful amount of creative use. And both numbers support the same broad conclusion: fiction is not an edge-case behavior in chatbot systems.

What do the researchers think this means for literature?

The researchers think AI fiction may force a rethink of what readers actually value. For some people, literature is not primarily a sacred exchange between author and audience. It is a service, an experience or a source of entertainment that should meet the reader’s needs efficiently.

Neel Gupta, a University of Washington doctoral researcher and coauthor of the study, says AI use exposes a more transactional relationship between reader and text than many people in literary culture are comfortable admitting.

Gupta argues that the data suggests some readers are not seeking formal innovation or artistic ambiguity. They are looking for entertainment that fits their preferences, and AI can supply that almost instantly.

That does not mean human writing is losing relevance. It may mean that human writing and AI writing are serving different audiences, and possibly different emotional or intellectual needs.

One likely outcome, the researchers suggest, is a more stratified market: a large volume of AI-generated stories for readers who want custom entertainment, and a smaller, more specialized space for human authors who prize experimentation, complexity or style.

That future would not necessarily eliminate literary culture. But it could narrow the market for books that depend on readers being willing to wrestle with ambiguity, dense prose or unfamiliar narrative structures.

Could AI fiction become better than it is now?

Yes. The authors say the biggest barrier today is not interest but capability, and model quality is improving quickly.

Current systems still struggle with extended structure, stylistic consistency, original characterization and ornate prose. They tend to perform better with formulaic genres that rely on recognizable tropes and familiar arcs. That is one reason fantasy, fan fiction and erotica are showing up so often in chatbot use: those categories are already built around conventions that models can imitate well enough for many users.

Where the systems still break down is in long-form coherence. They can produce lively scenes and quick exchanges, but maintaining a complex plot over a novel-length arc remains difficult.

If those weaknesses fade, the market for AI fiction could grow substantially. Better memory, improved reasoning and stronger long-range planning would allow models to sustain more elaborate books and richer character development.

That could widen the audience beyond hobbyists and role-play enthusiasts. It could also intensify concerns among writers, editors and publishers who already fear that generative AI could flood the market with low-cost, low-effort content.

What does the future publishing landscape look like?

The future publishing landscape may split into two tiers: mass-market AI entertainment on one side and specialized human literature on the other. In that scenario, the average reader could get faster, cheaper and more personalized stories, while a smaller audience continues to seek work that feels formally daring or culturally prestigious.

The study’s authors are not necessarily celebrating that outcome. But they do suggest it may be realistic.

• AI-generated books could dominate quantity.
• Human-authored fiction may move further into a niche.
• Readers may increasingly choose between customization and artistry.
• Publishing may need new labels, standards and business models.

How the trend affects writers, publishers and platforms

For writers, the rise of chatbot fiction is both a threat and an opportunity. It threatens the assumption that readers will only pay for polished human prose, but it also highlights demand for strong storytelling across formats.

For publishers, the more immediate challenge may be trust. Recent controversies involving suspected AI use in manuscripts have already damaged confidence in the industry. Several high-profile cases have led to pulled books, canceled deals and accusations that authors or publishers misled readers.

That backlash has centered on production ethics, originality and disclosure. But the new research suggests that the demand side is more complicated than the debate often assumes. If readers are already using AI fiction enthusiastically in private, public criticism of AI writing may not reflect real consumer behavior.

For platforms, the issue is discoverability and moderation. Tools that support role-play and custom storytelling may need to address sexual content, copyright concerns and harmful prompts while still preserving the flexibility users want.

For AI companies, fiction use cases may become a meaningful design target. A model that can maintain characters, preserve tone and remember user preferences across sessions could become more valuable to hobbyist storytellers than one that merely answers questions well.

Why this matters beyond the culture war

This story matters because it shows AI is not only competing with human labor in offices, schools and coding tasks. It is also becoming part of how people imagine, fantasize and entertain themselves.

That shift has implications for product design, moderation, intellectual property, publishing economics and even cultural taste. If readers grow accustomed to fiction that adapts to them in real time, the definition of what counts as a story may broaden.

At the same time, there are reasons to be cautious. The study is based on a public dataset with clear limitations. The estimates may not hold across platforms or user groups. And the popularity of AI fiction does not mean users will abandon human books, editors or authors.

What it does show is that the market for AI storytelling is larger than the most skeptical commentators believed. Some users are not just tolerating chatbot fiction. They are actively seeking it out, shaping it and returning to it again and again.

For Novaes, the technology has changed the way she reads as well as the way she writes. She began using AI to build her own stories, but now she also reads what other people create with it. In her case, the tool has become a gateway rather than a replacement.

That may be the most important conclusion of all: for at least some users, AI fiction is not killing reading. It is creating a new way into it.

Timeline of the key developments

Bottom line

The emerging AI fiction audience is not a novelty act. It is a real, if hard-to-measure, community of readers and improvisers who want stories tailored to their own tastes, however strange or specific those tastes may be.

That demand could reshape how companies build models, how publishers think about storytelling and how readers define literature in an era of infinite generated text.