The traditional boundary between author and reader is undergoing a fundamental transformation as generative artificial intelligence moves from the fringes of technical experimentation into the heart of personal entertainment. While the professional publishing industry grapples with the ethical and economic threats posed by large language models (LLMs), a growing demographic of users is bypassing traditional media entirely to co-create bespoke narratives that cater to their most specific, and sometimes idiosyncratic, desires. For these individuals, the value of a story is no longer measured by the prestige of a human author, but by the immediate, iterative, and judgment-free responsiveness of the machine.
The Shift Toward Collaborative Co-Creation
Petra Ferraz de Novaes, a 36-year-old software engineer based in Brazil, exemplifies this new breed of "power user." For Novaes, the barrier to storytelling was never a lack of imagination, but the granular labor of prose—the syntax, the pacing, and the relentless decision-making required to move a character from point A to point B. In 2024, she turned to AI to bridge that gap, discovering that the models’ tendency toward "hallucination"—the generation of unexpected or factually incorrect information—was not a bug, but a feature for creative endeavors.
Novaes utilizes specialized interfaces like SillyTavern to run open-source models locally on her computer. This setup allows her to maintain privacy and bypass the restrictive safety filters often found in commercial chatbots. Her sessions involve a blend of original concepts—such as elven warriors battling Amazonian queens—and established intellectual properties like Star Wars and Pokémon. The result is a never-ending, interactive quest where the AI fills in narrative gaps, often taking the story in directions Novaes herself had not anticipated.
This phenomenon is not isolated. A recent study by researchers at the University of Washington Information School suggests that a significant portion of AI interaction is now dedicated to fiction. "They know the stories are produced by an AI model, and they don’t care," notes Melanie Walsh, an assistant professor and coauthor of the study. "In fact, that seems to be a big part of the attraction."
A Chronology of AI in Literature and Recent Industry Friction
The integration of AI into the literary world has followed a rapid and contentious timeline. While early experiments in computer-generated text date back to the mid-20th century, the current era of "AI fiction" began in earnest with the release of OpenAI’s GPT-2 and GPT-3 between 2019 and 2020. However, the friction between human creators and machines reached a boiling point in 2023 and 2024.
- January 2024: Japanese author Rie Kudan won the prestigious Akutagawa Prize, later admitting that roughly 5 percent of her novel, The Tokyo Tower of Sympathy, was generated by ChatGPT.
- May 2024: A major publisher canceled a $2 million book deal following allegations that the author had used AI to supplement their work without disclosure.
- Mid-2024: Several literary journals, including Granta and Clarkesworld, faced backlash or were forced to temporarily close submissions due to a flood of AI-generated stories, leading to a broader industry debate on the "devaluation" of the art form.
Implicit in these scandals is the assumption that readers view AI-generated content as inherently inferior or fraudulent. However, the data emerging from user interactions suggests a parallel market where the "human touch" is secondary to personal gratification.
Analyzing the Data: The "WildChat" Findings
To understand the scale of this shift, researchers analyzed the WildChat dataset, which consists of over 500,000 prompts fed into ChatGPT. The findings, published in a 2024 study, indicate that more than a third of chatbot interactions involve fiction generation, including prose, poetry, fan fiction, and role-play.
The data highlights a "power user" dynamic. A small subset of users interacts with the models obsessively; the study cited one instance of a user generating thousands of variations of a single fan-fiction narrative set in the world of the video game Doki Doki Literature Club! over several months. Even when controlling for these extreme cases, researchers estimate that at least 7 percent of the broader user base engages in fiction generation.
There is, however, a discrepancy in the data. A 2025 study by OpenAI and the National Bureau of Economic Research (NBER) estimated that only 1.4 percent of ChatGPT prompts are fiction-related. This gap is likely due to definitions; the OpenAI study may not have categorized role-play or erotica—categories that often overlap with fiction—under the same umbrella. Furthermore, as users like Novaes move toward local, open-source models to avoid corporate censorship, their activity disappears from the datasets of major providers like OpenAI and Google.
The Allure of the Judgment-Free Machine
One of the primary drivers of AI fiction is the "judgment-free" nature of the interaction. This is particularly evident in the realm of erotica and hyper-specific role-play. For users like Razrien, a 40-year-old from Indiana, AI provides a space to explore sexual kinks that are either unavailable in mainstream pornography or too niche for human writers to produce at scale.
Razrien uses models like Google’s Gemma 4 to engage in role-play with fantasy characters. He notes that the ability to "direct the action" without being hemmed in by an author’s creative choices is a significant upgrade from traditional media. "It’s a digital choose-your-own-adventure book but with infinite swipes," he says. For Razrien and many others, the AI serves as a personalized entertainment service that prioritizes the user’s specific desires over narrative complexity or artistic merit.
Implications for the Future of Publishing and Literacy
The rise of AI-generated fiction challenges long-held idealistic notions about literature. Traditionally, storytelling has been viewed as a form of communication—a bridge between the soul of the writer and the mind of the reader. Neel Gupta, a PhD candidate at the University of Washington, argues that AI use reveals a more transactional reality for a large segment of the population. "This is about getting some entertainment—getting what they want," Gupta says.
This shift suggests a potential "tiered" future for the publishing industry:
- The Mass Market / AI Tier: A vast quantity of AI-generated or AI-assisted books that satisfy the "median reader" looking for genre fiction (romance, thrillers, erotica) that follows established tropes and predictable arcs.
- The Human / Prestige Tier: A smaller, niche market for human-authored fiction that emphasizes stylistic experimentation, narrative ambiguity, and the unique "voice" that current AI models still struggle to replicate.
Critics worry that this could lead to a "filter bubble" for storytelling, where readers are never challenged by perspectives other than their own. If a user can always direct the story to go exactly where they want, the transformative power of being surprised or unsettled by an author’s vision may be lost.
Conversely, there is evidence that AI fiction may actually increase literacy in certain demographics. Novaes, for instance, admits she was never a frequent reader of conventional novels. However, she now spends hours reading the AI-generated stories of others shared online. In this context, AI is not replacing reading; it is providing a new entry point into it.
Technical Limitations and the Path Ahead
Current LLMs still face significant hurdles in long-form storytelling. They struggle with "context windows"—the amount of text they can remember at one time—which often leads to narrative inconsistencies or characters "forgetting" key plot points in novel-length works. Furthermore, while they excel at tropes, their prose often lacks the "ornate flourishes" and "complex character development" that define literary excellence.
However, the technology is evolving. With the advent of larger context windows and more sophisticated "reasoning" models, the gap between AI-generated prose and mid-list human fiction is narrowing. As these models become more accessible and easier to run on consumer hardware, the trend toward private, personalized fiction is likely to accelerate.
The publishing industry, therefore, faces a choice: it can continue to litigate against the use of AI in professional circles, or it can acknowledge the emergence of a new category of "prosumer" fiction. As the boundaries between consumer and creator continue to blur, the ultimate legacy of AI in literature may not be the replacement of the author, but the empowerment of the reader to finally step inside the story.
About the Author
