Gordon Sheng spent months building InkOS, an AI-powered writing platform designed for China’s vast web novel industry. By mid-2026, InkOS had accumulated more than 50,000 users, authors who used the tool to generate plot outlines, draft dialogue, and accelerate the production of the serialized fiction that Chinese platforms publish in volumes that dwarf anything in Western publishing. Then the platforms started pushing back.
Rest of World reported on July 6 that China’s major web novel platforms, including properties owned by Tencent, ByteDance, and Baidu — are now actively restricting AI-generated content after a period in which they encouraged or tolerated it. The reversal reflects a collision between the economics of AI-assisted content production and the quality expectations of the readers who consume it, as well as the legal and reputational risks that platforms face when AI-generated content floods their catalogs.
The Numbers Behind the Crackdown
The scale of the AI content problem on Chinese web novel platforms is visible in the rejection data. In June 2026 alone, Tomato Novel, ByteDance’s web fiction platform, rejected 104,000 submissions flagged by its automated systems as AI-generated or AI-assisted. The platform has also introduced a daily word cap for submissions, a blunt instrument designed to make it economically impractical to flood the platform with AI-generated content at machine speed.
Jinjiang Literature City, one of China’s oldest and most prestigious web novel platforms, has taken a different approach, focusing on quality signals rather than volume limits. The platform has updated its submission guidelines to emphasize narrative originality and character consistency, qualities that current AI writing tools struggle to sustain across the long-form serialized fiction that defines the genre. Jinjiang’s approach reflects a bet that readers will self-select for quality, creating a market signal that platforms can use to filter AI content without relying entirely on automated detection.
Why Platforms Initially Embraced AI
The initial embrace of AI by Chinese web novel platforms was not irrational. China’s web fiction industry operates at a scale that is difficult to comprehend from a Western publishing perspective. The country’s major platforms host hundreds of millions of words of new content every month, produced by a professional and semi-professional author class that is under constant pressure to publish daily updates to maintain reader engagement and platform ranking algorithms. AI tools that could help authors meet those production demands were welcomed by both platforms and writers as a solution to a genuine productivity problem.
The economics also favored experimentation. AI writing tools reduced the cost of content production, and platforms that paid authors by the word had an incentive to encourage higher output. InkOS and tools like it offered authors a way to maintain the publication cadence that platforms demanded without burning out, a proposition that resonated with writers who were already treating web fiction as a full-time industrial enterprise rather than a creative pursuit.
(Related: China’s State Council Orders a National Survey on AI and Jobs — and the Numbers Are Staggering)
The Quality Problem and the Reader Backlash
The reversal came when the quality problem became impossible to ignore. AI-generated web fiction tends to be structurally coherent but emotionally flat, it can produce grammatically correct sentences and logically consistent plots, but it struggles with the character voice, emotional escalation, and narrative surprise that keep readers engaged across hundreds of chapters. Chinese web novel readers, who are accustomed to consuming fiction at high speed and have developed a sophisticated taste for the genre’s conventions, began identifying and rejecting AI-generated content in reader reviews and social media discussions.
The backlash created a reputational problem for platforms that had allowed AI content to proliferate. Web novel platforms compete for readers based on content quality and author reputation, and a catalog perceived as flooded with machine-generated filler undermines both. The 104,000 rejections on Tomato Novel in June alone suggest that the volume of AI-generated submissions had reached a level that threatened to overwhelm the platform’s editorial capacity and degrade the reader experience at scale.
What This Means for AI Writing Tools and Their Developers
The platform restrictions create a difficult environment for companies like InkOS. With 50,000 users already on the platform, Gordon Sheng faces a situation in which the primary market for his product, China’s professional web novel authors, is being squeezed between platform restrictions that limit AI-assisted content and reader expectations that demand the kind of quality that current AI tools cannot reliably deliver. The restrictions do not ban AI tools outright, but they create friction that reduces the commercial value of AI writing assistance for authors who depend on platform distribution.
The broader lesson from China’s web novel AI experiment may be that the economics of AI-generated content are more complex than they initially appear. Reducing production costs is valuable only if the resulting content can find an audience, and in a market where readers have strong preferences and platforms have reputational incentives to maintain quality, the race to the bottom that unrestricted AI-generated content enables is ultimately self-defeating.
(Related: China’s Universities Are Eliminating Language Degrees to Make Room for AI)