Webnovel Platform Sanctions Authors for AI-Assisted Content Creation
Qidian Chinese Network removed multiple webnovels from homepage rankings and recommendation feeds on August 18, 2026, after determining they contained excessive AI-assisted content creation. Authors used AI tools including GPT-3.5, Gemini, and Llama 3 primarily to revise grammar and descriptions rather than generate plots, yet faced sanctions that reduced their income-generating visibility. The platform and industry lack clear standards distinguishing permissible AI-assisted revision from prohibited direct AI content generation.
What happened
On August 18, 2026, Qidian Chinese Network (起点中文网), a major webnovel platform, processed multiple works deemed to have excessive AI creation markers. The affected works were not delisted but were removed from homepage rankings and recommendation feeds, losing access to important traffic sources.
Authors disclosed varying degrees of AI tool use in their works. Some authors acknowledged using AI to revise grammar and phrasing rather than generate core plot and dialogue. Qidian had issued a policy announcement on April 21, 2026, opposing the use of AI or automated means to replace genuine human authorship of core content and stating it would strengthen oversight of non-human automated text creation.
Reported impact
Readers reported finding large amounts of redundant environmental descriptions and filler details in some webnovels, which they attributed to AI generation. Some works contained visible AI prompt artifacts in the published text. The platform's concern centered on maintaining original content ecosystem and reader satisfaction, as paid readers objected to reading AI-written content.
Industry data indicated AI creation tools could produce 5 million characters of long-form novels in 48 hours. In February 2025, Tomato Novel disclosed detecting numerous accounts abusing AI technology for mass creation of poor-quality, homogeneous, and unreadable content, with some accounts updating over 100 works in a single day.
Index analysis
A study from the University of Southern California, published in Nature Human Behaviour, analyzed over 880,000 real texts across seven datasets to examine language diversity following large language model deployment. The research found that when LLMs assist in writing, text complexity variance narrowed by up to 50 percent, with AI-revised text showing 21 to 50 percent reduction in complexity variance compared to originals. Semantic similarity between original and AI-revised texts exceeded 0.95 in 87 percent of cases, indicating content preservation but stylistic flattening.
Governance controls
In April 2026, Qidian published guidelines opposing AI replacement of genuine human core content creation. Enforcement measures include suspension of recommendations, removal from rankings, work suppression, and delisting. In April 2025, 16 major webnovel platforms including Qidian, Tomato Novel, and Jinjia issued a self-regulatory pact requiring establishment of AI-assisted creation usage standards.
Industry actors acknowledged ongoing ambiguity about where the boundary lies between permitted AI assistance (such as proofreading and resource research) and prohibited direct content generation, with no clear legal or industry consensus on what constitutes excessive AI involvement in a work.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.