AI生成的内容正在大量进入传统出版社和网文平台，引发行业内对作品质量和创作真实性的关注。[1]一部名为《饭圈纪实：爱、数据和权力》的书籍因疑似包含大量AI生成内容，其豆瓣评分已跌至5.3分。[1]编辑五山检测某部书稿发现其AI生成指数达到60%，个别段落甚至高达90%以上。[1]
在已出版的作品中，科幻作家郝景芳承认其作品《银河学院》中AI写作比重占到一半。[1]网文编辑傅语发现作者提交的稿件中残留有AI指令，包含"润色时应减少极其疯狂惨烈宛如等密集修饰，把情绪压进动作里"等提示词。[1]
对于这一现象，业内人士看法不一。作家刘亮程指出AI产出的文字是"虚构的虚构，和真实隔了好几层"。[1]编辑五山认为"AI能够提高文字的下限，但很难对上限有所突破"。[1]不过，《饭圈纪实：爱、数据和权力》的第一作者马中红则回复称："实事求是地讲，我们就没有用AI写作"。[1]
当前，出版行业缺乏100%精准的AI检测手段，这使得区分真正的人工创作与AI生成内容变得困难。[1]
Artificial intelligence-generated content is increasingly infiltrating traditional publishing houses and online literature platforms, raising concerns about the quality and authenticity of published works. A book titled "Fandom Chronicles: Love, Data and Power" saw its Douban rating plummet to 5.3 points following suspicions that it contained AI-generated material [1]. Science fiction author Hao Jingfang acknowledged that her work "Galaxy Academy" incorporated AI writing, with the artificial content accounting for approximately 50 percent of the total [1].
Editors have begun detecting alarming levels of AI involvement in submitted manuscripts. Editor Wu Shan identified one manuscript with an AI generation index reaching 60 percent overall, with certain passages exceeding 90 percent [1]. In another instance, web fiction editor Fu Yu discovered that an author's submission contained residual AI instructions, including guidance to "reduce densely packed descriptions such as 'extremely crazy and tragic' during revision and compress emotions into actions" [1]. However, Ma Zhonghong, the first author of another work, disputed claims of AI involvement, stating: "To be honest, we simply did not use AI for writing" [1].
The publishing industry faces significant challenges in distinguishing between human and machine-generated content. Writer Liu Liangcheng characterized AI-produced text as "fiction of fiction, separated from reality by many layers" [1]. According to editor Wu Shan, while artificial intelligence can raise the lower standard of writing quality, "it is difficult for it to make breakthroughs in the upper limit" [1]. Currently, the industry lacks a detection method capable of achieving 100 percent accuracy in identifying AI-generated material [1].
