Preventing AI is just the beginning; the real test for Xiaohongshu is only just beginning.

``` On April 27th, Xiaohongshu released its first comprehensive AI content governance rules, establishing authenticity and original value as the fundamental standards for AI content governance, clearly encouraging AI as a creativity amplifier, while firmly opposing the use of AI as a tool for fabrication or as a machine for producing low-quality content. This highly significant statement can be said to pour cold water on the recent craze since March for “raising lobsters” (OpenClaw and other AI agents) in an attempt to fully automate the operation of social media accounts. UGC: Clear Boundaries To deeply understand the urgency and underlying logic of Xiaohongshu's governance stance, one must look at the "raising lobsters" movement that began this March. Countless speculators and black/grey market practitioners quickly sensed an opportunity and tried to use agents to generate thousands of seemingly real images, texts, and videos daily on social platforms like Xiaohongshu. This has struck at the very heart of UGC platforms’ business lifeline—trust capital. For Xiaohongshu, the essence of its community ecosystem is a recommendation and decision-making model built on real personal experiences. The reason users are willing to be influenced and eventually make purchases on the platform is fundamentally because they firmly believe there is a real person with real trial-and-error experiences and genuine consumer pain points behind the screen. Once the platform's defenses are breached and overrun by mass-produced fake check-ins and fabricated reviews generated by large models, the effect of bad money driving out good money will instantly destroy trust between users and the platform, erasing Xiaohongshu's unique moat compared to traditional search engines and e-commerce platforms. It is precisely because of these existential ecological concerns that Xiaohongshu, in its announcement yesterday, for the first time presented its attitude toward AI governance in a comprehensive and clear manner—with extremely strong wording. This AI governance stance is divided into two main sections: "AI content and behaviors encouraged by the platform" and "AI content and behaviors opposed by the platform." In terms of encouragement, Xiaohongshu welcomes creators to use AI for the following: First, using AI to enhance the informational value of works, such as making complex knowledge more visual for popular science content, making difficult information straightforward and easy to understand; Second, using AI to create distinctive virtual characters or produce outstanding secondary creations based on existing IP; Third, using AI for visual creation to produce aesthetic and narrative quality works, including art illustrations, paintings, and cinematic short films with plot logic. On the opposition side, Xiaohongshu is against using AI for fully automated account hosting, AI-induced account farming, and other behaviors that disrupt community order; against AI deception, including impersonating celebrities, fabricating identities, falsifying personal experiences, creating false information, spoofing classic works, or riding hot topics for low-quality marketing; against AI infringement of portrait and copyright; and against mass production of homogenized content, generating visually extreme or sensational images for clicks, or spreading harmful values or other low-quality creations. In terms of specific governance mechanisms, the platform adopts a dual-insurance system that combines voluntary declaration and mandatory system tagging. Creators are required to accurately indicate content involving AI participation, and for suspicious content attempting to conceal this, the platform's detection algorithms will forcibly add an unalterable AI label upon identification, and may impose the highest penalties such as limiting distribution or even account bans, depending on the violation’s seriousness. Since the start of this year, Xiaohongshu has dealt with over one million cases of various AI-related misconduct, including over 800,000 AI-hosted accounts and nearly 150,000 instances of AI-fabricated posts. Similarly, as a UGC platform, WeChat Official Accounts added new rules on March 27th, clearly classifying AI-generated, mass-scripted publishing, and automated creation tutorials that lack real human expression as violations—cracking down mainly on “account factory” style content arbitrage. The direct trigger was the viral story "A couple earns two million yuan a year writing WeChat articles with AI," which upon investigation revealed that most of the income came from 299-yuan course fees rather than platform traffic, and the involved account was permanently banned. Deeper Challenges Remain With the continued leap in capabilities of foundational models, whether it’s UGC with defined boundaries or PGC (professional content) platforms trying to use AI for efficiency and cost reduction, there is still no reason to be complacent. Beneath the craze, the entire industry is facing a major test brought by AI. First is the governance challenge of “ever-escalating offense and defense”—as platforms step up management, AI technology keeps advancing. For platforms such as Xiaohongshu, establishing rules is just the first step. The real test lies in the cost and accuracy of enforcement. Currently, open-source agents and models have reached an extremely realistic level in human-like expression. Platforms’ “AI detection algorithms” and black/grey market “AI camouflage techniques” are locked in a long-term arms race. False positives that restrict real influencers or false negatives that let fake content slip by can each trigger major public opinion crises and lead to a huge drain in community trust. As models become more lifelike in lighting, micro-expression simulation, or even in deliberately constructing logical flaws, the cat-and-mouse game between AI detection and camouflage technologies becomes endless. On the commercial side, content deflation and aesthetic fatigue are new issues introduced by AI. While AI solves the quantity and cost problems, it brings hidden quality pitfalls. The probabilistic nature of content generated by large models means they naturally lean towards formulaic results, further saturating the platform with homogenized content. An AI science popularization blogger explains this can lead to so-called "model collapse," that is, when a model repeatedly trains on low-quality, low-diversity, mutually contaminated data, its generative ability gradually degenerates, eventually falling into a state of monotony, repetition, and distortion. Furthermore, regardless of whether it’s a PGC or UGC platform, AI usage also brings about ethical “digital exploitation” and copyright grey areas. This may be the industry’s toughest nut to crack. For example, the boom in AI-generated short dramas might lead to structural unemployment for traditional entry-level scriptwriters, illustrators, and voice actors. At the same time, training of large models inevitably absorbs countless original works from UGC platforms. When an AI-generated short drama goes viral, how should the resulting income be distributed among the original creators whose data was used in training? Under the current legal and platform rule frameworks, this remains a minefield of grey areas. Technology has always been a neutral amplifier—it ultimately magnifies a platform’s core DNA. UGC veers left, still pursuing authenticity and interpersonal trust; PGC leans right, chasing ultimate efficiency and dopamine hits. In this profound industrial transformation, there is no absolute right path—only the most earnest questioning of one’s own commercial essence. Even greater challenges continue to emerge. As AI-generated content becomes ever-harder to distinguish from human creation, how should platforms keep balancing “encouraging innovation” and “risk prevention”? This question will require further exploration. Risk reminder and disclaimer The market has risks; investment should be cautious. This article does not constitute personal investment advice, nor does it consider the specific investment objectives, financial situation, or needs of individual users. Users should determine whether any opinions, views, or conclusions in this article are suitable for their specific situation. Investment based on this is at your own risk. ```