Behind Xiaohongshu's Regulations for AI Creation: A Value Choice of the Platform
Text by | Zhang Guo
In the memory of @One Side Painting Cats, a creator of Song Dynasty meticulous brush paintings, there is a clear timeline of changes in AI's involvement in creation.
In 2023, AI was just a distant technical term. At that time, she tried to turn the original paintings into animations using traditional video editing special effects, but the feedback she received was "it's better not to do it". At the beginning of 2024, she first tried to put the original paintings into an AI model, and the result still shocked her: the model couldn't accurately identify the small snake and the Totoro in the paintings, and the picture of the auspicious cranes flapping their wings had a sense of distortion like "crowing at the top of one's voice".
The change happened in 2025. With the improvement of the model's control over the first and last frames, based on the self - painted manuscripts, AI could already accurately restore the expressions in the works. The Totoro could start to naturally wave its paws and fly through the clouds while maintaining its complete body shape.
In the past, to achieve such dynamic effects, creators often had to rely on a professional animation production team and go through weeks of modeling, communication, and rendering. Now, One Side only needs an "AI assistant".
The iteration of AI technology is changing on a daily basis. This breakthrough has inevitably allowed the technology to penetrate into the daily creation of more people and pushed the content industry to a crossroads.
On the other side of the liberation of creativity and efficiency is the order challenge faced by the global content ecosystem. When a photo that has never been taken, a "personal experience" that has never happened, or even a person who doesn't exist can be mass - produced with assembly - line efficiency, the most core trust asset of the content industry is at risk of collapse.
Facing this irreversible technological wave, on April 27th, Xiaohongshu took the lead in presenting a systematic governance proposal through its first AI Governance Open Day: encourage AI as a creativity amplifier, oppose AI as a forgery tool and a low - quality content production machine, and all content created with AI participation must be actively marked.
This is also the first platform to systematically present AI governance principles since this round of technological explosion. It attempts to provide a reference standard for AI creation for creators and a sample for content platforms during the chaotic period of AIGC.
The Two Sides of AI: A Productivity Lever and a Hotbed for Forgery
In the traditional creative logic, technical "hard work" is still needed between creativity and the finished product.
This is especially true for the Song Dynasty meticulous brush paintings that One Side studies: on the white silk woven from silk, using water boiled with acorn shells as the background color, after going through processes such as line drawing, ink dyeing, separate dyeing on the front and back sides, and over - dyeing, a complete Song - style imperial court painting often takes at least two weeks for "One Side", a creator from the Central Academy of Fine Arts, to polish by hand.
One Side had been thinking about making these paintings from 1000 years ago come to life for several years. Under the traditional industrial path, it was basically difficult to achieve this at a low cost. A person in the film and television industry once mentioned this gap: a few years ago, when making an animation through 3D modeling, it took a full 15 days just for a 5 - second shot.
It wasn't until the revolution of AI technology that this cost was extremely compressed, and One Side also had a "readily available" digital animation team. The production time unit became hours.
The creator uses AI to make static Chinese paintings "come to life".
Fundamentally, the improvement of efficiency is not to drive people out of the creative scene, but to let AI, as a tool, become an "amplifier of creativity". Globally, this productivity revolution is happening in various forms: special AIGC units have begun to be added to international film festivals, and top brands such as Coca - Cola and Nike are trying to use generative technology to shoot brand films.
The involvement of AI in creation and even its integration into life is inevitable, but the real problem lies in how to use it.
In the face of the obvious efficiency dividends, the overflow of low - quality AI content is becoming a shadow over the entire industry. When the creative threshold approaches zero, the cost of false content and low - quality AI content also collapses.
In the past, behind valuable content, creators had to pay the cost of time, money, or the "sense of presence" of the physical body. This physical investment itself was a credit endorsement of "authenticity". Now, this physical constraint has been deconstructed by algorithms.
The whole world is experiencing this authenticity crisis. AI commentary videos that are hard to distinguish between true and false, virtual creators generated by code, and fictional content are squeezing the limited traffic pool with assembly - line efficiency. In the industry, there has begun to be a common phenomenon of using AI to batch - copy the tone of personal experiences and fabricate non - existent consumption experiences or life moments. Even more, some people use AI programs to implement account trusteeship, impersonating real people in the comment section for fully automatic and imperceptible interactions.
It's hard to distinguish between true and false AI - generated content.
This contradiction basically exists in all content ecosystems. On one hand, high - quality creators use AI to extend aesthetic creativity, taste, and knowledge, with the core being human will; on the other hand, low - quality content uses AI to replace or even forge the existence of "people", fabricating stories and experiences, with the core being traffic arbitrage.
These constitute the two - sided nature of the collision between AI technology and content, and this two - sided nature is not a battle of offense and defense for a single platform, but has become a test question faced by the global content industry.
Governance Evolution: Xiaohongshu Draws Three Boundary Lines
As technology continues to advance, platforms are also rewriting the rules simultaneously.
Taking Xiaohongshu as an observation sample, since this year, from scattered actions to presenting a systematic AI governance proposal externally, Xiaohongshu is gradually establishing a set of rules: from initially encouraging marking, to cracking down on AI - managed accounts, and then to refining the rules, the platform has begun to clarify one thing: AI can be used, but there are boundaries on how to use it.
At the AI Governance Open Day, Xiaohongshu clearly defined three boundaries for AI governance for the first time: encourage AI as a creativity amplifier, oppose AI as a forgery tool and a low - quality content production machine, and all content created with AI participation must be actively marked.
Excerpts from the AI governance proposal
An obvious change is that the past way of distinguishing whether content is AI - generated by "finding flaws" is fading out. For example, judging by "a person having six fingers" has little meaning today. After the model's ability has improved, it's becoming increasingly difficult to simply distinguish between true and false.
Therefore, "whether it's made by AI" is no longer the primary issue, and "whether it's okay to use AI in this way" has become the focus.
As early as January this year, in the "Community Convention 2.0", Xiaohongshu clearly stated: "If you use AI - assisted tools in your creation, please mark it actively." This was the platform's first statement on AI - generated content, aiming to encourage more creators to participate in marking and actively assume the responsibility of information transparency.
Subsequently, the rules began to tighten. In February, "Shuguanjia" issued an announcement, calling on users to actively add AI - generated and synthesized markings. For content that is not actively marked, the platform will add the markings uniformly after identification, and clearly crack down on the black - market behavior of selling "tutorials on removing AI markings".
Xiaohongshu has issued multiple AI governance announcements recently.
Regarding the "marking" governance, Xiaohongshu is taking both restrictive and supportive measures. While presenting the AI governance proposal, the platform has also supplemented the other side of the rules: high - quality AI - generated content with markings will receive more public - domain traffic.
When the boundary between true and false content is blurred, it's difficult for either the platform or users to make independent judgments. Therefore, Xiaohongshu emphasizes that marking is not just about putting a label, but about safeguarding users' right to know. Allowing AI - involved content to operate in the open, rather than lurking in the dark, is to prevent false information from ultimately eroding users' trust in creators and the platform.
On the other hand, Xiaohongshu clearly opposes the use of AI for black - and - gray - market activities: for "managed accounts" that use AI programs or third - party tools to achieve fully automatic posting and fully automatic interactions, the platform will take gradient measures until the accounts are banned.
Data shows that since this year, Xiaohongshu has cumulatively handled more than one million cases of various AI - related bad behaviors, including more than 800,000 AI - managed accounts and nearly 150,000 AI - forged notes.
As the first platform in the industry to say "no" to AI - managed accounts, Xiaohongshu believes that this kind of content is essentially mass - producing "pretenders that look like real people". It bypasses the most basic principle of "sincere sharing" in the community.
Essentially, this AI governance proposal is not to restrict, but to provide a clearer path for creators. Whether it's using AI to disrupt order, forge and infringe, or produce low - quality content, all are within the scope of governance; for high - quality AI - generated content, the platform has clearly shown its welcome and support.
Especially for content that is "aesthetic, narrative, and has real information increment", such as encouraging programmers with an algorithm background to popularize hard - core papers in an easy - to - understand way, or supporting designers and photographers to use AI to achieve complex visual ideas. Just as the creator "One Side" uses AI to make Song Dynasty Chinese paintings "come to life", this is human creativity leading technology, rather than technology replacing human thinking.
Behaviors such as using AI to impersonate real people, fabricate personal test experiences, or even plagiarize and infringe are on the list of direct crackdowns.
Ultimately, what this set of AI governance rules wants to do is very simple: to clarify the ambiguous areas. What can be used, what is a reasonable way to use it, and which behaviors are unacceptable are all stated in advance as much as possible.
Behind AI Governance is the Platform's Value Choice
For a long time in the past, many creators were in an ambiguous situation when using AI: will they be restricted in traffic if they mark AI? What degree of use is considered "excessive"?
When these questions remain unresolved, creators are likely to become conservative or, on the contrary, use AI only as a tool for low - cost and batch content filling.
The AI governance framework provided by Xiaohongshu this time attempts to shed light on the gray area. It conveys a core message: the platform does not reject AI, nor will it drive AI out of the community. What is truly differentiated is the use of AI and whether it brings real information increment.
At this Open Day, Xiaohongshu revealed that the platform is also establishing a traffic compensation mechanism for misjudgment scenarios, which also applies to misjudgments on the platform in the AI scenario. Through methods such as compensating traffic vouchers, the governance rules will not become a cold - cut - off measure.
Xiaohongshu has announced the launch of an appeal channel for AI - related note stickers.
At a time when AI has significantly lowered the creative threshold, the irreplaceable real - life experiences and real perspectives of real people have become scarce assets in the content industry. If AI - generated content is allowed to flood without restriction, the industrial - scale mass - produced false content will inevitably squeeze the living space of original creators, causing the content ecosystem to fall into a vicious cycle of bad money driving out good money.
The recent chain of disputes in the cultural and entertainment industry caused by AI - synthesized content essentially point to the same problem: when AI crosses the line to forge personal experiences and fictional real - life scenarios, the trust system will quickly loosen.
Looking back, Xiaohongshu's measures and detailed rules do not want the governance of AI - generated content to become a negative obstacle to creation, but rather to become a more definite usage guide - AI can be used to improve efficiency and express creativity, but it cannot replace real experiences, and it cannot be used to disguise as human experiences, identities, and judgments.
Essentially, any governance behavior is an external manifestation of the platform's value choice. As the first domestic community to positively and systematically answer "what the platform thinks of AI", Xiaohongshu's core consideration is not simply a technological decision, but about "what the platform wants to preserve". Xiaohongshu's answer still emphasizes the community's long - standing foundation: "sincere sharing and friendly interaction".
For example, the severe crackdown on "AI - managed accounts" is essentially because the behavior of machines simulating real people will dilute the space for real creators. In an ecosystem where hundreds of millions of people refer to Xiaohongshu for consumption decisions every day, this kind of interference with the sense of reality is a direct erosion of the foundation of mutual trust in the community.
AI - managed accounts
For the entire industry, we can no longer deny the efficiency and creativity of AI, and we can't try to go back to the old era without generative technology. What platforms can do is to redraw the boundaries: clarify which expressions must be based on real experiences, which generated content must be placed within the sunlight - based marking system, and which behaviors will damage the ecological balance of the community.
Xiaohongshu's AI governance proposal is an early response to the platform's attitude. As AI continues to penetrate into all aspects of content production, all platforms will face the same ultimate question: if generated content is allowed to flow in without distinction, trust will be diluted; if it is over - restricted, creativity will be suppressed.
Between