Amid the Personal Agent boom, Xiaohongshu enters the market with "Mishu".

Recently, the REDtech technical team of Xiaohongshu released an article, officially announcing a Personal Agent named "Mishu", positioning it as an "AI life buddy".

Judging from the officially disclosed application scenarios, Mishu is not in a hurry to demonstrate overly complex technical capabilities, but focuses on specific real-life problems: for example, finding small local shops loved by residents in an unfamiliar city, discussing what to order from the menu, comparing the actual usage experience of two products, checking the trial feedback of new beverage releases, or finding exhibitions and markets worth visiting on weekends based on recent notes.

Over the past decade, competition among internet platforms has always revolved around one core issue: how to get users to spend more time using their own products.

Search engines compete for search demands, short video platforms compete for user attention, e-commerce platforms compete for shopping behaviors, and local life platforms try to keep consumer demands for dining, entertainment and other services within their own ecosystems.

However, despite their different business models, the underlying logic is consistent: acquire users first, then convert user attention into commercial revenue through content distribution, ad display or transaction services.

The emergence of AI Agent is gradually changing this logic.

When AI is no longer satisfied with simply answering questions, but starts to understand user intent, call external tools, coordinate different services, and even complete a series of operations on behalf of users, what internet platforms are competing for will also change accordingly.

Users may no longer need to open multiple apps to search for information, nor repeatedly compare across different platforms. A single Agent has the potential to complete the continuous process from demand identification, information screening to service invocation.

In other words, the traffic entrances, information distribution and transaction links that were previously controlled separately by different internet platforms may be reorganized.

This change has been particularly obvious since the second half of the year. In September, Meta launched the personal agent Muse, and OpenAI subsequently released Dots; in the domestic market, Alibaba's Qwen Go and Ant Group's Lingguang 2.0 were unveiled one after another, while ByteDance and Tencent are also accelerating their related layouts.

AI is evolving from a tool that users need to open repeatedly and take the initiative to ask questions, to an assistant that can stay by users' side for a long time.

Now, Xiaohongshu has also entered this track. It may seem to be riding the Personal Agent trend, but if we put Mishu into Xiaohongshu's business accumulation over the past decade, we will find that its choice is somewhat different.

Alibaba has a transaction and service ecosystem, Tencent has social connections and application entrances, and companies like OpenAI are exploring more general-purpose personal execution environments. The most valuable resource in Xiaohongshu's hands is a large amount of consumption experience generated around specific life scenarios.

In the past, these experiences helped users discover products, find restaurants, and plan trips; now, Xiaohongshu hopes to let AI call them directly to participate in users' real-world decision-making.

Xiaohongshu "breaks into" the Personal Agent track via a detour

Suppose a consumer is preparing to buy a camera.

In the past, he might first look up the parameters on a search engine, then check real reviews on Xiaohongshu, then open an e-commerce platform to compare prices, discounts and after-sales services, and finally place an order. Throughout the whole process, each platform solves part of the problem, but it is always the user himself who connects these services together.

Now, Agent hopes to take over this work.

Users only need to state their budget, usage scenarios and preferences, and AI can try to complete product screening, information comparison, price inquiry, and even call transaction services. The process that originally needed to be completed across multiple applications has the opportunity to be compressed into one interactive entrance.

This is also the biggest difference between Agent and traditional AI Q&A, but judging from the current layout of various companies, Agents have not formed a completely identical technical route.

Take Alibaba's Qwen Go as an example. Its advantage lies in its ability to call ecological resources such as Taobao, Amap and Fliggy, to further connect user demands to transactions and services. In public tests, Qwen Go has demonstrated capabilities such as querying flight delays, contacting airlines to claim compensation, and finding the price protection entrance on Taobao.

Ant Lingguang 2.0 emphasizes the transaction processing capability of personal AI assistants, hoping to integrate application creation, tool invocation and specific services into the same interaction process; Meta Muse and OpenAI Dots explore continuous service capabilities from the perspective of long-term personal assistants.

Products from different companies are still at different development stages, but the focus of competition has gradually extended from the quality of a single response to how much personal information and action permissions users are willing to give to AI.

Xiaohongshu's entry point is different. The platform itself does not have a complete business ecosystem covering shopping, travel and payment like Alibaba, nor does it need to replicate all the capabilities of a general-purpose Agent in the first step. What Xiaohongshu can more easily give full play to its advantages are those consumption problems that have no standard answers but highly rely on life experience.

Take restaurant recommendation as an example. Ratings and menus can tell users the basic information of a restaurant, but questions like "is it suitable to bring parents there", "which dish is worth ordering", "is it easy to get a queue on weekend nights" often can only be answered with more specific consumption experience.

Such problems are difficult to solve solely by relying on product parameters, map ratings or search results. Users also need to refer to other people's experiences and make judgments combined with their own budget, taste, time and specific scenarios.

This is also the content value accumulated by Xiaohongshu in the past. Mishu introduced by REDtech has already included these scenarios in its product exploration scope. For example, users in an unfamiliar city can ask about nearby small shops worth visiting, check relevant reviews when they arrive at the restaurant entrance, and even ask about dish matching against the menu after entering the store.

One very interesting technology here is what Xiaohongshu calls "progressive perception".

Traditional searches often require users to describe the problem clearly at one time, but in real life, people's demands often change with the change of scenarios: when users first enter a restaurant, they may only want to know how the restaurant is; after seeing the menu, they will further consider the number of people traveling together, taste preferences and dish portions.

What "Mishu" does is try to gradually obtain on-site information during the task advancement process, combine previous conversations and personal backgrounds, and give suggestions at the right time. With the first-person perspective interaction of Rokid AI glasses, users can ask questions while looking around, without having to re-describe the environment every time.

In other words, search engines solve the problem of where to find information, community platforms accumulate how other people experience one thing, and what "Mishu" wants to do is put these experiences into a person's specific situation.

The platform that knows best how to drive product discovery is also learning to persuade users not to buy

Community experience can be said to be the most visible advantage for Xiaohongshu to build Personal Agent.

However, there is still a gap between content and decision-making. Xiaohongshu has a large number of user notes covering life scenarios such as catering, travel, fashion, home furnishing, digital products and beauty. Compared with standardized product databases, these contents can provide more specific details: whether a product is easy to use, whether the service of a store has changed, whether a piece of clothing is suitable for a certain body type.

It is undeniable that this information is very valuable for consumption decision-making. But the problem is that notes record one person's experience at a certain time and under certain conditions. This note is real, but it may not be universal; it can provide reference, but it may not be directly converted into an answer suitable for another person.

For example, a coffee machine gets a lot of positive reviews probably because of its good extraction effect. But for someone who only drinks coffee two or three times a week and does not like cleaning machines, this product may not be worth buying.

The value of consumption experience does not only lie in how many people praise it, but more depends on whether AI can know who is speaking, why he is speaking, and who this experience is applicable to.

The importance of this point to Xiaohongshu is self-evident.

In September this year, Xiaohongshu disclosed that from June to August, the platform cleaned up and disposed of nearly 120,000 fake notes created and published by AI. Some accounts use AI-generated images to forge experiences of restaurant visits, accommodation, medical treatment, beauty and other scenarios, fabricate personal experiences, and try to mislead user decisions.

In the past, such fake notes mainly affected ordinary users' search and browsing experience. After seeing a fictional restaurant review note, users may make a trip for nothing; after seeing a forged product experience, they may also make wrong purchase decisions.

When AI starts to read, organize and summarize these contents directly, the problem will become more complicated.

When ordinary users browse notes, they can judge whether the content is credible by themselves through the author's identity, image details, comment sections and differences between different notes; if AI directly compresses a large amount of information into a few conclusions, the source differences and opinion divergences in the original content may be weakened.

Even worse, when AI-generated content starts to enter the community and is used by other AI as an experience source, low-quality information may form a cycle between content production and information retrieval.

For "Mishu", what it needs to do is not limited to providing opinions for users. It needs to identify information sources, learn to distinguish commercial cooperation content, ordinary user experience, and AI-assisted generated content; retain different judgments when a product has a large number of positive or negative reviews; distinguish the timeliness of information, especially for demands such as catering, tourism and offline services, past experiences may not represent the current situation.

These capabilities do not sound as conspicuous as real-time multimodality and long-term memory, but they directly determine whether users are willing to trust AI's suggestions.

In addition, "Mishu" has a more important task.

If its goal is only to help users find popular products and popular restaurants faster, the higher the recommendation efficiency, the higher the product discovery efficiency may be. But if it wants to become a real personal assistant, it needs to consider whether users really need this product, and whether there are more suitable alternative choices for themselves.

Product discovery solves the problem of why users want to buy, while personal assistants often need to answer whether users should buy at all.

A product is very popular, and AI can tell users why it is popular; but when users have a limited budget, low usage frequency, or already have similar products, AI should also be able to give suggestions to postpone the purchase.

However, this point that is valuable to users means that the platform needs to re-understand the meaning of consumption conversion.

In the past, the commercial value of Xiaohongshu largely came from helping brands find the right crowd, stimulating interest through content, and then converting interest into purchase behavior. Now, Mishu needs to gradually try to stand on the side of consumers, help users screen products, compare experiences, and even identify things that are not worth buying.

A valuable consumption Agent should not only count how many purchases it has facilitated, but also pay attention to how many inappropriate consumptions it has helped users avoid.

Empowering both sides with AI, Xiaohongshu's business has to change

If we look at "Mishu" together with Xiaohongshu's commercialization moves this year, its significance goes far beyond just an AI product.

On August 7, Xiaohongshu released the NEXT AI commercial product solution, covering AI capabilities in multiple links for merchants from user insight, content production to ad delivery and transaction conversion.

In short, NEXT helps merchants better understand consumers, produce marketing content and run their business; Mishu hopes to help consumers understand their demands, screen information and make choices.

This is very interesting: merchants are starting to use AI to improve the efficiency of influencing consumers, and consumers are also starting to use AI to improve the efficiency of judging products.

Previously, Xiaohongshu mainly connected buyers and sellers through recommendation algorithms: merchants studied the platform's traffic rules, optimized products and content; consumers browsed notes, compared reviews, and then decided whether to buy.

Now, AI may gradually change this link.

Brands can use AI to analyze user preferences and judge what kind of selling points are more likely to attract attention; consumers can also use AI to sort out reviews, compare products, and even directly ask whether a certain product is suitable for themselves.

Both sides have obtained stronger information processing capabilities, and the platform can generate synergy in the middle, which can not only help merchants operate, but also help users make decisions.

However, a clear boundary line needs to be established between these two sets of AI.

Suppose a brand analyzes the core consumer group of a product through NEXT, and continuously produces content around these groups. The user then asks Mishu whether this product is worth buying, how should AI process these contents?

If it only gives recommendations based on the number of notes, interaction popularity and high-frequency selling points, while the merchant-side AI improves the efficiency of content production, it may also affect the information judgment of the consumer-side AI.

Therefore, the platform needs to let consumers know why AI recommends this product, what content it refers to, whether it contains commercial cooperation information, and whether there are opposite usage experiences.

After all, when merchants hope AI can help them sell better, consumers also hope AI can independently help them make judgments. Otherwise, AI is just a different presentation form of the past feed ads, and what users see is still what the platform wants them to buy.

In addition, creators on the Xiaohongshu platform may not like such changes very much.

Previously, users entered Xiaohongshu to obtain information through notes one by one. Creators got views, collections and interactions, and the platform maintained community activity through content distribution. When Mishu can directly read a large number of notes and organize them into a concise answer, users may no longer need to browse them one by one.

For consumers, this is an efficiency improvement; for creators, the way content is used has changed.

A note may be called repeatedly by AI, but it does not necessarily bring corresponding clicks and interactions to the author. If users only remember the answer given by AI, but do not know who this experience comes from, the motivation for creators to keep sharing may be affected.

Of course, if AI can accurately mark the information source, guide users back to the original note, and let content with unique experience, real details and professional judgment get more exposure, "Mishu" may also bring new opportunities for high-quality creators.

The key lies in how the platform faces and handles the use and distribution of community content by AI.

In this light, the birth of "Mishu" not only adds an AI entrance to Xiaohongshu, but also changes the way users consume information, and may also change the way merchants operate content, and even affect the value distribution of creators in the community.