Why is WeChat search connected to DeepSeek at this time? Late at night on February 15, Tang Chen discovered when searching on WeChat that WeChat search was being tested in grayscale to access DeepSeek-R1.Some WeChat users can click on the WeChat search box and enter the "AI Search" page. In the "Ask AI" entrance, two options are provided: "Quick Answer" or "In-depth Thinking".
"King Fried Combo"
According to the page introduction,"Quick Q&A" mode provides the most commonly used, fast and efficient answers.In fact, "Quick Question and Answer" is the AI question and answer function launched by WeChat in mid-2024: In addition to providing regular search results, WeChat search will also display AI-generated (AIGC) answers at the top of the results page.
"Deep thinking" mode, a more comprehensive answer provided by the DeepSeek-R1 model after long thinking.During use, this mode will display the process of thinking and reasoning, which is no different from the independent DeepSeekApp.
During the actual test, I asked "What is Tang Chen's content style like?" and "In-depth thinking" showed his thinking process in detail, logically analyzed the relevant search results, and extracted key information.
Finally, the "In-depth Thinking" mode provides a summary in five aspects: "In-depth analysis and logic", "Industry insights and forward-looking", "Case-driven and practical", "Cross-field perspective and criticality", and "Language style (professional yet vivid)", and lists typical content scenarios separately.
Moreover, "In-depth Thinking" also provides reading suggestions: Tang Chen's content is suitable for readers who pay attention to Internet industry trends and product operation strategies, and especially provides practitioners with in-depth interpretations that combine theoretical framework and practical reference.
The information source provided by this "deep thinking" includes both public account tweets and web page information, and it also supports continued questions at the bottom of the page. However, limited by the corpus and training effect, there are still problems such as information cocooning, historical answers cannot be found, and continuous answers cannot be answered.
The function introduction page of WeChat shows that this product involves the provision and use of DeepSeek open source large models. The DeepSeek series of open source large models are provided by DeepSeek. We would like to thank DeepSeek and the DeepSeek open source community for their contribution to the open source of large models and related research.
Currently, only some users are eligible for testing. This also means that the "DeepSeek whirlwind" has hit WeChat, and WeChat has also taken this opportunity to once again add AI in search scenarios, taking a step further towards the goal of building an intelligent knowledge center for social scenarios, and its "AI search" ambitions are also about to emerge.
Regarding the "WeChat + DeepSeek" cooperation, some netizens commented, "National sports-level AI + national-level applications, a truly king combination."
The sword points to “AI search”
What I am more concerned about is, why did WeChat search embrace DeepSeek at this time? There may be two reasons for this:
On the one hand, it is affected by the "DeepSeek effect".As an adherent of the open source route, the success of DeepSeek is considered a victory for the large open source model, and the value of open source has also been extremely magnified. Since the Spring Festival this year, major Internet platforms, mobile phone manufacturers, new energy vehicle companies, large model manufacturers, etc. have all connected to DeepSeek, hoping to take advantage of this open source wave.
A typical example is Baidu, which announced on February 14 that it will launch the Wenxin large model 4.5 series in the next few months and officially open source it on June 30. You know, Baidu previously emphasized "closed source" and was also the most determined opponent of "open source". One of the key reasons for Robin Li's attitude reversal is the impact of DeepSeek.
On the other hand, WeChat search embraces DeepSeek, and its own consideration lies in the four words displayed after the user opens the search box: AI search.In other words, WeChat’s greater ambition is to target AI search.
Previously, Yuanbao mainly promoted its AI search and integrated WeChat ecological content such as WeChat public accounts and video accounts as a competitive advantage with other similar AI applications.
This time, WeChat search is directly connected to DeepSeek, which is obviously close to the water. It will have a profound impact on the user experience, content ecology, and commercialization path of the WeChat ecosystem.
For example, at the level of WeChat content ecology, WeChat AI search integrates public account articles, video accounts, mini program services, Moments, web pages and other content to achieve a one-stop search of "social + information + services".Especially for content creators, AI search will solve the "content sinking" problem with higher efficiency: the value of long-tail content such as historical articles on public accounts and long and short videos on video accounts may be rediscovered through AI recommendations. This will further incentivize the production of in-depth, valuable content.
On this basis, WeChat’s user experience and commercialization path will also face reconstruction.The scattered content information in the WeChat ecosystem (official accounts, video accounts, mini programs, etc.) will be aggregated across scenarios by "conversational" intelligent interactions, changing the traditional search's reliance on keyword matching and presenting results more accurately and efficiently.
For example, if a user searches for how to take an online ride-hailing service at Beijing Capital Airport T3, I asked about the "Guide to the fastest ride-hailing ride at Beijing Capital Airport T3 Terminal" posted on WeChat. It will be directly displayed on the search results page, and may be linked to relevant public account articles or merchant stores.
Previously, Tang Chen had published in "WeChat wants to copy a "Zhihu" in the search box? 》 analyzed in the article, WeChat is already a national-level super App with more than 1.3 billion users, and different content carriers have been "grown" within it, such as official accounts, mini programs, friend circles, video accounts, take a look, listen, etc.
For a long time in the past, the content pool within the WeChat ecosystem was mostly fragmented and separated, making it difficult for users to obtain relevant information through a certain channel or channel. Search naturally becomes the key to unlocking the WeChat content library. Finding information through people complements recommendations, opening up the entire WeChat domain and connecting the entire WeChat ecosystem for effective information distribution.
Regarding the positioning of Souyisou, the WeChat team once said,WeChat is not a purely independent search engine, but a part of the WeChat ecosystem.Let better and more appropriate content and services create a more precise connection with users.
The search results it presents are based on the content of the WeChat ecosystem and mainly include three types: information, services, and brands.
Products such as DeepSeek are connected to WeChat, and large models have found suitable usage scenarios.WeChat can use AI to analyze user behavior and search content within the WeChat ecosystem, further personalize recommendations, and predict needs in advance even before users search.
For example, users who directly search for keywords such as "taxi-hailing" and "takeaway" can jump to the mini program without leaving WeChat, forming a "search-service" closed loop. A typical manifestation of this is that in October 2024, WeChat officially announced the launch of the "search direct" function of Souyisou.
This transformation from passive search to active recommendation is also a major change in WeChat search's goal of "AI search" to build a more intelligent search service system.
But for WeChat search, its advantages and challenges all stem from the closed nature of WeChat ecology. Even if AI search is about to come out, the privacy protection issues faced by the large model industry, the complexity of experience caused by over-reliance on AI, and the hallucination of large models will also be challenges faced by WeChat.
From this point of view, WeChat’s restraint is not entirely a bad thing: slower, maybe faster.