GuShiio.com Wen| Luo Chao Channel
Over the past 25 years, DeepSeek has had great momentum to restructure all industries, and the application that has been hit most is search.
-
WeChat regards AI search as the first step in connecting to DeepSeek, and some self-media claims that this move will reconstruct AI search;
-
Baidu AI Search immediately supported DeepSeek and welcomed this wave by embracing DeepSeek;
-
Zhihu’s AI product Zhihu Direct Answer is connected to DeepSeek, focusing on a knowledge base that aggregates knowledge within the station;
-
Weibo Smart Search DeepSeek-R1 In-depth Thinking Edition launched internal testing to better present hot events, vertical topics and other content;
-
……
Although AI search, smart search, and direct answer have different names, the head Internet platforms are competing to access DeepSeek’s product form, and they are all in-depth search.Some media also call this type of search DeepSeek search to pay tribute to DeepSeek.
I experienced it and found that although they are all DeepSeek drivers,However, different DeepSeek searches give different answers to the same question. The fundamental reason is the difference in content, which also means that different DeepSeek searches will have completely different values to users and the platform itself. It is no exaggeration to say that DeepSeek has launched a new round of search wars.
“DeepSeek search big horizontal review: content is the winner
WeChat AI search (Deep Thinking-R1), Weibo Smart Search (DeepSeek-R1 Deep Thinking Edition), Zhihu Direct Answer (Deep Thinking Edition) and Baidu AI search (DeepSeek-R1 Full Blood Edition) have different interactive interfaces, but the core interaction process is similar: after the user asks questions, AI search first takes a few seconds to tens of seconds to present the thinking chain, that is, the thinking process, and then gives the final result.
In this process, AI search has the ability to think deeply compared with traditional search. However, the in-depth thinking capabilities of different platforms come from DeepSeek’s thinking chain, so it is difficult to differentiate them.The basis for DeepSeek ‘s in-depth thinking is the search results of the site’s content, so content has become the most differentiated point for different DeepSeek search engines.
After experiencing DeepSeek search on the head platform, I firmly believe that content is the winner and winner of the AI search war.
Case1: Is it true that Lei Jun has become the richest man in China? Search time: February 27.
On February 27, as the market value of Xiaomi exceeded one trillion Hong Kong dollars, some media reported that Lei Jun’s worth once surpassed Zhong and became the richest man in China. With this question, I went to the DeepSeek engine, and the results were that Baidu, Weibo, WeChat, and Zhihu all gave a positive answer.
(From left to right, they are: WeChat AI search, Baidu AI search, Weibo Smart Search and Zhihu Direct Answer)
Zhihu just listed the reference sources and gave concise answers; WeChat, Baidu, and Weibo showed a detailed in-depth thinking process. Among them, Weibo and Baidu have detailed source bases. The former is based on Weibo content, and the latter is based on web page results.
(From left to right: Weibo Smart Search and Baidu AI Search)
Weibo Zhisuo’s in-depth thinking process surprised me. It mentioned that the content of Lei Junxi mentioning the richest man in an hour came from a verified account, and another content was that an ordinary account forwarded an article by Shangguan News, an authoritative media. Therefore, it has a certain degree of credibility.I feel that it is just like our technology media editors, with professional thinking logic and high credibility.
(The thinking logic shown by Weibo Zhisuo’s thinking chain)
The ability of Weibo Zhisuo to think like this is related to its content structure: the content source is not a third-party webpage link, but a named account. Information such as authenticated accounts (blue V, yellow V, etc.), media organization accounts, official corporate accounts, celebrity celebrity accounts, accounts of event parties, and accounts of fans of different sizes can become the basis for AI to think. This is its unique resource.In AI search, content is the primary factor of production.
Case2: Did Zhou Shouqian respond to Wang Xiaofei? Search time: February 27.
Not just entertainment gossip, I found that when searching for hot topics in society, Weibo would think deeply based on account attributes, presenting the fermentation process of hot events in the thought chain, allowing users to quickly understand the ins and outs.
For example, when I asked Zhou if he had responded to Wang Xiaofei?& rdquo;, Weibo gave a very clear answer, because Weibo has many unique chatter and short discussions on this topic, and AI can also point out that many big V’s have confirmed the authenticity.
(From left to right, they are: WeChat AI search, Baidu AI search, Weibo Smart Search and Zhihu Direct Answer)
As a square-style public discussion platform, Weibo has a natural content source advantage in hot events. So, if you ask some questions about knowledge, life, and interests, who is better?
I had a herniated lumbar disc and wanted to resume running when the weather got warm, so I asked DeepSeek to search for what should I pay attention to when running with herniated lumbar disc?& quo; Judging from the results,WeChat’s answer is the simplest, while Baidu’s answer is mainly based on common sense. Neither of them is of great reference.
(From left to right: WeChat AI search, Baidu AI search)
Zhihu gave quite professional answers based on the content of professional answers such as medical topics, which reflected the advantages of the knowledge base; there are many certified big V’s in the medical field on Weibo, and based on this, more professional answers can also be given. These two answers can really be used by me.
(From left to right, they are: Zhihu Direct Answer, Weibo Smart Search)
I decided to try asking some more life-related topics.
·························································
Recently, when I returned to Nantian in Guangzhou, I asked DeepSeek how to quickly dry clothes in Nantian. Every AI search gave a relatively accurate answer. Among them, the WeChat results were also automatically noted and made them more readable.
(From left to right, they are: WeChat AI search, Baidu AI search, Weibo Smart Search and Zhihu Direct Answer)
Case5: Recommend an electric vehicle with high-end smart driving cost about 200,000 yuan. Search time: February 27.
In the past 25 years, high-end smart driving has been relegated to 200,000 or even 100,000 models, and it has become a hot topic in car circles. I asked DeepSeek search people to recommend an electric vehicle with high-end smart driving with about 200,000 yuan. The recommended models by Baidu, Zhihu, Weibo, and WeChat are all highly overlapping. I asked the editors of Smart Car New Media Tram Connect to judge, and they all said that the recommendation accuracy is extremely high.
(From left to right, they are: WeChat AI search, Baidu AI search, Weibo Smart Search and Zhihu Direct Answer)
It’s not just a list of models. How can many models further match demand? WeChat did not give any suggestions on purchasing in the end, but Weibo, Baidu and Zhihu gave clear suggestions.
(From left to right, they are: WeChat AI search, Baidu AI search, Weibo Smart Search and Zhihu Direct Answer)
The thought chain of Weibo here is still worth praising again. For example, if a certain vehicle model is mentioned in N results, it will become its recommendation reason. Users can also click on the reference blog post to view the source content and check it carefully. According to the publisher and interaction of the content, Baidu and Zhihu will also give a reference to the content source of each conclusion. WeChat does not have it. It seems that the function is still relatively original.
(Weibo Zhisou)
After experiencing it, I feel that the differences between different DeepSeek searches are quite large. The fundamental reason is still that the content ecology is different.If DeepSeek is compared to a chef, in-depth thinking is its cooking process, and the content of the platform is the ingredients. Whether the final meal is delicious is as important as high-quality ingredients.
Based on this logic, I think this round of AI search wars launched by DeepSeek may make some platforms stand out and join the ranks of mainstream search engines.
DeepSeek changes the search landscape: content is more important
In this wave of war, Baidu is the defender. It has long been deploying AI search. However, it was previously driven by the Wenxin Model. The result is web content based on third-party and Baidu’s ecosystem. Now that DeepSeek is added, the summary will be more in place.
For WeChat and Zhihu, DeepSeek search can tap the content potential of the platform.Weixin Official Accounts have high-quality long texts, and AI search can aggregate the content. More importantly, by cultivating user habits,WeChat DeepSeek will eventually evolve towards chat-style intelligent assistant robots, which is the big guy of WeChat AI;Zhihu’s content advantages are professional knowledge and serious discussion, so direct answers will have significant advantages in medicine, science and other fields. DeepSeek will strengthen its status as knowledge-based vertical search.
For Weibo, DeepSeek search is incremental. Weibo content is not a collection of web pages,The classical search architecture based on crawler + index is difficult to accurately filter and sort content, and PageRank, a key search algorithm, cannot take effect. Therefore, in the past, the results of Weibo searches were basically sorted according to release time rather than relevance, which has been criticized by users.DeepSeek search completely gets rid of its reliance on traditional sorting algorithms, because it does not need sorting at all, it only needs to conduct in-depth thinking, analysis and re-integration of the results based on the thought chain.
At the same time, Weibo’s content traceability capabilities can be used by DeepSeek’s thinking, and AI can make more accurate judgments based on the account characteristics of the published content.
In the past, Weibo’s content distribution logic was based on hot interest discovery and attention flow, which led to the concealment of many vertical interest content, especially in digital, ACG, fashion, games, literature, tourism, health, food and other sub-fields. Interested and life-oriented content.
Many people now think of Xiaohongshu when it comes to life-oriented platforms. It is also focusing on search, announcing at the beginning of the year that its daily search activity has reached half of Baidu.However, with connecting DeepSeek, I found that the vertical and long-tail content on Weibo that is close to life also has the opportunity to be seen by users, which in turn attracts more creations to supplement the life-oriented content.
I used some vertical topics to search for Xiaohongshu. Then compared it with Weibo Zhixun, the latter has indeed used AI search to achieve dimension reduction. For example, I asked a question in the ACG field, when will “Lonely Rock Theater Edition” be launched in China?” rdquo;, what Xiaohongshu gives is a pile of user notes, which require me to click in one by one to carefully select them. Weibo will directly give a conclusion based on the discussions among ACG big V users.
(From left to right, they are: Xiaohongshu Search, Weibo Smart Search)
I asked for recommended hiking routes around Guangzhou in March. Weibo Smart Search analyzed all the answers and gave me a conclusion. It was equivalent to doing the screening work for me in Xiaohongshu.
(From left to right, they are: Xiaohongshu Search, Weibo Smart Search)
Another example is when I asked Keji how to solve the problem of hair loss. The content in the front of the little red book is recommended fish oil, etc., which is suspected of advertising. Moreover, the difference between advertising and content in the search results is different. Many of the content in the little red book may be underwater advertising, and screening costs are large.
On Weibo Smart Search, DeepSeek even reminded me that some results that are suspected of advertising can be eliminated, and some methods should be vigilant if they don’t work well, so as to truly screen content from the user’s perspective.The effectiveness of the results stems from the content advantages of Weibo,On the one hand, the Weibo platform has real-time hot attributes and time-sensitive content; on the other hand, it is of good quality, and publishers can be traced, privately sent, and interacted, especially for certified users such as media, celebrities, institutions, officials, and professional personnel (such as doctors). It represents the authority of the corresponding field, and the content is easier to be screened by people or AI.
(From left to right, they are: Xiaohongshu Search, Weibo Smart Search, Weibo Smart Search Thinking Chain)
It is an industry consensus that AI search is the big guy in the search market. Why has the little red book that strategically values search not been connected to DeepSeek? I think it has something to do with the fact that its content is mainly pictures and videos, which is not the content that DeepSeek is good at handling before the launch of the multimodal version. This is why Douyin is unlikely to connect to DeepSeek. As of now, searches for Xiaohongshu and Douyin are both a bunch of result lists for traditional searches, rather than the only answers.
The industry knows that search was once one of the Internet businesses with the highest technical threshold. AI can in fact be regarded as a business grown by search engines. NLP, deep learning and other underlying AI technologies are closely related to search. Because of this, Baidu and Google have invested heavily in the AI field.However, the emergence of DeepSeek directly changed the rules of the search game, and the threshold of technology has been leveled, because when everyone orders takeout, cooking skills may not be so important. At this time, the focus of AI search competition is content.
Traditional search engines are technology-driven. Whoever has strong technology can obtain more content and filter and clean out more effective content from it. The AI search competition is based on the platform’s own content ecosystem, this is not a question of more or less, but more importantly, the quality of the content (the key to determining the validity of the answer) and the structure of the content (who produces the content? It can assist AI decision-making), timeliness of content (outdated information, does not meet the current needs of users, and the importance of online search). From this perspective, Weibo, WeChat, and Zhihu all have their own advantages.
When it comes to AI search, technology can order takeout, but the content still depends on yourself.
How does DeepSeek redefine AI search?
These are actually two topics: 1. How does AI redefine search? 2. How does DeepSeek redefine AI search? Let’s look at the first question first.
Traditional search is based on the crawler + index underlying architecture, using a sorting algorithm to present a bunch of results to users in an arrangement or aggregation manner, and users then carefully select the content they really need from the results. Deep search is not divorced from traditional search, but it does the user carefully select this step and gives the only answer. This is the value of deep thinking.
This is also the goal that search engines have been working hard on: to give users only one answer, the answer they need most, less is more, and what you see is what you get.
In the AI era, the only answer will be more important, because IoT makes the search portal multi-scenario. For example, if a user asks a voice assistant questions while driving, TA can only give one answer, and it is impossible to give a bunch of answers for the user to filter.This also means that the value of search is no longer information retrieval aggregation and accelerated information flow, but truly becoming an expert assistant.
Therefore, the competition of traditional search experiences lies in the relative validity of the first three, first page, or first few pages of results; the competition of AI search experiences lies in the absolute accuracy of the only answer.
Experts must output effective and correct information to become industry authorities. Similarly, only when the answers in AI search are accurate enough can users use them with confidence and transform them into high-frequency and heavy users.
The AI model greatly improves the absolute accuracy of the only result,“”If we use the analogy of autonomous driving, the current accuracy of AI search is at least L4, which is almost unattended. Occasionally, hallucinations will occur and users need to take over, which means that the validity of the answer needs to be judged and verified. This is the reason why Weibo Smart Search, Baidu AI Search displays reference sources in the results and allows users to click, just like smart driving, gives users the right to take over the answer, which is the right to verify.
What many people ignore is that AI search didn’t break out in 25 years. After the outbreak of ChatGPT, AI searches have gone out of the streets one after another. Baidu AI Search, Zhihu Direct Answer, and Weibo Smart Search have all been launched long ago. There are also innovative AI search players such as Tiangong AI and Mita in the industry.The emergence of DeepSeek has redefined AI search.
On the one hand, its innovative application architecture greatly reduces the cost of each Token.ChatGPT is actually a type of search, but it is too expensive to be free. Two years ago, John Hennessy, chairman of Google’s parent company Alphabet, judged AI search that the conversation cost of artificial intelligence such as large language models may be more than 10 times that of traditional search engines, and the cost reduction will take a few years at worst. rdquo; The emergence of DeepSeek shortened this process.
On the other hand, its open source model has completely opened up the application situation of the AI model.DeepSeek has opened up the model architecture, model cards, model parameters, technical reports, evaluation results, etc., allowing third parties to access and commercialize it for free. As a result, we have seen Tencent, Baidu, Weibo, Zhihu and other platforms launch in-depth search for barrier-free access.
DeepSeek redefines AI search, and in fact will redefine the game rules, business model and market landscape of search. In addition to search giants such as Baidu and Google, platforms with content such as Weibo, Xiaohongshu, Zhihu, WeChat, Snowball, and Headline can do in-depth searches. Based on content differences, they all have the opportunity to make differentiated search services.
In the AI search war, the current discussion will end in fashion. After all, Xiao He only showed his sharp points.Deep search has not reached the L5 level in terms of reliability of results, and the existence of illusions will affect user trust to a certain extent;AI search has made the bidding advertising model that traditional search relies on ineffective, and a new business model must be formed to be sustainable. The head platform has not yet developed a business model; the DeepSeek multimodal model has not yet been launched, and multimedia content platforms have not yet embraced deep search.
It can only be said that with the rapid advancement of AI technology, search as a golden AI application scenario is bound to take the lead in AI, and a market with an annual scale of more than 200 billion US dollars is being reshuffled.
