At the moment when the AI wave is becoming more and more intense, the amount of information and application demand has ushered in a complete explosion, and the participating manufacturers urgently need to find the landing scene of large models, 2024 will undoubtedly become the "Shura field" of AI application field competition. Throughout the battle, changes are still taking place, new forces are constantly entering the arena, and traditional giants are also trying to interpret new ways of playing. Among them, AI search seems to have become a battleground for soldiers in the era of large models. For example, 360, which is obsessed with the search track, has launched a large-scale search product "360 AI Search", and has been launched on several Android app stores. Prior to this, whether it was traditional search players such as Bing, or Ali's Quark, "*Ask", or the launch of products such as "Search AI Assistant" at Station B, it is not difficult to see the ambition of players for the AI search track. There is no doubt that search is one of the Internet infrastructures, and it is the core means to undertake rigid demand, and has cultivated a "shelf-style" trading field. In the face of AI and large-scale model reconstruction, search as an application and search as an infrastructure have seen new opportunities. The search engine is one of the symbols of the classical Internet, which serves as a hub for information connections and holds the supreme authority over content distribution. Therefore, from the PC era to the mobile Internet era, players who are trying to drag them off the Iron Throne of domestic search have emerged one after another. 
However, whether it was 360, which had set off the "3B war", Sogou, or later headlines, they failed to subvert the leading position of domestic search. As a result, most players choose to retreat and create information cocoons in their own applications. The logic behind this is that search engines, as information portals, need to continuously absorb large amounts of data to improve the quality of search results. This means that in the "cold start" stage of a search product, there are often too few samples and uneven content - compared to mature products, the user experience of new search engines is often unsatisfactory, thus forming a high wall for user retention. In the past, there were not many solutions to this problem, so whether it was 360 search or search, or Toutiao trying to step into the search field, in the end, the search engine could only be implanted into its own products, with the help of its own traffic disk, forcibly imported users. But I have to admit that although the reputation of search in the Internet context is not very good, and it is not loved by "elite users", the "silent majority" in the Internet world will still give priority to thinking when searching - for a long time, this kind of unshakable mental advantage makes latecomers can only narrow the gap with the technology and content field, and it is difficult to surpass. Another way of thinking is to strategically abandon the "big and comprehensive" that search engines should have, and focus on vertical searches. Taking Quark, which is backed by Ali, as an example, it has accumulated word of mouth in the vertical group by focusing on the field of resource search and with the help of the network disk business - thanks to the characteristics of convenient resource retrieval and "not easy to turn gray", Quark has attracted some young users. From the perspective of the whole industry, even if vertical search players can achieve "small and beautiful" in a specific circle, the magnitude is still an egg hitting a stone compared to mainstream search engines. Once the search engine is too vertical, its users' minds are often limited by the scene, so that they will face obstacles in the process of expanding the circle in the future, such as users who are accustomed to using quarks to find film and television resources, and will never think of quarks when searching for a ** entrance. Statcounter data shows that as of December 2023, it is still at a rate of 66With a market share of 52%, it tops the Chinese search engine market, followed by Bing China, 360 Search and Sogou Search, while niche apps such as Quark are not on the list. This means that, according to traditional business logic, even if it is "rotten", it can firmly maintain its dominant position in the context of Chinese search. Fortunately, the swept AI wave gave other players a rare opportunity to overtake in corners. Overtaking in corners? In the context of the mainstream player models have been laid out, the charm of AI search for players is constantly emerging. As an important node in the application of large models, the search engine is not only a concentrated embodiment of the precipitation of players' large model capabilities in the past year, but also a subdivided application carefully created by many players, that is, "search as a service". For example, during the college entrance examination voluntary filling last year, the search engine was grafted with the AI voluntary filling function launched by the search engine, so as to introduce the large model application to the C-side; In the final preparation stage of the university, Quark also took the opportunity to launch the "AI Topic Assistant" to strengthen its search service capabilities with large-scale model applications. 
The introduction of AI capabilities into the evaluation dimension of search engines can also bypass the advantages of content resources, sorting algorithms, patents and other advantages accumulated by the search giants represented in the past. Compared with the era of lack of information and search engines, today's Internet information has long been overloaded - in the context of the rising number of media and information, whether it is the new "spam information" generated every day, or the homogeneous content that is copied from each other and copied virally, it has crowded the public eye. In this context, the user's search logic has also changed, from the pursuit of more comprehensive information in the past to the current "separation of the false and the true" in the massive amount of information, and the screening of available information. And this is precisely the area where AI capabilities can be demonstrated. The feature of AI search is that it can balance depth and efficiency, so as to output "useful" information for users more accurately, and present the information in the form of natural language dialogue. And this undoubtedly has a natural attraction for contemporary Internet users who are drowning in the sea of information. Therefore, looking at the AI search products on the market at this stage, "screening" has almost become an unavoidable keyword. This is in line with the British writer Neil Gaiman's words: "Google can give you 100,000 answers, but a librarian can give you the most accurate answers." AI is the "cyber librarian" that the public has been waiting for. 360, which has recently made efforts to make AI search, continues this route - when a user asks a question, AI will retrieve, read, and analyze the content of multiple web pages through the search engine, and output the results; Or by asking questions, read more web pages and conduct more detailed analysis; Kunlun Wanwei, on the other hand, adds a source function to its AI search product, so that the search results can be marked with the source of the generated content, thereby improving the credibility of the content. In other words, under the narrative of the widespread application of AI search, the old pattern will likely be subverted by new forces, and an offensive and defensive battle about AI search will also be staged. After all, for traditional search giants such as Google and other traditional search giants, the search business means real money, so even if they move closer to AI, the pace will inevitably tend to be slow; In contrast, players who are less tied to the past may be able to take a more aggressive approach to AI search. It's just that how this narrative line actually unfolds is currently torturing players from all walks of life who are deep into the battlefield. The vision of skinny realistic AI search is good, but at this stage, it is still a certain distance from the moment of large-scale explosion. According to the Photon Planet test, a number of products that claim to be AI search, the search results are similar at the content level, and there are certain logical problems - taking "who is the Nobel Prize winner in mathematics last year" as an example, many AI search applications only point out that relevant information cannot be retrieved, but fail to point out the objective fact that the Nobel Prize does not have a mathematics prize. A user who has tried many AI search products told Photon Planet that AI currently mainly plays the role of filtering information, provided that this information belongs to its own knowledge, otherwise it is far better to use Google to search slowly. "If you don't have the ability to discern, it's easy to be influenced by AI's 'hallucinations'. This means that there is still a lot of room for improvement in the accuracy and accuracy of AI search products at this stage. On the other hand, AI search is currently facing the problem of scarcity of user groups. At present, AI search users are more interested in cutting-edge technology geeks, rather than the general public. The "conversational" and "problem-solving" search experience of AI search can certainly cover the needs of college students, as well as customers such as investment and industry, but excessive instrumentalization has also put shackles on it to a certain extent. In other words, AI search may be able to win over the so-called "elite Internet users", but it does not necessarily provide a reason for the "silent majority" to use AI. As a simple example, if a user wants to go to a tourist attraction and make relevant strategies in advance, in the face of the AI's cold words, the user may be more willing to accept Xiaohongshu's notes with "temperature" and the comment area of amateur user reviews. In many search scenarios, users don't need short and precise answers from the "book of answers", but rather a systematic understanding of the cocoon. This means that if AI search cannot continue to evolve until it meets the demands of most search scenarios, then the products that players take a detour with the help of the AI wave may only be confined to a part of the aforementioned "vertical search engine", and the imagination space will be greatly reduced. Therefore, looking at the wave of AI search set off in the past two years, despite the hot surface, the vast majority of domestic players have not found a truly effective entry path according to the standard of reshaping the search track pattern, and most of them have changed their soup. Of course, if you take into account the possibility of AI iterative upgrades and the answers handed over by GPT Store, AI search can be a useful attempt for players who are lost in the fog of large model applications to find landing space.