AI Search Goes to War: Tencent Yuanbao Search and the Giants' Battle for the Information Gateway
On the occasion of the company's 28th anniversary, Tencent launched "Yuanbao Search" as a standalone web page, and it can also be selected in the engine settings of QQ Browser, making AI search a new battleground for giants competing for the information gateway. The article reviews the monthly active user landscape of AI assistants such as Doubao, Qwen, and DeepSeek, as well as commercialization paths including Doubao charging hotels a 12% comprehensive commission on orders and Yuanbao Search opening up externally through Tencent Cloud at 14 to 26 yuan per thousand queries. The author argues that competition is shifting from the number of results to content coverage, source quality, and answer credibility, and that platforms also need to maintain a balance between monetizing recommendations and user trust.
Why read this
AI score 57
The article reviews the products and monetization paths of giants such as Tencent and ByteDance in AI search, showing how the battle for the gateway is extending into content and transaction distribution.
Topics
Body · AI translation
On November 11, 2026, on the occasion of the company's 28th anniversary, Tencent launched a new search product, "Yuanbao Search," which has its own standalone webpage. Users can also upgrade QQ Browser to the latest version and select "Yuanbao Search" in the engine settings options.
As one of the earliest product forms in the internet industry, after years of development, the search market has long entered an oligopoly era with a relatively stable landscape. Today, a host of technology companies are once again extending their reach into the search market, largely because AI technology has brought new changes. Under the influence of AI technology, the competitive landscape and business models of the search industry may all face dramatic changes.
01
Unlike the PC internet era, when search was a core strategic entry point, after the rise of the mobile internet, specific needs were taken up by closed platforms, and recommendation mechanisms could precisely distribute some information that previously required active retrieval, so general-purpose search engines began to decline.
In this regard, in April 2017, in an interview with the media, then Sogou CEO Wang Xiaochuan said, "On the PC side, search is the center, the entry point for traffic, the core of all applications. On the wireless side, it is indeed not the most central position; it is just one of several applications and services. That is the reality, and you accept that reality."
Precisely because search began to deviate from the central position, mobile internet companies no longer competed for the search entry point, nor did they build products around search; instead, they carried out vertical exploration one after another in line with user needs. Correspondingly, users also adapted in sync to the isolated-island mobile internet ecosystem, satisfying their needs for shopping, entertainment, video and audio, and so on through individual isolated apps.
Image source: Baidu
Traditional search companies such as Baidu are committed to reconstructing their existing search businesses on top of existing retrieval and distribution systems; AI assistants such as Yuanbao, Qianwen, and Doubao rely on large models to organize content and services; and companies such as Tencent, DeepSeek, and ByteDance open their retrieval capabilities to developers to serve more AI applications. These layouts often occur simultaneously, and search has also evolved from a product for C-end users into a foundational capability for AI products to connect with external information.
Overall, although the layouts of technology companies differ, their strategic orientation all points to the battle for information entry points in the AI era. AI has extended search from presenting information to understanding needs and organizing answers, and has also made it possible for users' long-established search habits to be rewritten.
For traditional search companies, AI search is a defensive battle to protect their advantages in users and distribution; for new players, it is an opportunity to redefine the starting point for users to obtain information. The key to competition is beginning to shift from who can provide more results to who can make users think of them first when they want to ask a question.
02
Image source: QuestMobile
Data disclosed by QuestMobile shows that in September 2026, the top three AI-native apps by monthly active users were Doubao, Qianwen, and DeepSeek, with monthly active users of 419 million, 158 million, and 123 million, respectively, up 143.6% and 5076.6% year on year, and down 15.2%, respectively.
Image source: Tencent Research Institute
Because large models have strong task understanding and feedback capabilities, more and more users regard AI assistants as "intelligent secretaries" for solving everyday life problems. Survey data from the Tencent Research Institute shows that 95.6% of respondents had used AI large models in the past three months, among whom 80.9% regarded AI as their preferred tool for solving everyday problems, using services such as information queries, itinerary planning, and answering questions.
As more and more users' retrieval needs shift to AI tools, the problems of outdated knowledge and hallucinations in large models have also begun to gradually surface. The knowledge of large models comes from training data, and after training is completed the weights are fixed, making it difficult to provide the latest information. In addition, guessing answers based on the vague memory of model weights can also cause large models to hallucinate.
Image source: Tencent
Taking Yuanbao Search as an example, it supports retrieval of highly time-sensitive content and can obtain the latest information in real time, such as policy updates, news developments, weather changes, and stock quotes. In addition, the product also connects with Tencent's high-quality content ecosystem, covering content resources such as Tencent News, Sogou Baike, and Penguin Account.
Search helps AI move from a tool you can talk to toward an information service entry point that can be used repeatedly. But entry-point status does not depend on whether it can answer a particular question; rather, it depends on how to continuously lower the cost for users to obtain and judge information. When all AI tools have internet connectivity, competition will further extend to content coverage, source quality, answer credibility, and other aspects.
03
After finding the answer, how can commercialization be achieved?
An important source of revenue for traditional search engines is search advertising. After users enter keywords, the platform displays commercial results based on factors such as ad bidding and relevance, and charges advertisers on a per-click basis and in other ways. Clear search intent makes search an important channel connecting users and merchants.
In this regard, at the end of 2024, 360 Group founder Zhou Hongyi said, "Actually, doing AI search is also very painful for us. We are also disrupting ourselves, and for now we can't find a place to put ads."
Taking Doubao as an example, it is trying to direct users' life service needs toward transactions and then take a commission from them.
In August 2026, some hotel orders completed through the Doubao channel began to apply independent rates, with a software service fee of 11.4% and an additional 0.6% payment processing fee, for a combined rate of 12%. Doubao responded that merchants pay the channel service fee only after an order is completed, and cannot influence recommendations or rankings by paying.
However, whether the user scale can be converted into revenue sufficient to cover costs is still a question that AI assistants need to answer. According to LatePost's calculations based on the publicly available API prices of Volcano Engine, in May 2026, Doubao consumed tens of millions of yuan per day, while in the first half of the year, the product's daily revenue was less than one million yuan, mainly from e-commerce commissions.
Image source: Tencent Cloud
In parallel with the commercialization of consumer-facing products, Tencent is also selling search capabilities as a developer service. In October 2026, Yuanbao Search began to be opened to external parties through Tencent Cloud. Currently, the prices listed on Tencent Cloud's product page for the Express, Standard, and Flagship editions are 14 yuan, 20 yuan, and 26 yuan per thousand calls, respectively. For Tencent, search can both directly serve users and become a foundational capability invoked by other AI applications.
From charging on transaction channels to opening up retrieval services, the value of search is extending in different directions: toward users, it has the opportunity to participate in consumption decisions, connecting information needs to actual transactions; toward developers, it strives to become a foundational capability that AI applications can invoke on demand. These two paths mean that search's revenue sources are expected to be further broadened.
But an increase in revenue sources does not mean the profitability problem has been solved. Transactions need to prove that they can bring incremental orders to merchants, and retrieval services need to find a balance among quality, response speed, and cost. In the next stage of competition in AI search, greater emphasis will be placed on whether each demand can bring sustainable returns, rather than simply expanding the scale of users and invocations.
For internet giants such as Tencent, Alibaba, and ByteDance, AI search has a deeper significance: whoever understands users' needs first has a better chance of allocating content, goods, and services. Needs that were previously handled separately by different apps may in the future first be judged by AI, which then connects them to the corresponding services. Changes in the search entry point may therefore also bring about a redistribution of distribution power and transaction opportunities among platforms.
However, how many application boundaries AI can cross still depends on content licensing, service integration, and fulfillment capabilities. Giants possess ecosystem resources, but they also need to face the question of whether recommendations are overly biased toward their own businesses; new players can redo the interaction experience, but they still need to fill in content supply, transactions, and after-sales service. Model capabilities affect answer quality, while whether reliable services can be connected determines how far things can go after the answer.
Compared with the winners, what is more noteworthy in this round of competition is how platforms maintain the balance between the two when AI both helps users make judgments and earns revenue from recommendations and transactions. What users hand over is not only a search request, but also their trust in information filtering and consumption advice.
Platforms that maintain user trust during commercialization are more likely to turn temporary usage popularity into lasting entry-point advantages.
This article comes from the WeChat public account "Shijiao Farsight", author: Haotian, published by 36Kr with authorization.