To The Moon

Tencent is adding a crucial piece to its Yuanbao AI assistant.

On June 15th, Tencent announced that Yuanbao has officially integrated the Ima knowledge base. This allows users to directly access content from Ima's public knowledge repository during searches within Yuanbao, receiving AI-generated answers that include source citations and are traceable. By clicking on a citation card, users can jump directly to the original Ima source and add it to their personal knowledge base with one click.

For Yuanbao, integrating with Ima addresses a core and challenging issue for AI products: answer credibility. The competitive edge among leading AI assistants is increasingly shifting from model capability contests to competitions over context supply. User dissatisfaction often stems not from the model's inability to answer, but from not knowing the source of the answer, its reliability, or whether it can be further utilized.

On the surface, this is an integration of knowledge base capabilities. However, viewed within the context of Tencent's AI product evolution over the past year, it appears more like the latest step in Yuanbao's transformation from a simple chatbot into an entry point for an ecosystem-based Agent.

Tencent's vast ecosystem is undoubtedly Yuanbao's most significant competitive advantage. It is evident that Yuanbao is continuously extending deeper into Tencent's internal ecosystem. Previously, Yuanbao has progressively integrated content sources like WeChat, Tencent News, Tencent Sports, and Tongcheng Travel, gradually opening Tencent's internal content assets for AI utilization. Simultaneously, through "Yuanbao Pai," it supports the integration of more agent capabilities, enabling users to invoke tools to complete tasks without needing to switch applications.

As a personal knowledge Agent launched by Tencent in late 2024, Ima is transitioning from a standalone product to a foundational knowledge layer role. Beyond its integration with Yuanbao, it has also been incorporated into product systems like WorkBuddy and QClaw, creating a unified knowledge context through a "deposit once, call from multiple endpoints" approach.

At the recent 2026 Tencent Cloud AI Industry Application Conference, Dowson Tong, Senior Executive Vice President of Tencent and CEO of the Cloud and Smart Industries Group, stated that Tencent's approach to AI consistently aims for practicality, usability, and scalability. He noted that Tencent's rich product scenarios, interaction data, and ecosystem connections can provide high-quality context for models and enable models to invoke tools and connect systems, thereby truly completing task loops.

Consequently, as the AI race progresses, Tencent primarily adopts a logic of cross-pollination, embedding, and co-design for its AI product and organizational structure. Tencent recently launched an efficiency agent toolkit, structuring agent capabilities into a multi-layered product system centered on user needs.

For individual users, Tencent offers several products, including the AI assistant QClaw, the personal knowledge agent Ima, and the personal AI assistant Yuanbao. For workplace users, Tencent has introduced the "Buddy" family series, including the AI agent desktop workstation WorkBuddy Personal Edition and the coding agent CodeBuddy.

Even as the number of agent products grows, Yuanbao maintains a very important position within Tencent's AI strategy. Tong revealed that among all these agent products, Tencent's largest investment remains in Yuanbao. This is because Tencent still believes that chatbots represent the most universally demanded and widely applicable important segment.

However, judging by the monthly active user performance of standalone native AI apps, Yuanbao still faces significant pressure. QuestMobile data shows that in March 2026, the top three apps—Doubao, Qianwen, and DeepSeek—had monthly active users of 340 million, 166 million, and 127 million respectively, while Yuanbao ranked fourth with 57.346 million.

In terms of user engagement, however, Yuanbao performs better than Qianwen. In Q1 2026, Doubao averaged 54.8 uses per user per month, DeepSeek reached 41.7, Kuaidui AI was at 27.6, Yuanbao achieved 25.9, and Qianwen stood at 19.8.

The AI assistant market has now entered its second phase of competition. The first phase focused on models and user acquisition spending, while the second phase is about competing on ecosystem density and task completion capabilities.

Regarding Yuanbao's future development goals, Tong stated the clear objective is to win more users and improve retention rates. "My highest expectation for the Yuanbao team is to continuously improve the search service, introducing more real-time and authoritative professional data."

Moving forward, the real challenge for Yuanbao is to prove whether it can leverage Tencent's massive product ecosystem to build its own capabilities as a super Agent.