Qwen and Doubao closed China’s AI shopping loop. Paid ranking is the next fight.
Alibaba and ByteDance have moved AI shopping from demo to checkout before 618. Qwen can search, compare, and place Taobao orders inside Alibaba’s AI app.
Qwen-Taobao and Doubao-Douyin are China’s first AI commerce deployments at real platform scale ahead of 618 2026, a key mid-year shopping festival in China.
The two products reflect different theories of value capture: Qwen wants the assistant to remain visible; Doubao wants the transaction to feel effortless.
Both models run into the same conflict: users expect neutral recommendations, while platforms still depend on paid ranking.
China’s GEO market grew 215% YoY in Q2 2025, giving brands a new way to shape AI answers before disclosure rules catch up.
618 will not settle the business model. It will make the conflict harder to hide.
China’s 618 shopping festival, launched by JD.com in 2010, is the country’s second-largest annual e-commerce event. Every June, platforms run weeks of coupons, flash deals, merchant subsidies, and category-specific promotions. The festival tests logistics, pricing systems, recommendation engines, and consumer patience.
In 2026, 618 is testing something else: whether AI assistants can replace the search box as the first step in a purchase decision.
Alibaba and ByteDance moved first. Their answers to the same product problem look similar on the surface. Underneath, they point to very different assumptions about where value will sit in AI commerce.
Two products, two theories of value
On May 11, 2026, Alibaba linked its Qwen AI app to Taobao. Users can now search, compare, and complete orders through natural language inside Qwen. Within Taobao itself, a Qwen assistant helps with virtual try-on, discount calculations, and deal hunting. The catalog spans more than 4 billion products.
ByteDance moved in stages. Doubao’s first Douyin commerce integration began in December 2025, covering ticket bookings and selected shopping categories. In-app shopping gray-scale testing followed in March 2026. The full “Help You Choose” closed-loop feature went live in May 2026.
The two platforms enter 618 with different user bases. According to QuestMobile data, Doubao had roughly 345 million monthly active users in Q1 2026, compared with 166 million for Qwen. Qwen’s growth curve still matters: it reached 100 million MAU within two months of its November 2025 public beta launch.
Scale alone will not decide the race. Doubao’s user base was built through entertainment-driven engagement. Qwen’s was built closer to productivity, search, and shopping-adjacent intent.
618 will test which base converts when the question changes from “what should I watch?” to “what should I buy?”
The product design split is more important than the user count.
Qwen surfaces category results and keeps the Qwen frame visible. Ask for a product recommendation and Qwen returns a curated list. The user still sees Qwen as the filter and makes the final choice.
Doubao pushes harder toward a single answer. Ask the same type of question and Doubao can return a named product on a purchase page. The Doubao interface fades away as the transaction layer takes over.
Qwen is betting on assistant brand equity. The user should remember that Qwen helped make the decision.
Doubao is betting on transaction efficiency. The fewer steps users feel, the more useful the system becomes.
That split matters because AI commerce will not be judged only by recommendation quality. It will also be judged by who gets credit for the decision. If the user remembers the assistant, the AI layer gains long-term bargaining power. If the user remembers only the purchase, the commerce layer keeps the value.
China is hardly alone in this race. Amazon’s Rufus, Walmart’s Sparky, and ChatGPT’s shopping integrations all point toward agentic commerce. China’s structural difference is ownership. Alibaba and ByteDance control both the AI layer and the commerce layer, which makes the experiment faster and the conflict sharper.
Neutral recommendations collide with paid ranking
Qwen and Doubao have different product designs, but they share the same monetization problem.
AI shopping recommendations depend on trust. Users ask an assistant for help because they expect a better match than they would get from keyword search, which has been shaped for years by sponsored placement and merchant bidding.
Paid ranking works in the opposite direction. China’s e-commerce ad systems are built around auction logic. Merchants pay for visibility. The platform monetizes that visibility.
Alibaba’s customer management revenue is the clearest example. It remains the core of Alibaba’s highest-margin China e-commerce business and grew 8% on a like-for-like basis in fiscal year 2026.
The scale of that line is the problem. Customer management revenue is the largest single revenue stream inside Alibaba’s China e-commerce business. Any serious shift from keyword bidding to AI-mediated recommendations pressures one of Alibaba’s most profitable engines.
According to CIW’s analysis, Alibaba’s China E-commerce Group adjusted EBITA fell 40% YoY to RMB24 billion in the March 2026 quarter. That earnings base now supports Qwen app user acquisition, quick commerce expansion, and the broader AI push.
Alibaba is funding the AI layer from a weaker commerce core. The AI layer, if it works as promised, can reduce the value of the advertising revenue that core depends on.
ByteDance faces the same pressure from the other side. Douyin’s advertising revenue, driven by auction-based placement, is ByteDance’s primary commercial engine. Connecting Doubao to Douyin’s commerce layer places an AI decision system inside a platform that sells merchant visibility.
The tension is direct. The AI assistant earns user trust by making neutral recommendations. The ad layer earns money by selling ranking advantage.
Alibaba and ByteDance have both signaled that paid placements will stay out of AI recommendation results at launch. That buys time. It does not solve the business model.
A neutral AI assistant can reduce the return on merchant ad spending, especially for large brands that have spent years buying keyword visibility. A sponsored AI assistant risks losing the trust that made users try it in the first place.
A compromised AI shopping assistant becomes a familiar product with a new interface: search, but dressed as conversation.
That is why the platforms are quiet on monetization. The product can launch before the revenue model is settled. Trust cannot be rebuilt as easily.
