Alibaba has done something no Western marketplace has matched at this scale: it opened its entire Taobao and Tmall catalogue, over 4 billion products, to its Qwen AI assistant. Chinese shoppers can now describe what they want in plain language and let the AI browse, compare, place the order, and handle delivery and after-sales, all through conversation instead of keyword search. Qwen reached 300 million monthly active users across Taobao, Tmall, and Alipay, and Alibaba logged around 140 million first-time AI shopping experiences during the Chinese New Year campaign alone. For foreign brands selling in China, this is not a distant future feature. It is a change in how consumers find products, live now, on the largest e-commerce platform in the country.
What Alibaba Actually Launched
The integration gives the Qwen app access to the full Taobao and Tmall product catalogue, backed by AI agents that carry a skills library covering order management, logistics, and after-sales service. In practice a shopper can type or say something like “I need a gift for my mother who has trouble sleeping, under 300 RMB, delivered before the weekend,” and the assistant scans billions of listings, narrows by budget, brand, and delivery window, reads through the review database, and presents a short list. The purchase happens inside the chat. The assistant also offers virtual try-ons and a price-comparison tool so buyers can check they are getting a fair deal without opening five browser tabs.
This is the difference between a chatbot and an agent. A chatbot answers questions. An agent completes tasks. Qwen on Taobao is being built as an agent, and the shift from “search and buy” to “describe and let the AI buy” is the most important interface change in Chinese e-commerce since mobile overtook desktop.
Why This Matters for Foreign Brands
Discovery stops being about keywords
For a decade, winning on Tmall meant winning the search box. You optimised your title, packed in the keywords Chinese buyers typed, bid on the high-intent terms, and fought for ranking. When an AI agent does the searching, the game changes. The agent is not matching your title against a query. It is reading your full product data, your reviews, your specifications, your pricing, and your delivery promise, then deciding whether you fit what the shopper described. The brand that wins is the one whose data is clean, complete, and honest enough for a machine to recommend with confidence.
Reviews become machine-readable trust
Qwen is trained on Taobao’s massive review database. When the agent decides which products to surface, it weighs what real buyers said. This raises the value of genuine, detailed reviews and lowers the value of review manipulation, because an AI reading thousands of reviews spots the pattern of fake ones faster than a human scanning the first page. Brands that have invested in real customer satisfaction and detailed post-purchase content are better positioned than brands that bought their way up the rankings.
Structured product data is now a competitive asset
An AI agent can only recommend what it can understand. A product page written for a human browser, with the real information buried in a long image, is invisible to an agent that reads structured text. The brands that will be recommended are the ones whose specifications, ingredients, sizing, use cases, and compatibility are written out clearly in machine-readable fields, not baked into a graphic. This is a concrete, unglamorous task that most foreign brands have not done, and it is becoming the difference between being found and being skipped.
What This Does Not Change
It is worth being clear-eyed. Agentic shopping is growing fast, but it does not yet replace the discovery that happens on Douyin and Xiaohongshu, where Chinese consumers still encounter new brands through content and creators. What Qwen changes most is the consideration and purchase stage: once a shopper knows roughly what they want, the AI compresses the comparison and checkout that used to take many taps. Brand awareness is still built through content. The AI agent then acts on the awareness that content created. So this is not a reason to move budget out of Xiaohongshu seeding or Douyin livestreams. It is a reason to make sure that when the agent evaluates you against competitors, your store data holds up.
What Foreign Brands Should Do Now
- Audit your product data for machine readability. Every specification a buyer might filter on should exist as text, not only inside an image. If your key selling points are trapped in a graphic, an AI agent cannot use them.
- Invest in genuine reviews. The AI weighs real buyer sentiment. Post-purchase follow-up, product education, and customer service that earns honest positive reviews now compound into AI recommendations later.
- Keep pricing consistent. The agent surfaces a price-comparison tool. Inconsistent pricing across your Tmall, JD, and Douyin stores makes you look unreliable to a machine that checks all of them.
- Write for the question, not the keyword. Think about the real-life problems your product solves and make sure your listing answers them in plain language, because that is what shoppers now describe to the AI.
The Bigger Picture
China is often ahead of the West in e-commerce mechanics, and agentic shopping is another example. Alibaba is not experimenting at the edges, it has put an AI agent in front of 300 million people and connected it to its entire catalogue. Every other Chinese platform is building its own version. Within a year or two, a meaningful share of Chinese purchase decisions will pass through an AI layer that reads your store on the buyer’s behalf. The brands that treat their product data as something written for machines as well as humans will be recommended. The brands that do not will quietly disappear from the shortlist, without ever knowing why their traffic fell.
How the AI Decides What to Recommend
It helps to understand, in plain terms, how an agent chooses. When a shopper describes what they want, the AI turns that description into a set of requirements: price range, category, attributes, delivery window, and softer signals like “good for sensitive skin” or “suitable as a gift.” It then filters the catalogue against those requirements and ranks what survives. Ranking is where reviews, ratings, return rates, store reputation, price competitiveness, and fulfilment reliability all feed in. A product that meets the requirements but has a high return rate or inconsistent pricing gets pushed down. A product with complete data, strong genuine reviews, and reliable delivery gets pushed up.
This is why the unglamorous operational fundamentals now matter more, not less. Return rate, on-time delivery, customer service response, and pricing discipline were always good practice. In an agentic world they become ranking inputs that decide whether the AI recommends you at all. A brand cannot charm an agent with a clever campaign. It can only give the agent clean data and a track record the algorithm reads as reliable. The brands that treat these fundamentals as a checkbox will lose quietly to the ones that treat them as the product.
Sources: Alibaba Group corporate announcements 2026; South China Morning Post coverage of Qwen and Taobao integration 2026; EAC Ecommerce China Agency platform analysis
Harry Huang, Founder of EAC Ecommerce China Agency. Harry founded EAC to help foreign brands sell in China without the guesswork. Before EAC he worked on digital and ecommerce projects at Volkswagen in China, combining multinational reporting discipline with a practical grasp of how Chinese platforms convert traffic into sales. EAC is an independent agency running Tmall, JD.com, and Douyin stores for foreign brands, with one metric in mind: revenue per RMB spent.
Want your China store ready for the AI shopping era? Get a free audit from our team. Connect with Harry Huang: ecommercechinaagency.com/author/philip/
