Four billion items. That is the size of the product pool Alibaba handed to Qwen today.
This is not a search-box plug-in or another recommendation feed. Alibaba is putting the whole transaction chain in the hands of an AI agent: finding products, comparing prices, running virtual try-ons, placing orders, paying and handling returns or exchanges. In the Qwen app, a user can ask it to look at a pair of shoes, and the agent can browse Taobao and Tmall inventory, recommend options, compare sellers, check 30-day price history and finish payment through Alipay. The human step is reduced to the final confirmation.
This is not another shopping assistant
The weight of the move becomes clearer when set against what Western players are doing.
ChatGPT's Shopify integrations and Amazon's Rufus still mostly sit at the layer of answering shopping questions. After the conversation, the user usually has to move to a retailer page to pay, receive the goods and deal with after-sales service. The AI is an adviser, not the party that carries the transaction.
Alibaba is taking a different route. The agent holds the transaction from end to end: identity, payment, logistics and after-sales are kept in one flow. Wu Jia, Alibaba Group vice president, summed it up plainly as a move "from intelligence to agency".
That works because China's e-commerce system is both closed and deeply integrated:
- The Qwen app had reached 300 million monthly active users by early 2026.
- During the Lunar New Year period, it generated 140 million first-time AI shopping experiences.
- The product pool connects to 4 billion SKUs across Taobao and Tmall.
- Payment runs through native Alipay, without forcing users out of the AI interface.
- A skills library sits behind it for logistics and after-sales support.
For ChatGPT to build the same loop, it would have to persuade dozens of retailers, several payment companies and multiple logistics providers to support its protocols. Alibaba can plug its own infrastructure straight into the agent. The closed loop was already there.
Alibaba was pushed into this move
The important point is that Alibaba did not make this move in a vacuum. It was forced by rivals.
| Platform | Playbook | Current position |
|---|---|---|
| Pinduoduo | Algorithm plus low prices aimed at lower-tier markets | Eating into Taobao's mid-market users |
| Douyin e-commerce | Content, streamers and interest-based recommendations | Taking over the act of browsing |
| Taobao | Traditional search and shop-based logic | Pressed from both sides |
Over the past two years, Alibaba has tried structural separation, with cloud, e-commerce and local services operating more independently. The Qwen-Taobao integration moves in the opposite direction: AI capability built on the cloud side is being pushed back into the retail front line. In effect, Alibaba is admitting that separate battles were not enough.
The logic is straightforward. If users search, compare, decide and pay inside Qwen, Pinduoduo and Douyin's recommendation engines never get a chance to reach them. The AI agent becomes the new entrance, and the old front-end traffic battlegrounds -- homepages, search boxes and feeds -- all need to be revalued.
Why the West cannot easily follow
There is a structural problem here: Chinese e-commerce can run this model, while Western e-commerce cannot do it easily yet.
This is less about technology than ecosystem design.
Amazon has Rufus, but it has stayed at the advice layer because US consumers shop across platforms. A user might compare on Amazon, buy at Best Buy and return at Walmart. For Rufus to run the full chain, it would need to connect far more than Amazon's own storefront.
ChatGPT's Shopify route is more awkward. Shopify is merchant infrastructure, not the payment party and not a logistics operator. If OpenAI wants a full agentic transaction loop, every link has to be negotiated again.
Alibaba has Taobao, Tmall, Alipay and Cainiao logistics under the same roof. Once AI is plugged in, it is plugged into the whole family. The kind of closed Chinese ecosystem often criticised in the past suddenly becomes an advantage in the agent era: a closed loop is exactly what allows an agent to act.
The calculation behind the numbers
The 300 million MAU figure deserves a closer look.
Last year, the Qwen app was still at the million-user level. By early this year it had jumped to 300 million, largely because Qwen was embedded across Alibaba surfaces including Taobao, Tmall, Alipay and Quark. After this deeper e-commerce integration, the Qwen app has a new role.
It is no longer just another AI assistant. It is becoming the entrance for a shopping agent.
Alibaba has not disclosed core metrics such as conversion, repeat purchase rate or return rate. But one comparison matters: the 140 million first-time AI shopping experiences during Lunar New Year are close to one quarter's active buyer scale for Douyin e-commerce. If Alibaba can retain those users, Taobao's moat shifts from "the most goods" to "the AI that knows you best".
The second half of the year will be about conversion data. Whether an agent can replace the act of scrolling through products should become clearer within six months.
Sources: Alibaba integrates Qwen AI with Taobao for end-to-end agentic shopping (The Next Web), Alibaba to integrate Qwen AI with Taobao, CocoLoop, launch agentic shopping (Inside Retail Asia), Alibaba to merge Qwen AI with Taobao (TechNode)