After two years of chatting about AI shopping, why can we finally place an order today?
Text by | Wang Yi
A user said to Qianwen, "I want to lose weight. Please recommend some training equipment for me."
Qianwen's response was: It's not recommended to buy. The equipment you have already meets the basic needs for aerobic and strength training. The problem may not lie in the equipment. Then, the topic shifted back to the training plan and how to stay committed.
An AI shopping assistant advising you not to buy - this might be the most human - like aspect of an AI shopping assistant.
When an AI says "You don't need this", it truly understands your situation: what your goal is, what you already have, and where your bottleneck lies. "Reverse recommendation" based on in - depth understanding means that AI shopping has entered a new stage of truly assisting decision - making.
On May 11th, Qianwen was fully integrated with Taobao. Users can complete product selection, comparison, and place orders on Taobao by having a conversation with the AI in the Qianwen App. The "Qianwen AI Shopping Assistant" was also launched in the Taobao App.
Before this, global tech giants had been testing the waters in the AI shopping field for two years. Amazon stated that its AI shopping assistant, Rufus, had accumulated over 250 million users in 2025 and is expected to bring an additional $10 billion in annual sales to the company. In January 2026, Google also announced a partnership with retailers such as Walmart and Shopify to launch an AI shopping function in Gemini.
On the other hand, OpenAI high - profilely launched the "Instant Checkout" function of ChatGPT in September last year but announced its abandonment in March this year.
A global competition about "Can AI help people make purchases?" has begun, with different paths emerging.
01. Three Waves in a Decade, the Large Model is the Watershed
The scenario of AI helping people make purchases has appeared in science - fiction works for at least 30 years. In "Iron Man", Jarvis could complete an order just by speaking. In the 2013 movie "Her", the AI assistant could handle all daily affairs during a conversation. When Amazon launched Echo in 2014, Jeff Bezos' vision was "to be able to shop by saying a single sentence to the AI".
However, the progress in reality has been much slower than expected.
The first wave of attempts occurred during the era of voice assistants. Between 2014 and 2018, Amazon's Alexa Shopping and Google's Google Shopping Actions were launched one after another. Users could say to the speaker, "Buy me another box of milk." But it could only handle extremely simple repurchase instructions - clear product categories, clear brands, and clear quantities. At that time, natural language understanding could only perform the most basic field extraction. For slightly more complex expressions, voice assistants couldn't understand.
The second wave was Conversational E - commerce 1.0. Around 2018, various e - commerce platforms launched intelligent customer service one after another - Taobao's "Wenwen", JD.com's JIMI, and Amazon's customer service robot. But they were initially positioned as after - sales customer service rather than shopping advisors. In essence, they were "FAQ databases + finite state machines" and could handle returns, exchanges, and logistics inquiries. Asking them to help with shopping decisions? That was beyond their scope.
The third wave of attempts is the variable that has re - energized the entire AI field in recent years: the large model. The bottleneck of previous AI assistants ultimately boils down to a technical problem - machines couldn't understand what people were saying. The large model made it possible for the first time for "machines to understand intentions".
Since 2024, players have entered the field one after another: Amazon launched Rufus, and Perplexity launched "Buy with Pro"; in 2025, OpenAI integrated the shopping function into ChatGPT. AI shopping has entered the stage of real - world product competition from concept verification.
With the integration of Qianwen and Taobao this time, the form of the AI shopping assistant is more complete. It has truly become an "agent" involved in the core of transactions. It not only understands users' needs but also can mobilize the platform's capabilities to complete real transactions and services.
Making AI land in real - world business scenarios has been the most core proposition in the industry in the past two years. 2025 is generally regarded as the "Year of Agents" in the industry. AI has shifted from "answering questions" to "helping people do things" - programming agents, research agents, and customer service agents have emerged in various vertical fields. Shopping is the next scenario that everyone can think of: everyone is a consumer, and there is a natural transaction loop.
According to a report by Morgan Stanley Research, it is estimated that by 2030, the spending of Agentic in US e - commerce will reach at least $190 billion, accounting for 10% of the market share. Many research institutions such as McKinsey and Gartner have also released market forecasts for 2030. Although the specific scales vary, the industry has reached a high degree of consensus on the development trajectory of this market.
02. A Smart Brain Alone is Not Enough, Global Players Take Three Different Paths
AI will reshape all industries. What was once a prediction at the birth of OpenAI has now become a consensus. However, at the implementation level, the evolution of large models on the consumer side has been much slower than on the office side because it involves real products, transactions, and logistics.
There is a fundamental difference between AI shopping and AI writing or AI programming. AI shopping is not only about information processing. It needs to help you complete a series of actions in the real world: actually buy the products, deliver them to your home, and handle returns if there are problems. These aspects are beyond the capabilities of a single model and require a complete set of real - world business infrastructure to support.
From 2024 to 2025, global tech giants have tried to find answers. Large - model companies are eager to implement real - world business scenarios, and Internet trading platforms want to seize the opportunity of AI transformation. However, they have taken three different paths.
The first path is for model companies to seek external cooperation, with OpenAI as the representative. In September last year, ChatGPT launched the Instant Checkout function, allowing users to directly complete the checkout during a conversation by connecting to Shopify. However, according to The Information, the function only lasted for about six months before being discontinued - actual tests showed that users preferred to use ChatGPT as a product research tool rather than a transaction terminal, and the number of actually connected merchants was less than 20.
"Model companies can't access the most core data of e - commerce platforms, and e - commerce infrastructure cannot be fully opened to external companies. There is always a wall between the two companies. All AI can do is 'recommend + redirect', and it's difficult to go deeper." An e - commerce practitioner told 36Kr.
The second path is for e - commerce platforms to develop their own AI assistants, with Amazon as the representative. Rufus is directly embedded in the Amazon App. Users can have conversations with the AI, compare prices, and check reviews. According to Fortune, Rufus has accumulated over 250 million users, and the probability of buyers using AI to complete a purchase is 60% higher than that of ordinary users.
However, the limitations of Rufus are also obvious. Some industry voices point out that Rufus lacks stability and performs poorly in handling problems outside the in - station database, and the actual conversion rate of transactions through conversations is not high. Behind this is the fact that the performance of Amazon's self - developed large model has not reached the leading level.
The third path is the in - depth integration of large models and e - commerce ecosystems. Alibaba is the representative. The full integration of Qianwen and Taobao this time is the first in - depth integration of a 1 - billion - user - level e - commerce platform and a top - level large - model application.
An analyst who has long been concerned about AI told 36Kr, "The competition in AI shopping may seem like a competition in model capabilities, but in fact, it is a competition in ecosystem integrity. If you look at global players, there are actually very few that can meet both the conditions of 'a sufficiently smart large model' and 'a sufficiently complete e - commerce infrastructure'."
Ultimately, the choice of path depends on the company's own resource endowment.
Model companies lack transaction scenarios and fulfillment capabilities, so they can only seek external cooperation; e - commerce giants have a complete supply chain and data, but their large - model capabilities are limited; players with both large models and physical businesses have a better chance of creating a complete AI shopping experience. This is not only a choice of technical route but also a fate determined by the company's genes.
03. Qianwen's Integration with Taobao: It's Not Just "Adding Taobao"
Compared with the US market, China's long - term advantage in the AI field lies in more abundant and mature application scenarios for technology implementation. A large user base, diverse industrial forms, and rapid market acceptance together constitute complex and rich application scenarios.
Chinese tech giants have a comprehensive application layout in consumer scenarios. In the field of AI shopping, Alibaba has the most complete lifestyle service forms, and Qianwen has also joined the ranks of world - class large models. With this unique advantage, Alibaba has chosen to integrate the entire AI shopping chain within its ecosystem, building a hard - to - replicate barrier.
Alibaba has encapsulated all kinds of e - commerce capabilities accumulated by Taobao over more than 20 years - search, price comparison, order placement, logistics tracking, after - sales returns and exchanges - into "Skills" that can be called by AI. This AI shopping assistant may be the largest - scale commercial - level Agent application currently available for C - end users.
Now, all processes from product recommendation to order placement, fulfillment, and after - sales can be completed in the Qianwen App, rather than just a shallow interaction of redirecting links. This is a fairly comprehensive form of AI shopping service in the industry at present.
Teaching the model to understand when to search, when to compare prices, when to place an order directly, and when to advise the user to think again - these judgments themselves require a large amount of training in real shopping scenarios.
This understanding ability points to more critical data assets. Based on Taobao's 4 billion product library and more than 20 years of accumulated real - world shopping scenario data, Qianwen can accurately understand users' consumption intentions during conversations and provide more accurate recommendations.
The well - known investor Zhu Xiaohu once said that once the basic capabilities of large models become a relatively stable platform, the core of competition will shift to engineering implementation capabilities and the construction of industry data loops. And this is exactly the area where Chinese companies excel.
This actually reflects a deep - seated shift in the competition of large models. When the focus of AI development shifts to the realization of value at the application level, this change means that the competition is no longer just about the intelligence level of the model but also about the ability to transform technology into actual products, meet specific user needs, and achieve commercialization. As the importance of data increases, high - quality data will tend to be closed and become the unique moat of each company.
The significance of this may go beyond shopping itself. Since January this year, Qianwen has been gradually integrated with the service capabilities of Taobao Flash Sale, Fliggy, Gaode, Alipay, and other services within the Alibaba ecosystem. This full integration with Taobao further fills a key link in the consumer scenario.
Qianwen has transformed from a "conversation tool" into a super AI portal that can mobilize a complete set of lifestyle services - shopping, booking hotels, hailing a car, checking routes, and making payments, all happen during a natural conversation.
In fact, the same idea can also be seen in competitors. Doubao is accelerating its integration with Douyin E - commerce, and JD.com and Meituan have each launched independent AI shopping applications, trying to build barriers with vertical scenarios. A battle for the "AI lifestyle portal" is underway.
How to make users go from "giving it a try" to "being unable to do without" - this question still needs time to provide a definite answer. But one thing is becoming clearer: when people gradually get used to completing shopping during conversations, will the e - commerce and even the entire consumer market still be the same as it is now?
From the moment Qianwen advises you "Don't buy", the outline of the future may have emerged.