OpenAI loses to Qianwen; the chat window can't display the checkout counter

Title: OpenAI Loses to Qianwen, No Cashier in the Chat Box

Author: Dongcha Beating

Source:

Repost: Mars Finance

In March 2026, OpenAI officially announced the abandonment of Instant Checkout. Just six months earlier, this feature was praised by Silicon Valley media as a key function laying the foundation for the future of Agentic Commerce. Sam Altman even described it as a future cash cow for OpenAI at multiple events. Yet now, it has died with almost zero real transaction conversions.

Interestingly, on the other side of the Pacific, Alibaba’s Qianwen AI shopping is fully open for testing. Say “Get me a milk tea” in the chat box, and half an hour later, a delivery person will knock on your door.

Same AI, same chatbot—one is a crumbling fantasy retreating from the scene, the other is a real consumer power with actual money.

The same path, vastly different outcomes. The difference isn’t because the models aren’t smart enough.

When we talk about AI shopping, what exactly are we discussing? In the second half of 2025, this was the hottest track. ChatGPT had 900 million weekly active users. Even if only one in ten thousand casually bought something during a conversation, the revenue would be staggering. But between the high-profile launch and the six-month shutdown of Instant Checkout, many brutal stories unfolded.

No inventory, no supermarket

Rewind to September 2025, the day OpenAI launched Instant Checkout, and the entire retail industry was celebrating.

Shopify President Harley Finkelstein called it a new frontier in online retail. On launch day, Etsy’s stock surged 16%. Partners quickly jumped in—Etsy integrated US sellers’ products directly into ChatGPT, even covering transaction fees to speed up deployment. Walmart listed about 200,000 products. PayPal planned to connect its wallet to ChatGPT for checkout, and also promised to increase API and enterprise ChatGPT subscriptions. Stripe and OpenAI jointly developed the Agentic Commerce Protocol, aiming to set industry standards for AI-mediated transactions.

But six months later, the bubble burst.

OpenAI promised to connect with over one million Shopify merchants, but only about 30 actually went live. The landing page made for ChatGPT now silently redirects back to the official website. More alarmingly, internal data showed that while many users browsed and compared prices in ChatGPT, almost no one actually placed orders within the chat interface.

Walmart’s data shows that the conversion rate for checkout on the retailer’s website is three times higher than staying within ChatGPT. Forrester’s survey confirms this: among frequent AI Q&A engine users, completing purchases within the engine is the least adopted use case.

Why is shopping inside ChatGPT not working? Because OpenAI is trying to act as an e-commerce platform without owning any of the underlying infrastructure.

The shallowest reason is habit. People use ChatGPT like Google—to search for information and compare products. Once they’ve chosen, they go to a trusted site to pay. Asking users to input credit card info directly into a ChatGPT interface feels inherently unsafe. Users are willing to let AI help pick skincare products, but when it comes to paying, that cold, impersonal chat window can’t give them the reassurance they need.

And even if users are willing to pay, OpenAI can’t handle it.

As of February 2026, OpenAI still hadn’t built a system to collect and remit sales tax across US states. That’s infrastructure Amazon and eBay spent years developing. Fraud detection, returns, consumer protection—these are all complex issues. Ensuring real-time accuracy of prices, inventory, and delivery info across millions of products isn’t just a few lines of code; it’s a quagmire.

Forrester analysts pointed out that during its launch, Instant Checkout was riddled with errors, lacked support for multiple-item carts, had no promo codes, and even delivery info was opaque.

The most embarrassing part isn’t even OpenAI itself, but its partners.

PayPal invested engineering resources into integration and promised to increase API and enterprise subscription purchases. Now, the shopping benefits have evaporated, but the procurement commitments remain. According to insiders, PayPal and OpenAI are evaluating how to continue their relationship.

Etsy’s situation isn’t much better. It previously subsidized merchants’ transaction fees, but now it’s starting from scratch on its own ChatGPT app, with no clear fee structure yet. An Etsy spokesperson said they’re still unsure whether OpenAI will charge for in-app transactions.

Stripe’s position is somewhat better, as it already handles OpenAI’s subscription payments, which don’t depend on shopping features. But most partners lack this buffer.

For a company shifting focus to enterprise clients, such unpredictable cooperation is a huge risk.

Alibaba’s approach: a gift and a chain

On January 15, 2026, while OpenAI’s Instant Checkout was struggling, something happened on the other side of the world. Alibaba held a press conference in Hangzhou, announcing that the Qianwen app would fully integrate with Taobao, Alipay, Taobao Flash Sale, Fliggy, and Amap. At the event, Wu Jia, President of Qianwen’s consumer division, told his phone, “Help me order 40 Bawang Tea Ji’s Boya Juexian,” and half an hour later, a courier delivered it.

If we compare OpenAI’s Instant Checkout with Qianwen’s AI shopping, the key isn’t which model is smarter. The difference lies in who owns the entire chain—from product discovery to delivery.

Alibaba’s hundreds of millions of SKUs, native Alipay payments, and logistics network are all existing assets. When you say “Next week, I’ll hike in Siguo Mountain, what gear do I need?” it can generate a shopping list and place the order with a tap—smoothly, because there’s no cross-company data negotiation or profit-sharing involved.

That’s why domestic players like DeepSeek and Kimi can’t do this. Even with strong reasoning and long-text capabilities, without a real inventory and payment infrastructure, they can only provide a link. Alibaba has embedded the product catalog directly into the chat, while OpenAI is trying to make the chat window pretend to be a shelf.

But is this the perfect final form?

Alibaba succeeded because it dragged its massive commercial ecosystem into the large model. But when Qianwen acts as both judge and player, can its recommendations remain objective?

If I ask Qianwen which phone is best, will it prioritize recommending a product from Taobao for commercial gain? When an AI loses its neutrality and becomes a super shopping guide for its own e-commerce platform, can it still be called a general-purpose brain? Essentially, it’s using a heavy, old ecosystem to hijack a new technological entry point.

Qianwen’s “success” is not only a gift from the ecosystem but also a chain.

$50 billion and the elephant in the room

OpenAI also recognizes it lacks this heavy ecosystem. But that’s not the whole reason it’s giving up. The real elephant in the room is Amazon.

In late February, Amazon announced a $50 billion investment in OpenAI, becoming its exclusive third-party cloud provider for the enterprise platform Frontier. When your biggest backer is a giant that controls 40% of US e-commerce and is pushing its own AI shopping assistant Rufus, continuing to build a checkout system inside the app looks very unwise.

Moreover, this investment is a powder keg. Microsoft believes AWS hosting of Frontier violates its exclusive cloud agreement with OpenAI and is considering legal action. OpenAI’s lawyers are using technical terms like “stateful architecture” to bypass contractual obligations.

In the fierce competition among giants, OpenAI must make choices. In mid-March, CEO Fidji Simo announced a major strategic shift at an all-hands meeting: “We can’t miss this moment because of side projects.”

What’s driving this urgency is Anthropic’s rapid rise in the enterprise market. Its Claude Code and Cowork products have made it the top choice for corporate clients. Ramp credit card data shows that new enterprise customers choose Anthropic three times more often than OpenAI.

Last year, OpenAI spread itself too thin: Sora video generation, Atlas browser, Jony Ive hardware, e-commerce features, advertising, adult mode.

Now, they must refocus on two core battlegrounds: coding tools and enterprise clients. After all, earning from enterprise clients is much more reliable than scraping a few points of transaction fees in a chat window.

OpenAI expects enterprise revenue to account for half of its total this year, up from around 40%. To achieve this, it plans to double its staff from 4,500 to 8,000, mainly in engineering, research, sales, and product development.

In San Francisco, OpenAI signed a new lease, expanding its office space to over 1 million square feet.

The real battlefield for AI shopping isn’t the checkout

OpenAI’s retreat doesn’t mean AI shopping is dead. On the contrary, the upper funnel has been completely reshaped.

More than half of American consumers are now used to letting AI help make decisions. They no longer search for vacuum cleaners and scroll through ten pages of ads; they ask directly, “Which one offers the best value?” Discovery, research, comparison—all these actions are moving upstream, eroding the value of retail channels.

But the last mile of transaction closure doesn’t require smarter models; it needs more complete infrastructure.

OpenAI has clearly stated it will prioritize product search and discovery, and will add advertising within ChatGPT. This is its way of monetizing the discovery layer—much more realistic than building its own checkout system.

Finally, the most likely path for Amazon to replicate Alibaba’s success isn’t OpenAI, but itself. It has user profiles, product graphs, payment channels, and fulfillment infrastructure.

OpenAI’s failed attempt to build an e-commerce loop on its own, then taking money from the largest e-commerce platform, might ultimately turn it into a traffic gateway for that platform.

In China, Alibaba’s full-stack advantage has led Qianwen down a different path, but only Alibaba can make that work. Wu Jia, President of Qianwen’s consumer division, said a key point: “The overall agent competitiveness is very strong. Vertical agents are increasingly proven to be stage products. In the future, there won’t be many standalone AI applications as entry points.”

Translated into business language: the future winners of AI shopping loops will be those platforms that already have a complete ecosystem, not AI companies building everything from scratch.

You can’t grow a cashier in a chat box, but if the cashier is already in your store, placing a chatty AI beside it feels natural. That’s the most important lesson for AI shopping in 2026.