AI-powered shopping is disrupting cross-border e-commerce: Amazon is "shutting down" and ChatGPT is "dismantling its firewall."

At the start of 2026, ChatGPT accelerated its commercialization efforts, launching a transaction commission system and officially testing advertising features in the US market. The advertising cost has even been disclosed: per mille, almost three times that of Meta.

On the other hand, Google also announced a brand-new universal business agreement that covers the entire chain from product discovery, recommendation, ordering, payment, to order management and after-sales support, enabling AI agents (such as Google's AI Search or Gemini) to execute the complete shopping process across platforms and systems.

AI-powered conversational tools are beginning to resemble a new type of e-commerce platform. Although significant sales figures are not yet visible in the short term, they are already altering users' shopping habits to some extent. Adobe data shows that during the 2025 holiday shopping season, traffic driven from AI recommendations to retailer channels increased by nearly 700% year-on-year. Some organizations predict that by 2030, AI agents may influence 15%-25% of e-commerce sales.

Meanwhile, traditional e-commerce platforms seem to be getting wary. Amazon, eBay, and others have taken steps to ban AI agents from crawling and recommending products on their platforms (note: currently, eBay is the only platform that has integrated ChatGPT), shutting themselves down to create their own in-site AI shopping assistants, determined to protect their existing traffic.

Following the previous divergence between "independent websites vs. third-party platforms," the cross-border e-commerce industry seems to have reached another watershed: AI conversational tools represented by ChatGPT and Gemini, and traditional e-commerce platforms represented by Amazon and eBay, may be about to witness a new battle for traffic entry points in the era of AI e-commerce.

01 The Beginning of the AI Gateway War: "Open Networks" vs. "Walled Gardens"

Regarding AI shopping, the mainstream players are showing very different tendencies: one side is moving towards complete openness; the other is building walls first and observing before taking action.

The former, represented by OpenAI and Google, chooses to embrace all ecosystems, enabling cross-platform and cross-channel access to product information and shopping scenarios. Through natural language interaction, intelligent recommendations, and other methods, it transforms users' shopping intentions into concrete actions.

Over the past six months, OpenAI’s ChatGPT, Google’s Gemini, and Microsoft’s Copilot have successively launched in-site payment functions in conjunction with payment service providers, and have successively reached cooperation agreements with platforms such as Shopify, Etsy, Walmart, and Target. Retailers including Glossier, SKIMS, and Spanx have become the first brands to integrate ChatGPT.

Platforms like Amazon and eBay, on the other hand, adopt a relatively "closed" approach. They center their operations on their own data, traffic, and transaction systems, controlling resources such as search results, ad placements, and in-site recommendations to keep user traffic and transactions within their own ecosystem, creating a closed but deeply customized user experience. For them, this model maximizes advertising revenue and in-site transaction value.

Previously, reports indicated that Amazon had used technical means to restrict ChatGPT and Gemini from scraping product data on its platform; many users on the social media platform Reddit also reported that they could no longer see Amazon products at the top of AI recommendation results; eBay also recently updated its user agreement, explicitly prohibiting third-party agents and AI assistants from accessing the eBay platform without permission.

This is not the first time Amazon has clashed with an AI gateway. Previously, Amazon threatened legal action against Perplexity's AI browser Comet for its automated shopping behavior on its platform, arguing that it violated platform rules and could harm the user experience; Perplexity, in turn, accused Amazon of suppressing innovation and competition.

Image: The Information

Foreign media commented that this is consistent with Amazon's historically strict control over its retail business, such as its large-scale investment in building its own logistics, storing payment information, and developing in-site advertising.

Following the same line of thinking, in order to keep search traffic and advertising revenue within its own ecosystem and prevent them from being diverted by external AI entry points, Amazon is also accelerating the construction of its own AI shopping path. For example, Amazon's built-in AI shopping assistant Rufus has been constantly evolving over the past two years. In its Q4 2025 financial report, the company disclosed that Rufus has been used by over 300 million customers, and customers using Rufus have a 60% higher probability of completing a purchase. In the past year, Rufus has helped generate nearly billion in sales growth, far exceeding internal expectations.

An Amazon spokesperson said, "We see a significant opportunity to innovate and further improve the user's shopping experience. We can offer both an online store and an AI shopping assistant."

02 The landscape of cross-border e-commerce is changing: Independent websites are the first to reap the benefits of AI traffic.

Regardless of Amazon's stance, the series of actions taken by AI conversational tools such as ChatGPT and Gemini have already begun to shake up the entire cross-border e-commerce landscape, impacting various roles and links in the chain. However, unlike previous rounds of platform battles, this war has involved a sufficient number of players from the outset.

Under the logic of AI recommendations, traditional traffic allocation methods and strategies are undergoing tremendous changes. Unlike the past practice of competing for traffic through search keywords or platform algorithms, ChatGPT's recommendations are more based on natural language understanding and matching user intent, without relying on paid advertising slots.

The most direct and biggest beneficiaries are Shopify and the millions of independent website merchants behind it. Shopify is also the most proactive player in integrating various AI dialogue portals, participating in the construction of various protocols, and developing the Shopify Catalog to structure the products and information of millions of merchants. It has even opened the Shopify Catalog to brands that have not used Shopify to build their websites.

In multiple quarterly earnings calls, Shopify has mentioned the contribution of AI Shopping to independent website merchants: since January 2025, through the third quarter of 2025, traffic from AI recommendations in merchant traffic channels has surged sevenfold; and by the end of the year, orders attributed to AI recommendations have skyrocketed 15fold.

Of course, Shopify independent websites aren't the only ones getting a piece of the pie. According to Similarweb, by August 2025, about one-fifth of Walmart's e-commerce platform's referral traffic will come from ChatGPT. Other data shows that about 20% of ChatGPT's referral traffic is directed to Etsy, 15% to Target, and 10% to eBay.

Modern Retail states that approximately 36% of Walmart's referral traffic currently comes from AI recommendations. "Although referral clicks account for less than 5% of total website visits, if more shopping behavior occurs through AI entry points like ChatGPT in the future, it could have a significant impact on brands and retailers."

The forms and logic of advertising and business operations are also changing.

Google says it will allow merchants to offer discounts directly in AI-generated shopping suggestions, with AI determining when to display them to drive consumer purchases. Future plans include bundled sales and free shipping to further attract consumers to place orders directly through AI conversations.

Meanwhile, Google's Business Agent acts like a pre-sales consultant, accessing brand knowledge and product information to directly answer consumer questions from the brand's perspective, making the conversation interactive. Reportedly, brands such as Lowe's, Reebok, Michael's, and Poshmark have already integrated with Business Agent. In the coming months, they will be able to train their Business Agents based on their own data, gain relevant user insights, and offer related product discounts.

The AI-powered dialogue entry point has indeed brought tangible conversions to merchants. A cross-border marketing service provider stated that the conversions brought by AI assistants account for an average of 0.5%-1.2% of Shopify merchants' GMV, which is still in a relatively early stage; however, in the future, if merchants can allocate more budget to the GEO field, the GMV contributed by AI channels may reach 3%-8%.

Service providers are also seeing a resurgence thanks to AI shopping. For example, Stripe has become a crucial part of the payment process across various protocols and AI gateways; Worldpay, PayPal, and checkout.com have also announced that they will adopt Stripe and OpenAI's ACP standard; and KLAR, a payment company in the BNPL (Buy Now Pay Later) field, has joined Google's new Universal Business Protocol.

Even some payment service providers have chosen to join the market education effort. During a post-earnings conference call last year, PayPal executives stated frankly that AI shopping is still in its early stages and that PayPal needs to invest in marketing to educate the market.

In addition, a number of GEO service providers have sprung up like mushrooms after rain.

03 In the era of AI e-commerce, who can remain unscathed?

Last week, Amazon released its latest quarterly earnings report, stating that its North American retail revenue growth was 10%, a 1.3% slowdown compared to the previous quarter. Market research firm Dolphin Research analyzed that, considering the overall retail business in the US market has not seen significant changes, competition from AI-powered e-commerce appears to have already impacted Amazon's business to some extent.

However, Amazon's public statements do not seem to indicate that it is worried about this.

“I believe that ultimately many customers will choose to use the superior shopping agents offered by retailers. Because what consumers really value in retail shopping is a wide selection of products, low prices, fast delivery, and trustworthy service,” Amazon CEO Andy Jassy pointed out. “Third-party agents may be good at aggregating product selections, but retailers can provide a better experience in all four areas.”

"We need to consider how many consumers would prefer to use an intermediary between the retailer and the consumer, rather than a superior intermediary provided by the retailer that has their entire shopping history and accurate data, facilitating precise shopping?" he said.

As can be seen, whether in media interviews or in the earnings call after the report, Andy Jassy's attitude has been consistent: he is "very optimistic" that users will use Amazon's AI assistant more.

This confidence stems partly from the strong performance of Amazon's AI shopping assistant, Rufus, over the past year. Another reason is that completing purchases through AI dialogue is still in its very early stages, lacking real user data and fulfillment capabilities, and the overall user experience needs improvement.

Some market analysts have pointed out that placing an order on ChatGPT is not entirely the same as buying products directly from an independent website. For example, ChatGPT recommends products in search results only if the product details (such as pricing and inventory) are synchronized with the content on the independent website. However, there are not many merchants who can fully standardize information and comply with AI standards, leading to some oversights.

An OpenAI spokesperson also stated that they are researching how best to structure and standardize product data to ensure the accuracy of information, and are making improvements based on feedback from merchants and payment companies.

Even Amazon's own Rufus has been criticized by some retailers. In their view, Amazon's use of AI to obtain product listings on its website without the retailers' consent could lead to errors, such as displaying expired product information to consumers.

In an interview with The Information, Andy Jassy stated frankly that many AI-driven shopping tools offer a "not very good experience yet, and may obtain incorrect product information or pricing," but he believes that the situation will gradually improve over time.

It's worth mentioning that for Amazon itself, "closed" isn't absolute. Andy Jassy has mentioned on more than one occasion that he hopes to establish partnerships with third-party AI agents, but "it requires finding a way to exchange customer experiences and value that is meaningful to both parties."

“I believe that in the future, most retailers will have their own AI agents, and mature third-party AI agents will also emerge. But the people who shop on Amazon will still largely start with Amazon,” Andy Jassy said. “Part of the reason is the wide variety of products, low prices, and fast delivery. Another key point is that consumers are not very tolerant of you showing them ads that are not related to their search results, which is exactly the area that Amazon’s advertising team has been focusing on in the past.”

Ultimately, Amazon's current secret to dealing with the changing traffic sources in the AI e-commerce era is: to focus on user experience and maintain it consistently.

But this doesn't only apply to Amazon. In yesterday's conference call, Shopify expressed unprecedented confidence in AI shopping: "Under the new AI shopping system, the large model will not bypass Shopify for checkout, and complex processes such as logistics, payment, and analytics will always revolve around Shopify." Shopify's confidence stems from its access to billions of product data points and its structured capabilities.

"In 2026, we will enter a stage of large-scale application of AI shopping," said a Shopify executive.