AI shopping spurts up a new $8 billion circuit: retailers compete for chatbot recommendations, data sovereignty becomes the focus of the game

The Zhitong Finance App learned that as consumers increasingly turn to ChatGPT and Google Gemini for shopping suggestions, a dark retail battle over traffic entry and customer data is breaking out in full swing. Retailers such as Walmart (WMT.US), Ulta Beauty (ULTA.US), and Wayfair (W.US) are optimizing their websites in an effort to rank high in AI chatbot search results. According to Juniper Research, retail spending guided by AI agents such as Anthropic's Claude and Gemini is expected to reach 8 billion US dollars this year. Meanwhile, according to Adobe Analytics data, 41% of American consumers already used generative AI for online shopping in June, and visitors attracted by AI earned 41% more revenue per order than traditional channels.

AI recommendation traffic has exploded, and conversion rate advantages have been highlighted

AI shopping assistants are reshaping e-commerce traffic structures at an unprecedented speed. According to the smart commerce report released by Bain in May 2026, the number of shopping recommendations by ChatGPT in the US, the UK, Germany, and France has more than doubled year-on-year, and some retailers have reported that AI has now contributed 25% of recommendation traffic. The Bain survey also found that 30% to 45% of American consumers have used AI tools in the shopping process, and 64% of consumers say they have used or are willing to try using AI to complete shopping.

Adobe's data is even more astonishing: in the first quarter of 2026, AI-driven online shopping traffic surged 393% year over year, and this trend continued after the holiday season. The conversion rate for AI referral traffic was 42% higher than non-AI traffic, and this figure was still “38% lower” in March 2025 — an impressive reversal within a year. Shopify's data also confirms this trend: in the first quarter of 2026, the number of orders recommended by AI on the platform increased nearly 13 times year over year, the conversion rate was nearly 50% higher, and the average order value was 14% higher.

Josh Friedman, head of digital and e-commerce at Ulta Beauty, said that consumers who discovered its products through Gemini and ChatGPT had doubled their conversion rate and willingness to buy it. The company is partnering with Google to integrate the shopping cart and Ulta Beauty Rewards member rewards program into Gemini's AI shopping platform.

“Game of Thrones” for AI Traffic Entry: Open Network vs. Walled Garden

Tech giants and e-commerce platforms are showing very different strategic orientations around AI shopping. One party, represented by OpenAI and Google, chose full openness and cross-platform and cross-channel access to product information and shopping scenarios through natural language interaction and intelligent recommendations. Up to now, about 20% of ChatGPT's recommended traffic is directed to Etsy, 15% to Target, and 10% to eBay.

Traditional platforms such as Amazon and eBay, on the other hand, adopt a relatively closed strategy. Amazon used technical means to restrict ChatGPT and Gemini from scraping product data from its platform, while also speeding up the construction of its own AI shopping assistant Rufus. As of the beginning of 2026, Rufus has been used by more than 300 million customers, and the purchase conversion rate of customers using this assistant is about 60% higher. The sales increase brought by Rufus support in the past year is close to 12 billion US dollars. According to Sensor Tower data, heavy Amazon users now use Rufus in around 60% of their shopping sessions, and shoppers using this assistant are 2.74 times more likely to make a purchase.

Vince Koh, head of global digital commerce at Amazon AWS, made it clear: “When a shopper completes a purchase on the brand's own website... the retailer maintains a direct relationship with that customer”. AWS is encouraging retail customers to use AI platforms to promote products while ensuring that its own website remains the “best place to buy.”

Retailers' Counterattack: Own AI Assistants and Data Moats

Facing the traffic siphon of AI platforms, retailers are building lines of defense from two directions.

First, create your own AI shopping assistant. Walmart has launched Sparky, a generative AI-driven shopping assistant, which has been integrated into the Walmart App. The average customer price for Sparky users is around 35% higher, and weekly active users have more than doubled in the last quarter. Walmart has also adopted an “embedded” strategy — sending Sparky to the ChatGPT and Google Gemini platforms to directly serve consumers in AI chats.

Second, cultivate the data moat. Retailers generally believe they know their products and customers better than generic AI tools, which creates a unique competitive advantage. Raina Moskowitz, CEO of the wedding planning platform The Knot, said, “LLM is another starting point, another search point, and another planning point, but we have 30 years of data to help” couples coordinate their weddings. AWS's Koh notes that when shoppers search for specific products on the Kate Spade website, retailers can use information such as browsing history, budget, and style preferences to make personalized recommendations and build loyalty.

Instant Checkout Dreams Shattered: OpenAI's Strategic Turn and Retailers' Temporary Victory

The most iconic event occurred in March 2026. OpenAI has officially terminated ChatGPT's “Instant Checkout” (Instant Checkout) function, no longer supports users to place orders and pay directly within AI conversations, but instead guides users to third party e-commerce platforms to complete transactions.

One immediate reason for this decision is poor conversion rate data. Walmart executives once revealed that with the instant checkout function embedded in ChatGPT, the actual conversion rate was only one-third of the transaction transferred to the official website, a drop of 66%. An OpenAI spokesperson described this adjustment as a strategic shift: “Instant checkout is moving to apps where purchases can be made more seamlessly.”

Etsy's practice validates this logic. As one of ChatGPT's first integrated partners for instant checkout, Etsy found that ChatGPT excels at product discovery. Rafe Colburn, Etsy's chief product and technology officer, said that users who found products through ChatGPT usually return to the Etsy website to complete their purchase, and stated that “this marks the beginning of a deeper relationship, not just one that is always mediated by ChatGPT.” In May of this year, Etsy launched a native app within ChatGPT, further strengthening its position as a “discovery layer” rather than a “checkout destination.”

Retailers' “temporary advantage”: data sovereignty remains a moat

In the wave of AI shopping assistants, retailers still maintain a key advantage: consumers are still more willing to complete purchases on retailers' own websites. Although the traffic brought by AI platforms is growing rapidly, the end of the conversion funnel — the checkout process — is still firmly in the hands of retailers. As a result, loyalty programs are becoming the strongest moat for retailers in this data battle.

For example, Ulta Beauty has over 47 million loyalty members, and 95% of sales are completed through loyalty programs. The company is partnering with Google to integrate the shopping cart and Ulta Beauty Rewards loyalty program into Gemini's AI shopping experience. At the same time, Ulta launched its own AI shopping assistant “Ulta AI” built on Gemini Enterprise.

The underlying logic of this strategy is that even if shoppers discover products through AI platforms, loyalty points and personalized experiences will drive them back to the retailer's own channels to complete the transaction. As one industry observer said, “Loyalty data has become the new currency in the AI era — whoever has the most first-party shopping data can train the most accurate recommendation algorithms.”

As Ulta Beauty's Friedman said, “Whether it's Google search, affiliate marketing, or Facebook, interacting with customers on someone else's platform always comes at a cost. I don't think it's any different”. Retailers are willing to “pay taxes” in exchange for traffic on AI platforms, but refuse to hand over control of customer relationships.

However, the game is far from over. Bain & Company predicts that by 2030, artificial intelligence could affect 25% of e-commerce businesses in the US. According to the Nielsen IQ report, in e-commerce agents with artificial intelligence, AI agents are becoming actual decision makers, making evaluations on behalf of consumers — and often selecting between multiple retailers at the same time. As AI agents evolve from “recommenders” to “decision makers,” retailers' brand premiums and customer loyalty will be tested as never before.