Why AI Chatbots Favor Walmart — and What It Costs Retailers
AI chatbots recommend Walmart over just about any other store, UBS analysts found. Their tests show the bots steer shoppers toward Walmart and Costco more often than toward Amazon.com. The findings were published on Oct. 5, 2026. MarketWatch
Shoppers are increasingly asking ChatGPT and Google's Gemini what to buy. Retailers want to show up in those answers, but they do not want to hand over customer data to do it. Reuters
Walmart has locked in a spot inside both leading assistants. In October, Walmart partnered with OpenAI for shopping inside ChatGPT, then pursued a similar shopping partnership with Google's Gemini. Business Insider
NatWest and Bank of America have warned that shopping with AI agents could raise the risk of scams, fraud and data-privacy breaches. Reuters
The broader context here is who gets paid for the referral. With search, platforms took a cut through paid placement and auctioned keywords, where retailers bid for terms like running shoes. Chatbots squeeze search, comparison and checkout into one answer. That shortens the path to purchase. Small tweaks to model weighting, prompt wording or partnership deals can swing where sales go. For a merchant, a steady bias toward a rival works like losing the eye-level shelf, except the shelf is hidden and the ranking is controlled elsewhere.
Looking at what this means for retail profit and loss, the trade-off is clear. Being in the answer can cut customer acquisition cost, the cost of winning a new buyer, and lift conversion, the share of browsers who actually buy, on serious shopping queries. Handing over transaction data and customer identity weakens the retailer's own data used for personalization, stock planning, credit decisions and lifetime value estimates, a forecast of what a customer is worth over time. Retailers want the traffic without giving up the relationship. AI providers want detailed catalog, stock and checkout data to complete the sale. That negotiation will set take rates, the fee the platform keeps, data-sharing terms and who keeps the customer after purchase.
In my view, two questions will decide how this reprices risk for investors. First, whether recommendations stay like unpaid editorial or become paid placements with disclosure and auctions. An auction system would create a new ad cost to track, directly comparable to search cost per click, what advertisers pay each time someone clicks. Without paid placement, the bias is harder to defend against. Second, who pays when an agent misreads vague instructions or buys from a fake storefront. Banks are already flagging that risk. Higher disputes, account takeovers or scam losses would land on card issuers, merchants and payment firms before any efficiency savings arrive. Until those rules are set, AI checkout is a sales advantage with an unpaid bill for fraud still attached.
