Williams-Sonoma bars AI shopping agents without its written permission

A shopper’s approval is no longer enough to let an AI assistant buy from Williams-Sonoma’s websites. Williams Sonoma and Pottery Barn emailed customers this week about new restrictions, KQED reported October 13. Across the parent company’s brands, agents need written company permission to use accounts or make purchases on customers’ behalf.

An earlier ruling left room for website rules

The restrictions follow an August federal appeals court ruling in Amazon’s lawsuit against Perplexity. In that case, the court held that the user, rather than the AI company, was accessing the website when an agent shopped for them. But it also said companies could still set access rules through their terms of service.

KQED described that decision as weakening retailers’ ability to use anti-hacking law against shopping agents, while leaving contractual restrictions available. Williams-Sonoma’s updated terms use that latter route: they set conditions on automated access, including access a customer has approved.

Customer permission does not override the retailer

The new language covers an agent, assistant, bot or AI-enabled software—not just a particular shopping service. Customers cannot let those tools use their accounts or purchase for them without written permission from Williams-Sonoma. The policy applies across all brands under Williams-Sonoma Inc., rather than only the two brands that sent the customer emails.

• Automated tools interacting with the websites must identify themselves as automated.
• Williams-Sonoma reserves the right to block third parties, including those a customer has purported to authorize.
• Customers are responsible for all activity on their accounts, whether or not they authorized that activity.

The restriction is not a rejection of all AI on the retailer’s website. Its own chatbot introduced itself to KQED’s reporter as “Olive, your AI Sous Chef.” The company did not respond to KQED’s request for comment before publication, leaving its rationale for restricting customer agents unexplained.

in general, we should be encouraging consumers to find the tools that work best for them and to let them use them

Eric Goldman, Santa Clara University School of Law professor, speaking to KQED

Who should get to choose the shopping assistant?

Loyola Marymount University law professor Lauren Willis told KQED the new terms are likely legal for now. She wants a different legal standard: retailers should have to accept shoppers’ AI agents, and those agents should be required to act in consumers’ best interests rather than their providers’ interests.

Willis pointed to European Law Institute proposals that would require businesses to permit consumers to use digital shopping assistants in transactions. Her position pairs access with a duty to the shopper; it does not treat an agent’s involvement alone as a guarantee that the customer benefits.

Goldman also cautioned that AI agents can serve consumers, retailers or only the companies that make them, and that their loyalties are not always clear. Still, he questioned the business case for blocking them. Without an explanation from a website, he said, a block can look like an overreaction to AI.

The next challenge may be distinguishing an AI agent from the browser a shopper normally uses. Goldman predicted that many capabilities now associated with agents will become standard browser features. In his view, retailers’ ability to single out those tools will erode as the separation between agent software and ordinary browsing disappears. That is a forecast, not a change to Williams-Sonoma’s current rules.