Target updates terms for AI-agent shopping
Google Gemini integration shifts purchases through third-party assistants, liability for hallucinated recommendations lands on the customer
Images
Target has updated its online terms and conditions ahead of an integration that would let Google’s Gemini act as a shopping assistant for customers, according to Business Insider. The change anticipates a near-future where an AI tool can suggest Target products and help complete purchases on a user’s behalf—while the retailer makes explicit that the shopper must still direct and approve what gets bought.
That division of responsibility is the point. Once an AI agent sits between customer and storefront, errors become harder to assign. If the model misreads a request, recommends an unsuitable substitute, or invents product details, the consumer experiences the failure while the retailer can argue it merely fulfilled an approved order. The AI provider can say it only generated suggestions. The result is a familiar pattern from platform economics: the party that captures the data and controls the interface pushes liability outward.
For retailers, the upside is obvious. AI agents promise higher conversion and larger baskets by reducing search friction and inserting recommendations at the moment of intent. But they also pull the customer relationship away from the merchant’s own site and into the agent’s layer, where prompts, preferences, and purchasing history become the valuable asset. If the agent becomes the default “front door,” brands compete not just on price and availability but on how they are represented inside a model’s recommendation logic.
That has knock-on effects for advertising. Sponsored placement already shapes search; an agent that “helps you decide” is an even more powerful choke point. The commercial question becomes whether the AI interface is neutral, auction-driven, or bundled into broader partnerships—while the legal question becomes what disclosures, if any, apply when a recommendation is effectively an ad.
Target’s updated terms read like early scaffolding for that world: a contractual reminder that even when software does the shopping, the risk of a bad purchase still sits with the human who clicked approve.