AI shopping agents reshape retail buying & checkout

Retailers and technology groups are rolling out AI shopping agents that can search for products and complete purchases on a customer's behalf. The shift is gaining support from new industry protocols, including OpenAI's Agentic Commerce Protocol and Google's Universal Commerce Protocol.

Large retailers are testing services that shift online shopping away from a sequence of clicks and toward a chat-based exchange, with software acting on the user's behalf. Often called agentic commerce, the model allows an assistant to search catalogues, compare options, add items to a basket and pay, while staying within limits set by the customer.

In practice, a shopper could ask an assistant to buy dinner ingredients or source household items within a set budget, and the software would complete the transaction with little further input. The customer still sets preferences for spending, brands, and delivery, but the agent handles the work that would usually occur across multiple web pages or apps.

The technology behind this process relies on product data feeds, merchant systems and payment interfaces. At checkout, agents can use encrypted credentials and one-time payment tokens. At the same time, the retailer continues to process the order, manage fraud checks, and handle refunds or customer support as usual.

Competing standards

Two emerging standards are central to making AI-led shopping work across platforms. OpenAI's Agentic Commerce Protocol, developed with Stripe, is an open standard that lets AI assistants access product catalogues, prices, and inventory, then initiate checkout for the user.

Google's Universal Commerce Protocol serves a similar purpose but is more closely tied to Google's own services. Developed with partners including Shopify and Wayfair, it allows AI tools to browse retailers during Google search journeys and complete purchases inside Google interfaces using Google Pay.

Both approaches aim to remove friction from eCommerce, but they differ in emphasis. ACP is presented as platform-neutral and aimed at merchants, while UCP remains rooted in Google's consumer touchpoints, even as it is described as compatible with other standards.

Retail tests

Walmart is one of the more visible retail adopters. It introduced Sparky, a generative AI shopping assistant inside its mobile app, to answer product questions, summarise customer reviews and recommend items.

Walmart has also moved into conversational buying through a partnership with OpenAI that allows shoppers to buy goods through ChatGPT using Instant Checkout. This brings payment into the same conversation in which the customer makes the request.

Amazon has taken a slightly different route with its Buy for Me beta. When a product is unavailable through Amazon, the feature allows its AI system to place an order on another retailer's website using the customer's stored address and payment details.

Google is also expanding retail-facing AI services. Its Business Agent tool lets consumers chat with stores through Google Search, effectively placing a digital sales assistant within a retailer's search listing and offering direct purchase options in the same interface.

Retailers, including Lowe's and Macy's, have enabled these agents to answer queries in a brand-specific style and guide users to checkout. Start-ups are also entering the market. Perplexity, for example, offers shopping tools that query retailers' product interfaces, surface offers, and support instant checkout within a chat environment.

Data pressure

The rise of shopping agents is changing what matters in digital merchandising. If software is making the buying decision, product listings need to be legible to machines as well as attractive to people.

That puts new weight on structured product data, detailed attributes and reviews. Research cited by industry participants suggests that incomplete product information can sharply reduce the likelihood that an AI agent will select an item. At the same time, strong review histories can improve ranking even when prices are not the lowest available.

For merchants, this could shift competitive advantage. Instead of focusing only on web design, paid placements, or conversion tweaks, retailers may need to pay closer attention to how well automated buyers interpret their catalogues.

The change also extends to payments. Visa and Mastercard are developing interfaces that would allow approved AI agents to spend within set customer budgets, while fintech groups are building tools to make agent-led payments secure and routine.

Together, payment infrastructure, open protocols and retailer adoption are pushing the concept beyond the experimental stage. Forecasts from industry analysts point to rapid growth in the use of AI shopping assistants over the next few years, suggesting the sector is preparing for a market in which software increasingly decides what to buy and where to buy it.

As that transition gathers pace, retailers face a new test: not simply winning a shopper's click, but making their products visible, credible and easy for algorithms to choose.