AI may change how shoppers find and compare products before it changes how they pay.

Online shopping usually starts with a search box. A customer types what they want, scans a list, opens a few product pages, and compares options.

AI assistants change that first step. A shopper can describe a need, add constraints, and ask for recommendations. The assistant helps narrow the choice.

The retailer may still make the sale, but it may no longer be where the customer began.

That shift could affect retail economics before AI handles checkout. It could change which products shoppers see, where retailers spend to attract them, and how much they know about the path to purchase.

Walmart’s approach has changed

In October 2025, Walmart and OpenAI announced plans to let customers shop Walmart through ChatGPT using Instant Checkout.

By March 2026, OpenAI said it was moving away from that initial checkout approach. The company said it would let merchants use their own checkout experiences and focus on product discovery. OpenAI also described a Walmart experience within ChatGPT that would connect shoppers to Walmart’s environment, with account linking, loyalty, and Walmart payments.

The change is worth watching. It suggests that AI shopping does not have to mean handing the entire transaction to an assistant. The assistant can help a customer find and compare products, while the retailer keeps more control of checkout and the customer relationship.

These announcements tell us what the companies are building. They do not show how many people use the experience or how much it affects sales.

Retailers are testing different routes

Walmart and Google announced plans to bring Walmart and Sam’s Club products into Gemini. The proposed experience included account linking, recommendations based on previous purchases, connections to existing carts, and local delivery options.

OpenAI has also said that Target, Sephora, Nordstrom, Lowe’s, Best Buy, The Home Depot, and Wayfair have integrated with its Agentic Commerce Protocol for product discovery. Shopify product data is available in ChatGPT, with purchases continuing on the merchant’s own store.

The details vary, but the underlying change is clear: product discovery is spreading beyond retailer websites and apps. Some shoppers may see products in a general assistant. Others may enter a retailer’s own experience from there.

The traffic question

If customers do more research in an assistant, retailers may see changes in their website traffic. Some shoppers could arrive after they have already narrowed down their choices. Others might make a decision without visiting the retailer’s site at all.

That could affect paid search, affiliate referrals, and the value of a product-page visit. Fewer visits would not automatically mean fewer sales. A retailer might receive fewer, more qualified visitors. Or it might lose visibility to competitors that appear more often in recommendations.

Retailers need to measure what actually happens. They should not assume that AI discovery will either increase sales or destroy existing channels.

The attribution problem

If an assistant helps a shopper compare products, the retailer may not know what influenced the decision. The customer might click through, return later, or buy in a store. Existing analytics may not capture the assistant’s role.

That makes it harder to evaluate marketing spend and customer acquisition costs. Retailers should find out what referral, product-view, and conversion data each platform provides. They should also identify what remains invisible.

Product information matters more

An assistant can only compare products using the information it can access. Incomplete descriptions, outdated stock levels, and unclear delivery or returns details make it harder to recommend a product confidently.

Retailers should treat product data as part of the shopping experience. Accurate attributes, availability, price, delivery estimates, and policies help customers make decisions, whether they are browsing a retailer’s own site or asking an assistant for help.

Retailers also need to explain what makes their products different. If every product is represented with the same generic details, an assistant has little reason to recommend one over another.

What retailers can do now

Retailers do not need to predict when AI will start completing most purchases. They can start with a few practical steps:

  • Keep product descriptions, attributes, prices, stock, delivery, and returns information accurate.
  • Test how products appear in AI shopping experiences.
  • Decide which parts of the journey should lead back to their own channels.
  • Track referrals and conversions where platforms allow it.
  • Separate partnership announcements from customer adoption and measurable business results.

I would treat AI assistants as a developing discovery channel, not simply as a new checkout system. The practical questions are whether a retailer’s products can be found, whether the information is reliable, and whether the retailer can still serve the customer well after the recommendation.

Sources

These are company announcements, so the article distinguishes announced capabilities from independently measured customer or commercial outcomes.