AI shopping in ChatGPT and Gemini: Retailers and brands caught between reach and loss of control

Whether sneakers, cosmetics, or coffee, a growing number of retailers, initially concentrated in the United States, are bringing product search, shopping assistance, and, in some cases, payment directly into AI platforms such as ChatGPT, Gemini, and Microsoft Copilot. This creates a new sales channel, but one that runs through an interface the retailer does not own.

Until now, many retailers have tried to keep customers within their own ecosystems through proprietary apps, websites, and digital assistants. Recent examples point in a different direction: Retailers are integrating their assortments and commerce systems into the AI offerings of major technology companies. Within these environments, they become product and fulfillment providers, while the platform increasingly shapes the actual shopping experience.

JD Sports, Nike, and Gap Inc. are taking this approach the furthest. They plan to enable product discovery and checkout directly within third-party AI environments. JD Sports is preparing to offer purchases, including payment, through Copilot, followed later by Gemini and ChatGPT. Nike and Gap are adopting Google’s Universal Commerce Protocol, which is designed to let shoppers discover and purchase products in Gemini and AI Mode in Google Search.

The companies describe these integrations to reach customers where purchase decisions increasingly take place. Nike even emphasizes that it will continue to control the customer relationship. But that claim is open to question: When Google or Microsoft provides the shopping assistance, product comparisons, cart, and payment interface, it controls significant parts of the customer journey, even if the retailer continues to handle the order and fulfillment.

Sephora and Starbucks are initially testing a less extensive version of this model. The Sephora app in ChatGPT provides personalized product recommendations, while integrated checkout has been announced for a later stage. With Starbucks, customers can discover and customize beverages and start an order in ChatGPT. To complete checkout, however, they must still switch to the Starbucks app or website.

These examples show that the transition is happening gradually: Product discovery moves into the AI platform first, followed by shopping assistance and, potentially, checkout. At each stage, the platform gains more influence over the customer interface.

Retailers can reach customers in the places where product discovery and purchase decisions are increasingly beginning. Instead of building their own AI assistants, they can integrate product data, prices, inventory, and checkout functions into existing platforms through technical interfaces and commerce protocols. This brings shopping assistance, product selection, and payment closer together.

But the added reach comes with tangible drawbacks:

• Retailers have limited control over how they present their brands and provide guidance within a third-party interface.
• The platform influences which products are visible and which competing offers appear alongside them.
• Retailers may receive less data about how purchase decisions are made.
• Fees, advertising models, and access rules could create new dependencies.

This development potentially goes beyond conventional online marketplaces. An AI agent does not merely display products; it can interpret needs, compare offers, and initiate a transaction.

Companies such as Nike emphasize that they will retain the customer relationship despite integrating with third-party platforms. McKinsey’s “Rewiring Retail in Europe” report challenges that claim. If Google, OpenAI, or Microsoft takes over product discovery, shopping assistance, and checkout, these companies will control significant parts of the customer interface. In the most extreme scenario, the retailer would be left primarily with assortment management, order processing, and logistics. McKinsey therefore warns that retailers could shift from orchestrating the shopping experience to becoming barely visible “fulfillment engines.”

McKinsey does not, however, advise retailers to withdraw from third-party AI environments. Retailers should become “agent ready,” provide machine-readable product data, and secure their visibility in AI-generated answers. At the same time, they must determine which parts of the customer journey they want to control themselves.

A presence in ChatGPT, Gemini, or Copilot can provide an important additional sales channel. But retailers should not confuse reach with control. Whoever recommends products, shapes comparisons, and initiates the purchase also plays a defining role in the customer relationship.

Retailers must therefore decide not only which AI platforms they want to sell through. They must also determine what role their brand will play within those environments and which parts of the customer relationship they are unwilling to hand over to the platform operators.