The Race to Own AI Shopping

Ecommerce companies may soon compete not only for shoppers, but also to “own” the AI through which those shoppers buy.

Conceptually, the idea is not new. Competing for customers has long included competing for attention and proxies. Take search engine optimization. Ecommerce marketers work hard to rank in search engine result pages to win a shopper. The search engine is a proxy.

Generative engine optimization seeks similar visibility, but another layer of AI shopping is emerging.

Commerce Agents

On September 2, Anthropic released “Building Commerce Agents with Claude,” a collection of software patterns, “guardrails,” and implementations to help companies build AI agents for retail and other commerce applications.

The blueprint includes two examples.

Shopping agent

First, there is a consumer-facing shopping agent. This agent resides on a store’s website or app and connects to the product data, checkout software, and other services.

Claude poses an example. A shopper could tell the agent she needs a tent, sleeping bag, and stove for a weekend trip with two children. The agent would search the store’s product catalog, select compatible items, compare those products, and add them to the shopping cart, all while considering the customer’s preferences and previous orders.

After the sale, the same shopping agent could answer questions about delivery, returns, exchanges, and refunds.

Claude claims retailers (unnamed) running its shopping agents have experienced a 35% increase in order value and a 60% improvement in conversion rate.

The claim is interesting given that many enterprise retailers have deployed similar agents.

Merchant agent

Anthropic’s second example agent works behind the scenes with a store’s internal systems.

An ecommerce manager could ask the agent which products to discount to clear old inventory and free up cash. The merchant agent might analyze inventory levels, track sales velocity, recommend price changes, and even draft marketing campaigns in response.

The blueprint’s guardrails would require human approval before the merchant agent could implement significant changes.

Similar merchant agents already exist in some ecommerce platforms, but the Anthropic blueprint should make it relatively easier to build custom implementations.

Shoppers Want AI

Anthropic’s announcement comes as more consumers appear willing to involve AI in shopping during the 2026 Christmas season.

Bain & Company, an Anthropic partner, commissioned an August survey of 1,105 U.S. shoppers for its holiday 2026 forecast. Twenty-four percent of online buyers plan to start their holiday shopping on AI platforms such as Claude, Google Gemini, and ChatGPT, up from 17% in 2025.

Roughly 60% of respondents plan to start on retail or brand websites, up from 51% last year.

The two findings are not necessarily contradictory.

Some shoppers may ask ChatGPT or Gemini what to buy before visiting a store. Others may go directly to a merchant’s site and use its AI agent instead of search or navigation menus.

Still others might use both an external and on-site agent during the same purchase.

2 Models

Bain’s survey describes two emerging models for AI commerce.

In the first, an external platform owns the AI relationship or at least the shopping interface. ChatGPT, Perplexity, or another assistant finds products from multiple sellers and helps complete the transaction.

Shopper → AI platform → Merchant

OpenAI, for example, introduced Instant Checkout with Stripe and the Agentic Commerce Protocol in 2025, allowing supported purchases without leaving ChatGPT.

Google has taken a similar approach with its Universal Commerce Protocol, developed with Shopify, Etsy, Wayfair, Target, Walmart, and others. The protocol connects AI experiences with existing merchant and payment infrastructure.

The second agentic shopping model puts the merchant closer to the center.

Shopper → Merchant’s AI → Merchant

Anthropic’s blueprint helps make this approach practical. A retailer can offer conversational product discovery while retaining its own catalog, checkout, customer relationship, and brand experience.

Ecommerce may be entering a race to own the AI through which customers shop.

Customer Relationships

Piece by piece and innovation by innovation, the industry is assembling an agentic AI commerce stack. As these two AI shopping models emerge, direct customer relationships will likely become increasingly important.

Merchants that know their customers, retain first-party data, and maintain direct channels such as email, loyalty programs, and on-site experiences should be better positioned — whether shoppers arrive through an outside AI platform or use the merchant’s own agent.