Skip to main content

Sitation Blog

The Fight Over Who Gets to Shop for Your Customer

November 12, 2026

The Fight Over Who Gets to Shop for Your Customer

Battle lines are forming over AI shopping agents, and every online seller has to decide how far to let them in.

It’s been a fascinating week in agentic commerce, but it’s been easy to miss amid the coverage of the Jensen Huang scandal (side note – yikes). A consequential split is taking shape over whether businesses will let a customer’s AI agent through the front door, and how far inside it can go.

The Verge reports that eBay appears to be blocking agents such as Meta’s Muse from browsing listings. United Airlines is restricting third-party agents from booking flights, citing account security and unintended transactions.

At the other end, Walmart announced Agent Profiles and an Agent Gateway in October, with customer permissions and spending limits. Walmart says agents will see the same prices as people. Its approach is controlled access, with rules about what an agent can do.

These are different boundaries around discovery and checkout, with very different risks and potential benefits. Together, they raise a critical commercial question…

How much of the shopping experience will businesses allow customers to delegate?

In some ways, agents are the new Instacart. You give someone a shopping list and preferences, and they do their best to fill the order. An AI agent can take that model beyond groceries to a flight or a car. It can also misunderstand the assignment, which makes permissions and approval important.

At our trade event in New York City last week, a participant shared that she is in the market for a new car, and had her Muse (the Meta AI personal agent) arrange test drives at local dealerships. If a dealership didn’t respond to her agent’s messages, she dropped the brand from consideration. In her words, “I’m not going to inconvenience myself.”

Is that crazy? Maybe! But it’s sticking with me. Our personal assistant and shopping agents are becoming the front-line filter for our consideration set. Dealerships lost the opportunity before a salesperson ever spoke to her.

What this means for brands

Agentic buying is coming fast, whether you like it or not. Here are a few things I’m thinking about:

Brand equity has to survive delegation. “Buy my usual coffee” carries a brand preference into the task. “Find good coffee under $15” leaves more room for substitution. Loyalty still matters, but brands need to understand whether it shows up in the customer’s instructions and whether their product information gives the agent a reason to preserve it.

Impulse buys may move upstream. An agent sent to replenish essentials may never encounter the checkout display or browse the recommended products. Some spontaneous purchases could disappear; others could originate in the assistant’s recommendations. Retailers should measure what happens to basket size when the shopper delegates – if it’s not on the list, it probably doesn’t make the cut.

Price elasticity could become more visible. When comparing equivalent products takes less effort, small price differences may shift more orders. But an agent instructed to prioritize a trusted brand or delivery tomorrow could preserve a premium. The task the customer sets matters as much as the price the retailer sets.

Who owns the payment flow?

At the same time, there’s also a battle over who owns AI payments. Visa and Mastercard are building services for agent-initiated payments. Stripe and OpenAI developed the Agentic Commerce Protocol to connect agents and merchant checkout. These systems can work together, but the commercial stakes remain: who holds the payment relationship, earns transaction revenue, and sets the terms of access? Winning the recommendation doesn’t automatically mean controlling the purchase.

I understand the caution. Businesses need to prevent unauthorized purchases and protect customer accounts. But blocking an agent can also mean turning away a customer who has deliberately chosen to shop that way.

It’s a growing trend that can no longer be ignored.

Next steps

Our current advice to commerce teams is to test that experience yourselves. Give an agent a realistic buying task involving your products. Watch where it drops your brand, loses access, or fails to complete the order. Then decide which barriers protect the customer and which simply cost you the sale.

OpenAI now allows you to register your business and start measuring mentions. Definitely do that! And consider getting your shipping policies and at least a basic product catalog up and running.

The dealerships that never replied are a useful warning. Sometimes the first competitive hurdle is being willing and able to answer the customer’s representative.

There’s a lot of traffic and potential customers out there, both human and bot. Don’t miss out as things change!

Put AI to work on your product data

We help teams choose useful AI use cases, run a pilot with people in control, and scale what works across their PIM and product content.

AI digital transformation

Book a free AI readiness review