Retail media works because people scroll. A brand pays Amazon or Walmart, its product shows up first, a person sees it and clicks. A Reuters analysis published on November 4 asks the obvious question: what is that slot worth when the one scrolling is an AI agent?
It matters beyond Amazon and Walmart. If agents pick products from data instead of from paid placement, the thing a seller has to win is no longer an auction run by the retailer but the agent's own reading, request by request, of which product fits. That is a different game.
The size of what is at stake
Emarketer puts US retail media spending at $74.4 billion this year, and Walmart and Amazon hold a combined 88% of it. Brands are expected to spend 22.6% more on it in 2026 than the $60.7 billion they spent last year.
This is not a side business. Amazon's ads run at about 10% of its net sales. Walmart's CFO John David Rainey said in August that advertising, membership and marketplace together made up 45% of the quarter's operating income growth.
Now set that against Morgan Stanley's estimate from July: AI agents could handle 10% to 20% of US e-commerce spending by 2030. Every one of those purchases skips the page where the ads live. All of them.
Amazon closes the door, Walmart opens it
The two companies answered in opposite ways. Amazon banned Meta's Muse agent in September, saying it entered the store without permission and captured customer credentials. This week it told Reuters that outside agents should work openly and respect a retailer's choice about whether to take part.
Amazon also won a court order in March that blocked Perplexity's Comet agent from its store. The 9th Circuit overturned that ban in August, and the case is still going.
Walmart went the other way. On October 21 it rolled out infrastructure for third-party agents: each agent gets a profile tied to the shopper's Walmart account, and the shopper can approve purchases or set spending limits. It partners with Meta's tool and has started testing ads inside Sparky, its own chatbot.
Both moves protect one thing. Amazon keeps agents out so the ad page stays the front door. Walmart lets agents in on its terms, then puts ads inside the assistant it controls.
What an agent reads instead of a banner
The most useful line in the piece is not about either retailer. Advertising for bots, Reuters writes, may depend less on visual or emotional appeal and more on structured product data that matches what the chatbot was asked.
So P&G, Nestle and L'Oreal are already reworking their product information, field by field, so that an agent can parse it without guessing what a phrase on a label was meant to say. A banner tells a person a mood. A product record tells an agent a size, a price, an ingredient, a return window. The agent compares records. The record is the pitch.
Kearney's Katherine Black sketches where this could go: agents searching at a shopper's price point, with retailers setting price floors and ceilings, and retail media turning into a bidding system run between machines.
Why the threat is slower than it sounds
Not everyone is alarmed. Andy Jassy said last week that third-party agents have had no meaningful effect on Amazon's ads yet. Morningstar's Brett Husslein calls the shift an evolution, not an existential threat. Mondelez said in April that retailer sites still drive most of its volume.
Amazon's ad revenue did grow only 11% in the third quarter, its slowest pace in nearly four years, by LSEG data. But nobody in the piece ties that slowdown to agents, and neither should we. The honest reading is that the money has not moved. What has moved is the question every big advertiser is now asking its agency.
The bigger picture
For a consumer brand the fight is over a shelf. For a B2B company it already happened, with no ad market in front of it. When a buyer asks ChatGPT or Claude which tool fits their team, there is no sponsored slot to buy. The answer names vendors from what the model found and trusts: comparison pages, reviews, docs, threads.
P&G is rewriting product data because an agent reads data, not banners. A startup's version of that work is making sure the pages ChatGPT and Claude read describe what you do, for whom, and against whom, in the words a buyer types. Money cannot shortcut it. There is no slot to pay for.
ClawWorld runs AI visibility work for B2B startups. Start with a free baseline report: 10 buyer questions, who ChatGPT and Claude name instead of you, and the pages they cite.