Retail Media Is a $196B Bubble Built on Disintermediation Risk.

Retail media is one of the fastest growing categories in advertising, north of $196 billion globally, and every brand I talk to is being told, correctly, that they need to be spending there. Amazon Ads, Walmart Connect, Instacart, the whole ecosystem of “pay the retailer to get seen inside their own storefront.” It’s a genuinely effective channel right now. I’m not arguing otherwise.

What I am arguing is that the entire model quietly assumes something that’s about to stop being true: that a human is doing the browsing, seeing the sponsored placement, and making the click.

The Assumption Underneath It All

Retail media works because humans still browse shelves. That’s changing.

The entire sponsored placement model is built on visual real estate. Your product shows up higher, bigger, with a “sponsored” tag, in front of a human eyeball scrolling a results page. That works beautifully when a person is the one doing the scrolling and deciding.

AI shopping agents don’t scroll. They don’t get visually influenced by a bigger product image or a strategically placed badge. An agent tasked with “find me the best value protein powder under $40” is going to evaluate based on whatever criteria it’s optimizing for, price, reviews, ingredient match, and it’s going to do it by querying data, not by being nudged by a sponsored placement at the top of a page it isn’t even really “looking at” the way a human does.

The Real Risk

Disintermediation: the agent becomes the shelf

If AI shopping agents become a meaningful share of how purchases get discovered and decided, the retailer’s storefront stops being the decision environment. The agent is. And retail media networks have built almost their entire monetization model around controlling that decision environment through visual placement. If the decision environment moves to a layer they don’t control, and can’t easily sell sponsored placement inside, a huge chunk of that $196 billion is optimizing for an audience that’s about to stop looking.

This is disintermediation in the classic sense, a new layer inserts itself between the business and the customer, and the businesses that don’t control that new layer lose leverage fast. Retail media networks currently have no clear, proven mechanism for “sponsoring” a placement inside an AI agent’s decision logic the way they do inside a visual search results page.

What This Means For Brands Right Now

Don’t panic, but don’t go all-in blind either

• Keep investing in retail media, it’s still working and will be for a while, but treat it as a channel with a shrinking shelf life in its current form, not a permanent pillar.
• Start paying attention to how your products are described, reviewed, and structured in data feeds, not just how they look visually. That’s the layer AI agents actually read from.
• Diversify discovery investment beyond a single retailer’s ecosystem. Being dependent on one platform’s sponsored placement model is riskier now than it was three years ago.
• Watch which retail media networks start building genuine AI-agent-compatible sponsorship models early. That’s the signal for who actually sees this coming versus who’s still selling banner ads with better targeting.

Quick Answers

A few direct questions, answered directly

What is disintermediation risk in retail media?
The risk that AI shopping agents insert themselves as the primary decision-maker between brands and buyers, bypassing the visual, sponsored-placement model that retail media currently relies on.

Is retail media still worth investing in?
Yes, in the near term it remains an effective and fast-growing channel. The concern is structural and longer-term as AI-driven shopping behavior grows.

How do AI shopping agents choose products?
Generally by evaluating structured data such as price, reviews, specifications, and stated criteria, rather than by visual browsing or being influenced by sponsored visual placement.