IAB Debate: Will AI Kill the $72 Billion Retail Media Business? Industry Splits — 'I'll Gladly Have an Agent Buy My Dental Floss. Not My Jacket.'

"I don't care what dental floss I use. I'll gladly have an agent execute that on me." Scott Collins, senior director of product strategy at Moloco Commerce Media, was drawing a line that divided the entire room at IAB Connected Commerce this week. One side of the line: low-consideration products where AI agents could automate repurchasing. The other side: everything consumers actually want to choose for themselves. The session, chaired by IAB VP of AI Caroline Giegerich, was framed as "Will AI kill retail media?" The actual argument was more useful than that headline suggested.

The real debate

The Drum's analysis of the panel captured the framing precisely: "This is not really a debate between apocalypse and nothingburger. It is a debate about where the margin shifts start to happen and who captures the value when they do." Collins set the stakes. If AI agents remove indecision from the purchase path, some of the clicks, impressions, and ad placements retailers currently monetize will shrink. "It doesn't take a lot of agentic commerce to start changing the economics of retail media business," he told the room. That single line was the most important of the session.

The math behind the alarm is straightforward. US retail media will reach 71.98 billion dollars in 2026, according to eMarketer's March forecast. That revenue depends on consumers searching within retailer platforms like Amazon, Walmart.com, and Target Circle, where sponsored placements sit next to organic results. AI agents that handle product search externally — inside ChatGPT, Perplexity, or Gemini — reduce the value of those placements because the consumer never sees them.

The category divide

Collins offered a matrix that reframed the argument. Low-consideration, low-preference products like dental floss, paper towels, and printer ink are candidates for full agent automation. The consumer does not care about brand. Replenishment is a solved use case waiting for better technology. High-consideration products — a jacket, a laptop, a vacation — are different. The consumer wants to browse, compare, feel, try on. AI can help discover and narrow the options. The final decision still needs human involvement.

Andrew Lipsman pushed back. The industry, he pointed out, "has already had plenty of shots at automating replenishment." Amazon Subscribe and Save, DTC subscription boxes, grocery ecommerce auto-delivery — none of them have reached mass adoption. If consumers did not want automated dental floss from those tools, why would they want it from an AI agent? "This is not a burning consumer problem waiting for AI to solve it," he argued. Collins called the comparison a "false equivalency." Subscription tools are passive schedules, he said. Agentic AI is proactive intelligence that adjusts to context, price changes, and preference drift.

Whose data wins

David Verklin raised the structural question. Retailer AI tools like Amazon's Rufus and Walmart's Sparky have something general-purpose AI lacks — "history, behavioral context and actual purchase data." "One thing about agents is they're only as good as the data you feed them," Verklin said. Ask Rufus for allergy medication and the recommendation is shaped by what you have actually bought before. Ask ChatGPT the same question and the model has to infer your needs from scratch.

That is the retailer's structural advantage. They already know what you buy. But the advantage only activates if consumers start their shopping journey on the retailer's surface. If the first touchpoint moves to ChatGPT, the retailer never gets to use the data it spent a decade collecting. Adobe Analytics reported on April 16 that AI-driven traffic to US retail sites grew 393 percent in Q1 2026 and now converts 42 percent better than paid search. The first touchpoint is already moving. The debate at IAB was whether retailers can hold the second one.

Mubboo's Take

The dental floss versus jacket distinction is the most useful framework published this year for understanding where AI commerce will actually work. Low-consideration replenishment products may eventually shift to fully automated AI purchasing. High-consideration products — where the consumer cares about features, quality, value, and how something fits into their life — will always require comparison and judgment. Mubboo exists in the high-consideration space. Our shopping coverage at mubboo.com and AU buying guides help consumers decide which robot vacuum to buy, which flight route to take, which hotel to book for a family reunion. These are not dental floss decisions. They are jacket decisions. And jacket decisions need editorial content that AI agents cite, not AI agents that buy on your behalf.