- Researchers found that AI shopping agents skipped sponsored products across every model they tested.
- Review volume also changed pricing power, with brands able to raise prices by 19% to 37% after doubling reviews and still be selected at the same rate.
AI shopping agents may create a problem for one of ecommerce’s biggest revenue engines: sponsored product placement.
Recent research from Columbia Business School and Yale School of Management tested how Claude Sonnet 4, GPT-4.1 and Gemini 2.5 Flash behave when shopping in a mock Amazon-style marketplace.
The finding marketers should pay attention to is simple: across every tested model, AI agents skipped sponsored products.
They did not rank them slightly lower. They avoided them.
AI agents may have a sponsored product problem
Retail media has grown around a simple idea: brands pay to appear in front of shoppers at the moment of purchase.
That model works when humans are browsing search results, product grids and category pages.
AI agents behave differently.
In the Columbia and Yale study, the agents selected products based on signals such as reviews, ratings, price and product information. Sponsored placement did not help. In the tests, it appeared to work against the products instead.
That matters because retail media depends on visibility being valuable. If AI agents ignore or discount sponsored listings, the value of that placement changes.
Reviews became a pricing advantage
The same research also found that reviews had a measurable effect on agent selection.
When researchers doubled a product’s review count, brands could raise prices and still be selected by AI agents at the same rate.
The pricing advantage varied by model. Claude Sonnet 4 allowed a 19.4% price increase. GPT-4.1 allowed 37.4%. Gemini 2.5 Flash allowed 17.2%.
That turns reviews into more than a trust signal.
In an agent-driven shopping environment, review volume may become a direct commercial advantage because it changes how much a brand can charge while remaining competitive.
Why this matters for retail media
The risk is not theoretical.
McKinsey has estimated that agentic commerce could influence between $3 trillion and $5 trillion in global retail spending by 2030, including up to $1 trillion in the United States.
If even part of that activity moves through AI agents, ecommerce visibility will depend less on what shoppers see and more on what agents trust.
That could pressure brands that rely heavily on paid placement inside retail marketplaces.
It could also reward brands with stronger review profiles, cleaner product data and more consistent third-party information.
The data layer becomes the storefront
For brands, the practical implication is that product information becomes harder to treat as a backend detail.
An AI shopping agent needs clean product names, accurate prices, current availability, clear specifications, review signals and structured data it can parse.
If that information is incomplete, inconsistent or outdated, the product may never make the shortlist.
This is where agentic commerce connects directly to SEO and AI visibility. We recently covered how Google is turning search into an AI agent platform, and ecommerce appears to be one of the clearest areas where that shift could affect real buying decisions.
The open question: do agents browse or use APIs?
One important caveat is that not all AI agents will shop by browsing websites like humans.
Many agents will use APIs, product feeds, marketplace data or direct integrations instead.
That makes the challenge broader than website optimization alone. Brands may need to keep both their public product pages and their structured product feeds accurate, complete and machine-readable.
In other words, the agent may not care how good the landing page looks if the underlying product data is weak.
The takeaway for ecommerce brands
The Columbia and Yale findings point to a simple but uncomfortable possibility.
If AI shopping agents continue to favor reviews, ratings, clean product data and organic trust signals over sponsored placement, ecommerce brands may need to rethink where they invest.
Paid visibility will still matter for human shoppers.
But for AI agents, the stronger advantage may come from the signals that cannot be bought as easily: real reviews, accurate data and product information that machines can trust.
