Blue Yonder: 46% of AI shoppers have returned an item

AI shopping recommendations are driving returns, and the cause is your product data. Blue Yonder's 2026 Consumer Retail Returns Survey found that 24% of U.S. consumers have bought a product after an AI tool recommended it, and 46% of those buyers returned one in the past year. The top reason was "different than described," at 36%. Retail Dive picked it up this week with the line that matters: most product data was written for a search bar, not for a machine suggesting products.

What actually happened

The survey was fielded by a third party in September 2026 across more than 5,000 consumers in the U.S., Australia, France, Germany and the U.K., including 2,000 Americans. It is a vendor survey from a supply chain software company, so treat the numbers as directional. The pattern is still hard to argue with:

• Returns for AI-recommended purchases: 36% "different than described," 23% wrong size or fit.
• Returns overall in the U.S.: 42% wrong size or fit, 20% "different than described." The AI case nearly doubles the "not as described" share.
• 49% of shoppers who returned an AI-recommended item say they are more likely to return those than purchases made directly through a retailer.
• Gen Z: 43% have bought on an AI recommendation, and 56% of them returned one.
• Only 20% of respondents would trust AI to pick their size or fit.
• 82% would shop less or stop entirely at a retailer that limits returns because they return too often. 51% have chosen one retailer over another purely for a better returns policy.

Blue Yonder CEO Duncan Angove called it "not really an AI problem" but "a data problem." An AI assistant, in his words, is only as good as the product information it can find.

Why it matters for your business

If a shopping agent is reading your catalog, it is reading your attributes, not your hero image. A listing that says "cozy oversized hoodie" tells a human something and tells a model nothing. The agent fills the gap with its own guess, the customer gets a different product than the one the agent described, and the return comes back to you with the shipping cost attached.

Here is what we do on client catalogs, in priority order:

1. Measured dimensions on every variant. Chest width, length, sleeve, weight, in numbers. Not "runs small." If you sell furniture, every dimension plus box dimensions.
2. Material and care as structured fields, not prose. Agents parse fields. They skim paragraphs.
3. Variant-accurate images. The photo on the blue variant shows the blue variant. Agents and shoppers both use the image to confirm what the text claims.
4. Schema.org Product markup with size, color, material and a machine-readable returns policy. The 82% figure means your returns policy is a conversion input. Publish it where the agent can read it.
5. Fit feedback loops. Feed your "did it fit?" return reasons back into the listing. That data is the thing Angove says you are missing.

We covered the discovery side in AI shopping agents prefer marketplaces. This is the delivery side. Winning the recommendation and then eating the return is worse than not being recommended.

Sources: Blue Yonder via Business Wire, Retail Dive.