46% of AI-Advised Shoppers Returned a Purchase, Survey Finds

Nearly half of U.S. shoppers who bought an item on an AI recommendation — 46% — returned at least one such purchase in the past year. That's the headline number from Blue Yonder's 2026 Consumer Retail Returns Survey, released November 2, and it lands squarely on the desks of marketing leaders who have been racing to plug AI assistants into the shopping journey.

The problem, at least according to Blue Yonder's CEO, isn't the AI. It's the data underneath it.

What are shoppers sending back, and why?

The survey, now in its fourth year in the U.S., polled more than 5,000 consumers across the U.S., Australia, France, Germany and the U.K., including 2,000 Americans. The top reasons for returning an AI-driven purchase were blunt ones: the item was different than described, or it was the incorrect size or fit.

So the machine recommended something. The shopper believed it. The product failed to match the promise. Back in the box it goes.

Two supporting figures sharpen the picture for CMOs:

• Just one in five U.S. consumers trusts AI to recommend the right size and fit.
• More than eight in 10 would cut back on shopping at a retailer that limits their returns.

Pair those and you have an uncomfortable squeeze. Shoppers barely trust AI to get sizing right, yet they will punish retailers who tighten return policies to absorb the cost of AI's misses.

Whose problem is this, really?

Blue Yonder, for the record, provides an AI-driven platform and multi-tier network for supply chain strategies. It has a commercial interest in the diagnosis. But the diagnosis is worth reading in full.

"Shoppers are taking AI's advice faster than retailers are getting ready for it, and nearly half of them are sending something back," Duncan Angove, CEO of Blue Yonder, said in the release announcing the survey.

Angove's verdict on the root cause is short. "That's not really an AI problem. It's a data problem."

He elaborates: "An AI assistant is only as good as the product information it can find, and most product data was written for a search bar, not for a machine suggesting products for someone. When what arrives doesn't match what shopper expected, it goes back in the box."

That last line is the whole story in one sentence. Product catalogs were engineered for keyword search — titles, attributes, thin descriptions. AI shopping assistants consume that same data and make personal recommendations from it. When the underlying record can't answer "will this fit this person," the assistant guesses. The customer discovers the guess at the front door.

What should marketing leaders take from this?

The survey points to a gap between AI adoption on the storefront and data readiness behind it. Marketing teams investing in AI-driven recommendations are, on this evidence, funding a returns pipeline unless product information improves first.

The stakes are measurable. Eight in 10 shoppers ready to reduce spend at a retailer that clamps down on returns means the cost of bad recommendations can't simply be passed to customers through stricter policies. The alternative — richer, structured, machine-readable product data — sits closer to the merchandising and content operations that CMOs increasingly oversee.

The 46% figure comes from shoppers who acted on AI advice, a self-selected group. Still, with 5,000 respondents across five countries and four years of trend data behind the survey, it's a signal worth budget attention.

AI recommendations are converting. They're also coming back. Angove's framing gives CMOs the defensible position: fix the product data, or keep paying for the boxes.