AI news · AI shopping · returns
Nearly half of AI-led US shoppers sent a buy back, survey says
Blue Yonder finds wrong descriptions, more than wrong sizes, drive returns of AI-recommended items, which its chief executive blames on product data written for a search bar.
Among the roughly one in four US consumers who have bought something on an AI assistant’s advice, 46 per cent sent such a purchase back in the past year. The figures come from a Blue Yonder survey of more than 5,000 people in five countries, 2,000 of them American, reported by DC Velocity and, on 3 November 2026, by Retail Dive. The firm sells supply chain software, so returns are its market.
The reason for sending things back is what stands out. Fit is the usual reason across all US returns, at 42 per cent, with description at 20. On AI picks it turns round: 36 per cent differed from their description and 23 per cent were the wrong size or fit.
Blue Yonder’s chief executive, Duncan Angove, puts the blame on retailers’ product data, most of which he says was “written for a search bar”. The survey measures shoppers who have returned at least one item, so it gives no return rate per AI recommendation.
Fixing that text is writing work, and it has a known price. Copy.ai’s Chat plan was $29 a month, or $24 on yearly billing, and Jasper Pro was $69, or $59, at our 3 November check (our price sheet).
Items that differ from their description make up 36 per cent of these returns against 20 per cent of returns overall, so a retailer counting on AI-led sales should check its product text before its sizing data.
Sources
- Retail Dive: AI recommendations are leading to returns. Checked
- DC Velocity: Blue Yonder: 46% of people who shop with AI have returned what they bought. Checked
More AI news from
- OpenAI puts ChatGPT search at a fifth of Google’s volume
- Canva restructure hits 1,271 roles as its growth slows
- Supermetrics lets Claude draft campaigns on six ad platforms
- Shopify says orders from AI search rose 3.3 times in a year
- EU court adviser says Gemini can breach press publishers’ rights
- Google ad strength buys more reach than sales, Adalysis finds
- Smartly’s new tier sets modelled ad lift beside platform figures
- YouTube takes AI tracks for copyright claims beyond the US
- Anthropic’s FlashProve aims to prove which AI model answered you