In China, AI-Generated Photos of "Defective" Goods Used to Win Refunds, and Shopping Platforms Fight Back With AI

Fake damage photos for refunds

Some Chinese buyers have used generative AI to make perfect products look broken. They then send the fake photos and ask for their money back. In one case in Shanghai, a plush toy photo had odd outlines and clear light and shadow mistakes, but a 50-yuan refund was still approved. In Hubei, a seller's photo of a rotten banana looked the same as the seller's own photo taken before shipping, and the refund was also approved.

Refunds without returning the item

In China, many online stores let buyers keep cheap or fresh goods and get only their money back. This saves shipping cost and time, and it helps both sellers and buyers. But it also makes it easy to cheat with fake photos showing tears, stains, or mold. Sellers find it hard to prove a photo is fake. Saying a picture looks strange is not enough to refuse a refund. Hiring an expert to check the image can also cost more than the product itself.

How well the AI checker works

• Alibaba's Taobao and Tmall added an AI tool that finds fake images. Sellers can report a suspicious buyer photo from their chat screen in the Qianniu tool.
• Taobao says its AI detection model is 95% accurate. When it calls an image fake, it is right 95% of the time.
• The model's recall is 50%. That means it finds only half of all the fake images.
• Rival JD.com and ByteDance's Douyin have also started testing services that spot AI-made fake images.

Why it matters

This shows a new kind of fight between AI tools. Cheaters use AI to make fake proof, and platforms use AI to catch it. Honest sellers can lose money, and honest buyers may face more checks. Chinese regulators are also updating rules to stop platforms from forcing sellers to accept refunds without returns.

What to watch

• Whether the new AI checkers catch more fakes, since the 50% recall rate is still low.
• How the updated online platform rules are written and enforced.
• Whether JD.com and Douyin launch their fake-image checkers across their whole platforms.