AI-generated fake images used in e-commerce refund fraud in China
Buyers on Chinese e-commerce platforms have been using AI-generated fake images of product defects to claim refunds without returning merchandise. A fig seedling seller and other merchants reported cases where AI-fabricated photos were approved for refund by platforms within minutes, prompting the National Anti-Fraud Center to launch an AI content verification tool to detect fraudulent images.
What happened
Buyers on e-commerce platforms in China have been submitting fake AI-generated product images to claim refunds without returning goods. Zhang Xiansheng, a fig tree seedling seller, received a 'refund only, no return' claim with an image showing completely yellowed seedlings, despite shipping healthy seedlings in pots that could not dry out in two days. The buyer claimed the seedling had died in transit.
Zhang Xiansheng identified the image as AI-generated based on horticultural knowledge: the yellowed leaves were fully extended rather than shriveled and drooping as real dehydrated plants appear. Despite his objection and request for return-for-refund, the buyer requested platform intervention. The platform approved the refund within minutes based on the fake image.
Similar cases affect multiple product categories: clothes racks photographed breaking at identical intervals with precise fractures, plastic phone cases with AI-generated jagged edges, and beaded bracelets appearing as hollow shells. Some images even bore watermarks indicating generation by AI tools like 'Doubao'.
Reported impact
Zhang Xiansheng stopped selling on the platform after this incident, initiating shop closure procedures. Multiple merchants reported the scheme has become widespread, with the fraudulent refund mechanism threatening normal e-commerce operations.
Industry experts stated that AI-fabricated image fraud constitutes civil fraud and, if unchecked, will threaten the normal order and healthy development of e-commerce.
Index analysis
The National Anti-Fraud Center App launched an 'AI Content Verification' function supporting detection of AI-generated images, videos, text and audio. Testing by reporters found all suspected AI-generated 'after-sales images' submitted to the verification tool were identified as containing AI generation traces, while genuine photographs were identified as unmanipulated.
Detection methods use small models trained on AI-generated content patterns combined with large models to identify physical defects and logical inconsistencies in images. Industry observers stated that current mature verification tools can effectively distinguish real content from AI-fabricated content with minimal false positives.
Governance controls
The National Anti-Fraud Center launched the AI Content Verification tool as a public welfare function in March 2026 to help identify fraudulent content. Experts noted that platforms have an obligation to adopt such technical means to maintain transaction order and protect merchant and consumer rights.
Beijing University of Posts and Telecommunications researchers recommend using AI to counter AI fraud. Verification tools can significantly reduce the cost for merchants to identify fakes and serve as evidence for disputes with platforms.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.