CCTV Investigation of E-commerce AI Model Images: Three-Legged Deformities and Hand Anomalies Frequently Appear, Merchants Accused of Concealing AI Nature
A CCTV News investigation found that before Double 11, a large number of clothing seller show images on e-commerce platforms were AI-generated, with some model images showing distortions such as three legs and hand poses that violate human anatomy, yet shop customer service still claimed they were real photos. A clothing e-commerce manager said that generating a full set of images with AI takes only a few minutes and costs about 30 to 40 yuan, while sending clothes out for a photo shoot costs 600 to 800 yuan per set and takes at least 3 to 4 days. Xie Yongjiang of Beijing University of Posts and Telecommunications said that if merchants falsely claim real photos and the difference between AI images and the actual product is enough to affect purchasing decisions, it may constitute false advertising or even consumer fraud.
Why read this
AI score 55
The report uses merchant cost data and expert statements to present the prevalence and distortion risks of AI model images in clothing e-commerce.
Topics
Body · AI translation
As Double Eleven approaches, AI models on e-commerce platforms have once again drawn attention. In some product images, the "models'" hand poses clearly violate human anatomy, and some even have three legs... Many consumers wonder whether AI models can truly reproduce how clothing looks. Have AI model images already become a trend in the apparel e-commerce industry? Can "generative" seller show images objectively present clothing quality and let consumers choose with confidence? Fast speed, low cost: AI-generated seller shows have become an industry trend. Opening major e-commerce platforms and searching for apparel products, all kinds of model display images can be seen everywhere. A reporter randomly searched for a certain brand's long-sleeved knit sweater, and the main product image was a model wearing the item. The reporter uploaded the image to the National Anti-Fraud Center App's "AI Content Identification," and the test result showed that the "image contains traces of AI generation," but the store's customer service claimed that the "product display image is a real photo." △ Model image of a knit sweater sold by a store on an e-commerce platform. Another woolen suit jacket on sale had a short video attached to the product link featuring the same model and the same scene. The reporter observed that the model's voice and lip movements did not remain consistent at all times, and this video was also detected by the National Anti-Fraud Center App as "containing traces of AI generation." Multiple apparel e-commerce practitioners confirmed that the use of AI model images is quite widespread in the industry today, especially since the beginning of this year, when many merchants have launched a "generative" seller show model. Zhang Zhitao, the head of an apparel e-commerce company, did some calculations: "In the past, when we sent sample garments out to be photographed, it took at least 3 to 4 days to get the final images. After switching to AI, a full set of images can be generated in a few minutes. Previously, mailing a set of images for photography usually cost 600 to 800 yuan, but now the cost of AI generation is only thirty or forty yuan." The reporter learned that a certain scale of AI model image industry has already formed on the market, including subdivided fields such as agency production, prompt instructions, and vertical AI tools for apparel. In the AI model image production systems developed by many technology companies, merchants only need to upload clothing photos, and the platform can automatically generate model effect images. "Models" grow three legs? Early AI was prone to body distortion. Under cost reduction and efficiency improvement, how effective are "generative" seller shows? Wang Jun, the head of an apparel e-commerce company for middle-aged and elderly people, said that in the early days, when the technology was immature, AI models were prone to body distortion, but with technological iteration, the degree to which AI model images reproduce apparel products has also improved significantly. However, the apparel industry has its particularities, and "generative" seller shows still cannot completely replace high-definition real photos of product details for the time being. △ Some "generative" seller shows in which model distortion appears. Wang Jun said: "When we tested products in the early stage, we basically used AI generation for all of them. After testing the styles and forming basic data, we would use real people for real photos of 'hit products.' Because for the subtle parts of some materials, it is difficult for AI to present them accurately." In addition to color differences and fabric texture, some netizens also pointed out that ordinary people's body shapes vary greatly, and "generative" seller shows can only identify the appearance and design of clothes, while the effect when worn varies widely. Xue Yan, the owner of a clothing wholesale studio, admitted frankly, "Whether a piece of clothing is comfortable to wear, whether it constrains the neck or the armpits, whether trousers bunch up at the waist when you sit down, whether they are comfortable... these real feelings when worn cannot be shown by AI models. The images are virtual, but clothing ultimately has to be worn on real people." Experts: Concealing the AI attribute, serious distortion of effect may be suspected of false advertising. Xie Yongjiang, director of the Internet Governance and Legal Research Center at Beijing University of Posts and Telecommunications, said that objectively speaking, the current application of AI model images in apparel e-commerce does indeed have shortcomings. AI model images are difficult to satisfy consumers' needs to see the real effect clearly and judge how the clothing feels when worn. At the same time, vertical virtual fitting solutions bound to sample garment scanning, fabric collection, and 3D fabric computation can theoretically achieve high-fidelity display, but the cost is relatively high, and it is currently difficult to popularize among small and medium-sized merchants. In the future, it is more suitable as a supplement to real photo materials rather than a complete replacement for real photos. Xie Yongjiang pointed out that the law does not prohibit the normal use of AI models, but relevant practitioners must clarify two risk boundaries: first, whether there is a substantive difference between the image display effect and the actual product, and whether it misleads consumers into purchasing; second, whether merchants deliberately conceal it. If a merchant falsely claims that it is a real photo, while the cut and fabric in the AI image differ greatly from the actual product, enough to affect consumers' placing orders, it may constitute false advertising and may even be identified as consumer fraud. ▌ Source of this article: CCTV News WeChat public account.