AI product images flood e-commerce: merchants save ninety percent of costs, buyers struggle to tell real goods

The Double Eleven major promotion has entered a white-hot stage, and the "face-changing" on e-commerce pages is also accelerating. More and more merchants are handing product images, detail pages, and even buyer shows in comment sections over to AI generation: images are becoming more and more perfect, real people fewer and fewer, and it is becoming harder and harder for consumers to clearly see a product's real material and lighting effects. When "shopping by looking at pictures" turns into "opening blind boxes blindfolded," is AI a sharp tool for reducing costs and increasing efficiency, or a hidden risk that overdrafts industry trust? From merchants to platforms, from regulators to consumers, a battle of offense and defense and a game surrounding AI selling goods is unfolding. Phenomenon: When product pages are filled with "AI images" The Double Eleven event has strong discounts, and Ms. Xie, a resident of Jinan, prepared to take the opportunity to buy a pair of shoes online. After opening several stores, she found that not only were the model videos and images on the product display pages astonishingly perfect, but even the buyer shows in the comment sections had an indescribable fakeness. Only after careful identification did she realize that all the promotional materials related to the products were AI images and AI videos. "I don't know when it started, but when buying things you can't see real product images anymore, like a large blind-box-opening scene." Ms. Xie's experience is not accidental. On social platforms, many netizens have reported similar problems. Some netizens left comments complaining: I placed an order because the online pictures made my heart flutter, but the actual product made me furious. Merchants use AI to play a fine game of "fraud"; someone wanted to buy a lounge chair "full of relaxed vibes," but not a single detail of the product received matched. Some consumers even said that after ordering pre-sale women's clothing shown with AI model images, they found upon delivery that the "95% pure cotton fabric" originally marked on the detail page had been quietly changed to a fabric that "contains no cotton." A reporter's review found that this round of AI image penetration into e-commerce is far faster than expected. From product main images and model videos to detail pages and buyer shows in comment sections, AI-generated materials are replacing real photographed images in batches. As AI-generated exquisite images spread from product detail pages to comment sections, even the last line of trust defense, the "buyer show," has been breached, making it increasingly difficult for consumers to confirm real materials, cuts, and how items look when worn before placing an order. "For most customers, their ability to identify AI product images is not as strong as that of us professionals," admitted Chu, an e-commerce merchant mainly selling hanfu and ancient-style performance costumes. In his view, professionals may be able to judge by first impression that an image "is fake," but ordinary consumers often lack this ability and are very easily misled by highly realistic AI images. Cui Lili, deputy dean of the Institute of Digital Economy at Shanghai University of Finance and Economics, also pointed out that exquisite AI-generated images may conceal flaws in real environments and small details that consumers care about, increasing communication costs and even return and exchange costs for both buyers and sellers. Merchants: AI saves real money According to the "AI Reshaping Operations: 2026 China E-commerce AI Application White Paper" jointly released by Taotian Group and "Tianxia Wangshang," among nearly a thousand merchants it surveyed and interviewed in depth, more than 95% had already applied AI tools, more than 60% were using them frequently every day, and 88% of merchants planned to continue maintaining or increasing AI investment. Why are merchants so enthusiastic about AI? The answer is first of all cost; the cost advantage brought by AI is real. Chu did the math for the reporter: previously, making a decent set of product images often cost one to two thousand yuan; now, with AI-generated images, the cost is only ten-plus to twenty yuan, directly cutting costs by more than 90%. On social platforms, a women's clothing graphic designer said the industry ecosystem is undergoing obvious changes. In the past, women's clothing product images required organizing real-scene shoots and then handing them to graphic designers for post-production retouching; now many brands choose to cut graphic design positions and instead have operations staff directly use AI tools to complete image generation, thereby achieving "cost reduction and efficiency improvement." Beyond cost, AI also greatly compresses the production cycle for product visual materials, thereby accelerating the pace of clothing launches. Chu introduced that under the traditional model, from arranging a shoot to receiving usable final footage often required waiting two weeks or even longer, while AI image generation can produce images the same day. The shortened cycle directly empowers product development, making it possible for small and medium-sized merchants to quickly test styles and launch new products at high frequency. For small stores with limited capital and manpower, AI has removed real-life shooting, the heaviest fixed cost, and the threshold for realizing the business idea of one person operating an online store has dropped significantly. A reporter's search found that a large number of AI image generation service providers for clothing merchants have already emerged online, with businesses covering basic image generation, customized solution image generation, and more, with quotations ranging from 10 yuan to 500 yuan. At the same time, AI women's clothing image-and-text sales tutorials can be seen everywhere; for a dozen or so yuan, one can buy a material package containing prompts for thousands of popular women's clothing styles, lowering the entry threshold for clothing merchants. At the level of image quality, the realism of AI-generated materials continues to iterate, and some works have reached a level that is almost indistinguishable from the real thing. Chu admitted that whether for mainstream large models overseas or in China, image generation capabilities have improved greatly. "Some AI images are hard to distinguish as real or fake at a glance even for industry professionals." In addition, the application of AI model images is no longer limited to small and medium-sized e-commerce sellers. Many netizens have found that high-end women's clothing brands have also begun testing AI model materials. A consumer in Shanghai reported that in the promotional materials of some luxury goods stores on Huaihai Middle Road, AI model images have already appeared in large numbers, and AI visual solutions are penetrating from a money-saving tool for small sellers into the high-end fashion track. Contradiction: Technological equality and trust overdraft Chu believes that AI has achieved a kind of "technological equality." Early e-commerce online images were also deceptive, but at least there was a threshold: only good photographers could produce good images; now, as long as they are willing, any merchant can generate good-looking images with one click. For example, with the long-criticized practice of fake orders, AI provides more convenient tools for such operations. As AI pushes the threshold for generating virtual "buyer shows" to nearly zero, those highly polished positive reviews in comment sections that are detached from reality have also become another veil blinding consumers. In Chu's view, it is understandable for merchants to use AI for reasonable optimization and truthful display of product advantages; but if the products themselves are of poor quality and prone to not matching the goods, using retouched AI images to beautify and conceal this is essentially fooling consumers. "Customers may be deceived once, but it is hard to generate repeat purchases, and in the end the merchant is the one who suffers." It is worth noting that AI, this "double-edged sword," does not only point at merchants. Mr. Chu mentioned that some buyers also use AI to forge images of product defects, "P" holes into intact clothing and "add" mold spots to fruit, in order to apply for refunds without returning goods, harming merchants in reverse. This offense-and-defense battle around AI is drawing both buyers and sellers into the same net. Industry insiders analyze that the shooting costs saved by AI may ultimately translate into return costs, trust costs, and platform governance costs. When product images cannot accurately present the products themselves, these costs will sooner or later be recalculated into every transaction. "The old PS images at least showed what the product looked like; now AI images are everywhere, and even the material of the product received is fake" "I have already blocked many sellers; in the future I may have to shop more at offline stores" "When I see a merchant using AI images, I just skip and don't buy"... On social platforms, many consumers have expressed their dislike of AI images. Cui Lili warned that if it is too complicated for consumers to distinguish real information, it will reduce the purchasing experience and even reduce the willingness to buy online, "In the long run, this is not conducive to building trust and a sustainable relationship between buyers and sellers." Governance: Labeling measures take effect, but the game is still in a "vacuum period" In fact, in the face of the surging chaos of AI in e-commerce, regulators and platforms have not left it unchecked. The "Measures for Labeling AI-Generated Synthetic Content" and the "Cybersecurity Technology - Labeling Method for AI-Generated Synthetic Content," which took effect simultaneously on September 1, 2025, explicitly require that content such as text, images, and videos generated by AI must "reveal their identity": on the one hand, explicit labels in the form of text, graphics, etc. must be added in appropriate positions so that users can clearly perceive them; on the other hand, implicit labels must be embedded in file metadata for backend traceability and verification, and images are also subject to the unified dual-labeling requirement. The "Measures for the Supervision and Administration of Live-Streaming E-Commerce," which took effect on February 1 this year, stipulate that when live-streaming rooms use AI-generated character images or video footage to carry out business activities, they must complete labeling in accordance with national standards and continuously remind consumers that the content is AI-generated. Taobao launched a platform-wide special governance campaign against fake AI images in March 2025, building recognition models and combating distorted AI product images through a combination of source interception, cleanup of existing content, and labeling reminders. After Taobao's backend connected to the AI fake-image recognition model, within one month it had cumulatively intercepted nearly 100,000 fake images involving excessive PS and AI exaggeration; Douyin announced an upgrade to its AI content labeling function on September 1, 2025, launching creator labeling tools and metadata read-write capabilities to implement AI content annotation and traceability requirements... However, in the vast e-commerce ecosystem, the implementation of rules still faces numerous obstacles. Mr. Chu admitted that the platform's ability to recognize AI images is "very limited," "As technology strengthens, new technologies emerge and iterate, and this system cannot stop it," with both sides caught in an endless "mutual game." He believes that the industry is currently in a "vacuum period," "let the bullets fly for a while," and that truly forming a healthy relationship may still require a period of market cooling-off. Cui Lili suggested that platforms can require merchants to proactively declare and label, or prompt consumers to verify real and objective details, "but this will certainly increase transaction costs." Trend: In the AI era, the bottom line of trust must be upheld The AI wave is unstoppable. The "Report on the Development of Generative Artificial Intelligence Applications (2026)" released by the China Internet Network Information Center shows that as of the first half of 2026, the number of generative AI users in China exceeded 700 million, with a penetration rate of over 50%. The Chinese Academy of International Trade and Economic Cooperation of the Ministry of Commerce believes that AI is upgrading from a single-point tool to the underlying infrastructure of the e-commerce and retail industries, achieving systematic transformation across the entire chain. Beyond technology inclusiveness, governance issues such as content authenticity, labeling compliance, and products not matching descriptions will become accompanying propositions for the long-term development of AI e-commerce. A research report by Shenwan Hongyuan pointed out that AI has moved past the pilot stage and entered a period of large-scale value realization in the e-commerce industry, becoming e-commerce infrastructure. AI has lowered the threshold for content production, and risks such as fake materials and AI-fabricated buyer shows have simultaneously amplified; platform governance and the implementation of regulatory rules will determine the long-term health of the industry. Analysys believes that the focus of the next stage of AI e-commerce is not just efficiency improvement, but the establishment of a trustworthy AI transaction system: AI content labeling and the identification and governance of falsely generated content will, like recommendation algorithms and intelligent fulfillment, become indispensable components of this infrastructure; whether the inclusiveness of technology and consumer trust can be balanced is the core test for platforms and merchants. "All the phenomena in the development of e-commerce to date are the result of friction among multiple parties' interests." Cui Lili believes that the direction of future regulation is to allow new technologies and new tools to be applied within a reasonable scope, such as complying with commercial credit, protecting personal privacy, and adhering to business ethics; platforms should promptly identify the reasonableness of technology applications; and government departments should, in light of unreasonable behaviors emerging in the market, quickly iterate relevant laws and regulations and provide the administrative supervision they should.