The holiday shopping season often comes with the inevitable – a purchase that doesn’t quite meet expectations. For many, a quick photo of a damaged item is all it takes to secure a refund from online retailers. It’s a system built on a fundamental assumption of trust. But what happens when that trust is systematically undermined by the very tools we’ve come to rely on for creative expression and innovation? We’re entering an era where generative artificial intelligence is not just creating art and music, but also crafting elaborate deceptions to exploit e-commerce refund policies.
The Rise of the AI-Fabricated Refund
Across Chinese social media platforms like RedNote, a growing chorus of complaints is emerging from e-commerce sellers and customer service representatives. They’re reporting an alarming increase in refund requests accompanied by seemingly doctored photos and videos. These aren’t just simple, easily debunked fakes. We’re seeing sophisticated attempts to fool even seasoned reviewers.
One seller recounted a customer’s claim of shredded bedsheets, but the Chinese characters on the shipping label appeared nonsensical – a subtle but telling detail. Another buyer presented a picture of a cracked coffee mug, a damage that looked less like a typical ceramic break and more like paper being torn. "This is a ceramic cup, not a cardboard cup. Who could tear apart a ceramic cup into layers like this?" the exasperated seller questioned, highlighting the sheer implausibility of the alleged damage.
The Products Most Vulnerable to AI Scams
Certain product categories are proving particularly susceptible to this new wave of AI-driven fraud. Sellers have noted a spike in fraudulent claims involving:
- Fresh Groceries: Imagine claiming your delivered produce was rotten or your live seafood arrived deceased. AI can easily generate images that depict these scenarios with unsettling realism.
- Low-Cost Beauty Products: For inexpensive items, the cost of returning the product often outweighs the refund amount. This makes them prime targets for scammers who can claim the item was damaged or unusable, knowing the seller is unlikely to pursue a return.
- Fragile Items: From delicate ceramics to electronics, anything prone to breakage is a potential target. AI can convincingly simulate cracks, dents, and other forms of damage.
These categories are often the focus because sellers frequently opt not to require customers to return the damaged goods before issuing a refund. This expediency, designed to improve customer experience, inadvertently creates a fertile ground for fraudsters.
A Case Study: The Case of the Errant Crabs
One particularly notable case, which garnered significant attention on Chinese social media and even prompted police intervention, involved a seller of live crabs. A customer submitted photos and videos claiming that most of the crabs arrived dead, with two having miraculously escaped their packaging. The accompanying videos showed the deceased crustaceans being prodded with a finger, a graphic attempt to prove their lifelessness.
However, the seller, Gao Jing, a seasoned crab farmer with over 30 years of experience, immediately spotted inconsistencies. "My family has farmed crabs for over 30 years. We’ve never seen a dead crab whose legs are pointing up," she stated in a video addressing the incident. The giveaway, however, was the crabs’ biological makeup. The initial video depicted a mix of two males and four females, while a subsequent clip showed three males and three females. Adding to the absurdity, one of the crabs in the second video appeared to have nine legs instead of the standard eight.
Gao reported the fraud, and the subsequent police investigation confirmed the videos were fabricated. The buyer was detained for eight days, marking a significant regulatory response to this emerging AI-enabled scam.
The Global Reach of AI Fraud
While the Chinese e-commerce landscape has been an early adopter and showcase for this trend, the problem is far from isolated. Forter, a New York-based company specializing in fraud detection, reports a substantial increase in AI-doctored images used in refund claims, estimating a rise of over 15 percent since the beginning of the year, with this figure continuing its upward trajectory globally.
Michael Reitblat, CEO and cofounder of Forter, notes that this trend gained momentum in mid-2024 and has since accelerated as AI image-generation tools have become incredibly accessible and user-friendly. He points out that the AI doesn’t need to achieve photographic perfection. "Frontline retail workers and refund review teams may not have the time to closely scrutinize each picture," he explains, suggesting that even imperfect AI creations can be enough to slip through the cracks.
Organized Crime Enters the Fray
What’s particularly concerning is that this isn’t just the domain of isolated individuals. Reitblat reveals that organized crime groups are employing the same AI-generated image tactics to orchestrate refund fraud on a massive scale. In one alarming instance, a single group submitted refund claims totaling over a million dollars. They used AI-altered images depicting damage to various home goods, submitting these claims within a narrow timeframe to overwhelm automated systems. To further obscure their tracks, the fraudsters utilized rotating IP addresses, a common tactic to mask their true locations and identities.
The Arms Race: AI vs. AI
In response, some e-commerce sellers are attempting to use AI to fight AI. One Chinese toy seller demonstrated a method where they feed refund request photos into an AI chatbot designed to detect doctored images. However, as Reitblat cautions, these AI detection tools are still in their nascent stages and are far from foolproof. Furthermore, even if a chatbot flags an image as potentially manipulated, e-commerce platforms may not always automatically side with the seller. This creates a complex and evolving landscape where trust, a cornerstone of online commerce, is under constant assault.
The Broader Implications for E-commerce
This situation mirrors a previous backlash on Chinese digital marketplaces, where sellers were criticized for using AI-generated product photos that created unrealistic expectations for consumers. Shoppers lamented that online buying had devolved into a gamble, where the received product bore little resemblance to the advertised image.
These two trends – AI-generated product images and AI-generated damage claims – are, in essence, two sides of the same coin. They both highlight the precarious position of trust in e-commerce. The widespread availability and increasing sophistication of AI are making it harder to operate under the assumption that most online interactions are honest.
Existing safeguards, such as AI watermarks embedded in generated images, are often easily circumvented or removed. For e-commerce platforms to maintain functional systems built for human interaction, they must innovate. This could involve implementing more robust verification processes, revising refund policies to better account for sophisticated fraud, or developing stronger accountability mechanisms specifically designed to address AI-enabled scams.
The future of online shopping hinges on the ability of platforms and sellers to adapt. As AI continues to evolve, so too will the methods of those who seek to exploit it. The challenge lies in staying one step ahead, not just through technological solutions, but by fostering a renewed sense of digital integrity in an increasingly AI-influenced world.
This article is an adaptation of insights originally featured in Zeyi Yang and Louise Matsakis’ Made in China newsletter.