AI is being exploited to fabricate evidence, and online sellers are facing a new wave of scams.

Over the past decade, e-commerce has experienced explosive growth based on a seemingly simple yet incredibly fragile foundation: trust . Buyers believe that the products they receive will be exactly as described. Sellers believe that customers will be honest in their feedback. And platforms believe that visual evidence such as images and videos provided by users reflect objective truth.

Images of defective, broken, scratched, or damaged products during shipping have long been considered "gold standard evidence" for online platforms to verify refund requests. With just one convincing photo, buyers can easily receive their money back without going through a complicated verification process. This mechanism is designed to protect consumers and enhance the shopping experience.

However, this very simplicity and humanity are being subtly exploited by generative AI. As technology can create increasingly realistic fake images and videos, the already fragile trust system of e-commerce is beginning to show worrying cracks.

Gone are the days of crude photo editing tricks; the new wave of fraud is using AI as a "technological accomplice," creating illogical refund proofs that are nonetheless capable of bypassing automated verification systems. Sellers become victims, while platforms face a difficult dilemma: protecting customers while maintaining fairness within the ecosystem.

1. Online shopping platforms and the reliance on "real-life" images.

Most e-commerce platforms today, from global giants to local ecosystems, operate on a common assumption: the majority of buyers are honest. Therefore, refund processes are often designed to prioritize customer experience, minimizing barriers and procedures.

With millions of transactions occurring daily, requiring a physical inspection of every product is impractical. Instead, images and videos provided by customers become the primary authentication tools. A photo of a cracked cup, a video showing a dented box, or a damaged package of food is often enough for the system to make a decision.

This method worked effectively for many years because creating fabricated evidence requires effort, skill, and time. Not everyone has the ability to stage a convincing "defective product" scenario. The cost of fraud often outweighs the benefits, making this behavior uncommon.

But the emergence of Generative AI has completely reversed that balance.

Today, with just a few lines of description (prompt), someone with no design skills can create images of damaged products with incredibly realistic lighting, textures, and backgrounds. AI video can even simulate motion, shadows, and physical details with astonishing accuracy. The technical barriers have virtually disappeared, paving the way for fraud on a large scale.

2. AI "helps" create illogical refund proof.

In China, one of the world's largest and most competitive e-commerce markets, a wave of AI-powered refund scams is rampant. On social media platforms like RedNote and Douyin (China's TikTok), numerous shop owners have shared their hilarious and frustrating stories of receiving bizarre and illogical refund proofs.

The common thread in these cases is that the images look very real at first glance, but the more closely you examine them, the more illogical they become.

A bed linen seller reported receiving a photo of a product torn into tiny pieces. The fabric was frayed, the tears looking natural enough for the automated system to accept the refund request. However, upon closer inspection of the label, the seller discovered that the Chinese characters on it were completely meaningless, randomly put together.

Another case involving household goods further angered the selling community. A buyer sent a photo of a ceramic cup that was "cracked," but the crack had a texture similar to torn paper. Ceramic surfaces are dense and brittle; they couldn't possibly peel like that. "Who could tear a ceramic cup like tearing paper?" the seller asked helplessly.

These illogical details often only become apparent when humans observe closely. Meanwhile, automated review systems are easily fooled by images that appear "realistic enough" overall.

3. The live crab incident on Douyin: When AI crosses the line of petty fraud.

The peak of this wave of fraud was an incident on Douyin in November involving a seller of live crabs. Customers sent videos showing the crabs were dead upon receipt, even poking them with their fingers to prove the product's condition.

For most sellers, this would probably be sufficient evidence to accept a refund. However, the shop owner noticed something unusual from seemingly small details. According to him, there had never been a case of a dead crab with its legs pointing upwards in the way shown in the video.

The seller's suspicion prompted him to investigate further. When comparing the two videos the customer sent, he discovered that the number of crabs and their sexes didn't match. In fact, in one frame, a crab appeared to have... nine legs.

The incident was reported to the authorities. After an investigation, the police determined that the video was faked using AI technology. The buyer was detained, and this is considered the first AI-generated refund scam in China to be publicly prosecuted.

This incident not only shocked the seller community but also sounded an alarm for the entire e-commerce industry: AI is no longer a testing tool, but has become a real means of committing crimes.

4. Gaps in the refund policy

According to many merchants, fraudsters don't randomly select products to target. They aim for categories with the most flexible refund policies, where sellers are less likely to request returns.

Fresh food is a prime example. For hygiene and safety reasons, many platforms allow refunds without requiring returns if the product is damaged. Inexpensive cosmetics and fragile items also fall into this category, as the return shipping cost can sometimes exceed the product's value.

Policies designed to protect the consumer experience inadvertently create blind spots for fraudulent behavior. When combined with AI generation, these blind spots are amplified into major vulnerabilities.

Sellers face a difficult choice: accept a refund to avoid negative reviews and trouble, or insist on filing a complaint at the risk of losing customers and wasting time. In many cases, small financial losses, when accumulated over time, will erode profits and business motivation.

Although China is currently a hotspot, this phenomenon is not limited to a single country. Forter, a New York-based fraud detection company, reports that the number of AI-generated or digitally manipulated images in refund requests has increased by more than 15% globally since mid-2024.

This period coincided with the explosion of easy-to-use, inexpensive, or even free AI image creation tools. Tools once reserved for professional designers have now become accessible to the general public, leading to unintended consequences.

Michael Reitblat, CEO of Forter, warns that organized crime has also quickly entered the scene. No longer just small-scale fraud by individuals, many groups have built sophisticated processes: using AI to create images of damaged or broken products, combined with networks of fake accounts and constantly changing IP addresses to send mass refund requests.

In a short period of time, several groups carried out scams totaling over $1 million, overwhelming the review system and making tracing them extremely difficult.

Essentially, e-commerce cannot function without trust. Every process, policy, and experience is designed based on the assumption that the majority of users behave honestly. When this assumption is shaken, the entire ecosystem faces a crisis.

Some sellers are starting to turn to AI to fight AI. Image analysis tools, capable of detecting tampering or generative effects, are being deployed as a new layer of defense. However, this is an unequal technological race. As generative AI becomes increasingly sophisticated, the detection capabilities of its counterparts must also be constantly upgraded.

Meanwhile, e-commerce platforms are often hesitant to fully side with sellers, due to the risk of harming the buyer experience. Just one overly rigid policy could trigger a wave of negative reactions from genuine customers.

If platforms tighten refund processes, require more proof, or extend processing times, honest buyers will be the first to suffer. But if no action is taken, fraud will continue to increase, shifting the risk costs to sellers and ultimately the entire ecosystem.

5. Perspectives from design and technology

From the perspective of a seasoned designer, this issue isn't just about financial fraud; it's also a crisis of image authenticity. For decades, images have been considered the most powerful visual evidence. "Seeing is believing."

Generative AI is breaking that unspoken convention. When images can be created without physical objects, the concept of “visual evidence” must be redefined. This impacts not only e-commerce, but also extends to journalism, legal matters, and media in general.

For the design industry, this is a time to reconsider the ethical role of technology. AI is a neutral tool, but how people use it will determine its social impact. When the line between creativity and fraud becomes blurred, responsibility lies not only with the end user, but also with the developer and the distribution platform.

E-commerce is built for people, based on trust between people. But as AI increasingly intervenes in that process, maintaining the system becomes more difficult than ever.

The scam of using AI to create fake product images to trick people into giving refunds is just an early sign of a larger problem: when technology overrides existing societal norms. How we react today will shape the future of digital commerce for years to come.