What looks like a simple customer complaint may no longer be what it seems. Across food delivery, e-commerce and insurance, AI-generated images are being used to fabricate evidence and secure fraudulent refunds. As platforms built on trust struggle to keep up, the line between genuine grievances and digital deception is rapidly blurring.
Take for instance, food delivery platforms like Zomato and Swiggy Instamart, which are grappling with a growing wave of false complaints driven by manipulated claims of damaged or contaminated orders. A few months back, customers began using AI tools to falsely claim refunds. Deepinder Goyal, entrepreneur, investor and co-founder of Zomato, said in a podcast discussion that users have increasingly turned to digitally altering food images before filing complaints — adding fake insects, inserting foreign objects, or making intact items like cakes appear smashed or spoiled.
At Zomato, a sudden spike in complaints about ‘smashed cakes’ raised red flags internally. Investigations revealed that many claims were backed by AI-edited images designed to secure refunds or replacements. The problem, Goyal noted, is far from isolated. Manipulated visuals are now commonly used to falsely suggest poor delivery quality.
Smashed Cakes
Alongside AI-driven tactics, older forms of deception still persist. Some customers reportedly tamper with their orders-placing hair in food, damaging packaging, or altering items themselves — before photographing and submitting complaints. Each refund comes at a cost, often borne by restaurants and delivery partners, making it increasingly difficult to balance customer satisfaction with fairness.
Over the past five years, such false complaints have grown steadily, driven by both economic pressures and easy access to these services. Generative AI has only accelerated the trend, enabling highly realistic manipulation with minimal effort.
Customers can now fabricate receipts, replicate brand logos, and create convincing proof of purchase. Products can be digitally altered to appear defective-cracked cosmetics, torn clothing, or damaged electronics-often presented through multiple angles to enhance credibility. A perfectly intact item can be made to look faulty within seconds. This is not harmless behaviour. Even when carried out by otherwise legitimate customers, it amounts to theft.
High Cost of Synthetic Deception
The financial impact is substantial. The Reserve Bank of India has proposed compensation of up to Rs 25,000 for small-value fraud losses. Meanwhile, Indian banks reported 13,469 cases of card and internet-based fraud in FY 2024-25, amounting to `5.2 billion in losses-lower than the previous year, but still significant. Insurance fraud is also evolving. In one recent instance, an image of a damaged Land Rover was altered with a fake number plate and reused across multiple claims in the United Kingdom. Cardiff-based insurer Admiral recorded a 71% rise in fraud during 2025 compared to the previous year, partly blaming the increased use of AI software to manipulate evidence.
The Insurance Fraud Bureau in the UK said the industry was heavily concerned about AI-generated claims and was investing in technology to tackle the threat.
The insurers have identified fabricated documents, exaggerated damage, and entirely imaginary assets in AI-generated submissions.
Even as online marketplaces are facing similar challenges, sellers report suspicious refund requests backed by images with clear inconsistencies — such as nonsensical shipping labels or unrealistically damaged products — pointing to the increasing use of AI-generated visuals in everyday deceptions.
A viral example underscores how easily these systems can be exploited. A customer ordering eggs on Swiggy Instamart received a tray with only one cracked egg. Instead of sharing the actual image, they used Gemini Nano to edit the photo, making it appear that the entire tray was damaged. The altered image was convincing enough for customer support to approve a full refund without further checks, sparking widespread concern. The problem lies not in the technology itself, but in outdated verification systems. Many platforms still rely on trust-based mechanisms that were not designed to detect sophisticated, AI-generated evidence.
To counter this, companies are beginning to adopt stronger safeguards. These include in-app photo and video capture to prevent the use of external images, AI tools to detect manipulation, and behavioural analysis to flag suspicious activity.
Zomato, for instance, uses an internal ‘karma score’ to assess the credibility of user complaints, giving more weight to trusted customers while scrutinising repeat offenders. The platform has also tightened policies around AI-generated images in restaurant menus and is expanding manual review processes for flagged cases.