.png){kind=logo}
Zomato founder and CEO Deepinder Goyal said some customers now use AI-edited images to fake food damage and claim refunds. He shared the details on YouTuber Raj Shamani’s “Figuring Out” podcast episode that aired on January 3.
Deepinder Goyal said some users digitally add insects, hair, or foreign objects into food photos. He also said AI tools can make intact items look damaged, including cakes that appear smashed or smudged.
He linked the trend to a sudden rise in cake-related complaints. Goyal said Zomato saw about a 5% spike in cake damage reports after the pattern emerged.
The issue drew wider attention after a Mumbai bakery, Dessert Therapy, alleged that a customer submitted an AI-generated damaged-cake image. The bakery said the image showed errors that suggested AI generation, including inconsistent text and unnatural details. The customer sought a refund of Rs. 1,820 on a Rs. 2,500 cake, the bakery said.
Goyal said Zomato uses an internal “karma score” to judge disputes between customers and delivery partners. The system compares past behavior patterns to decide which side appears more credible in a specific complaint.
He said the platform does not always reach a clear answer in real time. As a result, he said Zomato often refunds customers while it keeps delivery partners active, unless repeat patterns create strong evidence. He said the company absorbs losses in a large share of disputed cases, which he put at roughly 50% to 70%.
Goyal also said fraud does not only come from customers. He said Zomato terminates about 5,000 delivery partners each month, citing repeated fraud patterns. He listed scams such as marking orders delivered without delivery and cash-on-delivery misuse around “no change” claims.
Also Read: Zomato’s Response to Vendor Mislabeling at Hyperpure
The same tactics that hit food delivery also affect wider e-commerce returns. Fraud teams now face images that look realistic enough to pass fast checks, especially when platforms rely on photo-based verification.
Competitors also invest in stronger fraud controls. Swiggy, for example, has highlighted device-first risk intelligence work with SHIELD to identify abuse patterns across users and delivery partners.
Goyal said inaccurate refunds can hurt restaurants and delivery partners as well as the platform. As AI image editing and solution-based tools become easier to use, food delivery fraud teams may rely more on pattern scoring, device signals, and deeper verification to reduce false refund claims while keeping service disputes fair.
