Marketplace Refund Impersonation Controls: Fake Damage Photos, Fake Support Agents and Redirected Refunds

Three Ways Impersonation and Synthetic Media Hit Marketplace Refunds

Refunds attract three distinct kinds of fraud, and each needs its own control. The first is fabricated evidence: a buyer claims an item arrived damaged and backs the claim with an AI-generated or edited photo or video. The second is fake support: a scammer poses as the marketplace or a seller and contacts a buyer who is waiting for a refund. The third is redirection: an attacker takes over a buyer or seller account and changes where the refund or payout goes.

Fabricated Damage Photos in Refund-Without-Return Claims

Refund-without-return policies exist so a buyer does not have to ship back a spoiled or worthless item. They rely on the buyer's photo being honest. Generative image tools have made that assumption weak. The OECD AI Incidents Monitor logged an incident dated 23 July 2025 describing consumers in China using AI tools to create images and videos of damaged goods to claim refunds without returns, citing Chinese media reports of AI-generated photos of rotten durian and videos of destroyed products. Coverage cited by the OECD included lawyers saying the conduct could amount to fraud.

The same technique works on any platform that accepts uploaded photos as proof. It does not need sophisticated tools: a genuine photo of the item can be edited to add a crack, a stain or mould, and the result often looks more convincing than real damage.

Controls start with how evidence is captured. Ask buyers to take photos inside the app rather than upload from a camera roll, so each image is bound to the claim, the device and the time. Screen images for signs of generation and pixel-level editing, and check for the same image, or a near copy, appearing across claims and accounts. Send flagged high-value claims to a reviewer, and use forensic analysis when a seller disputes the outcome.

Scammers Posing as Marketplace Support

A buyer waiting for a refund is an easy target for someone pretending to be the marketplace. The FTC's guidance on refund and recovery scams describes callers who claim to be from a company or government agency, promise to recover money, and then demand an upfront fee, payment by gift card, crypto or wire, or bank and card numbers to deposit the refund. Voice cloning and synthetic video make those approaches more convincing, for example a call that sounds like a support agent the buyer has spoken to before.

The FTC's Trade Regulation Rule on Impersonation of Government and Businesses took effect on 1 April 2024, letting the agency seek civil penalties and redress against scammers who impersonate businesses. Enforcement comes after the harm, so marketplaces should design refunds to leave scammers no opening: issue refunds only through the platform to the original payment method or in-app balance, never ask buyers for card numbers or remote access, publish the only legitimate support channels, and show in-app warnings when a buyer has an open refund.

Protecting the Refund Destination

A refund that goes to the wrong place is a loss even when the claim is genuine. Attackers who take over a buyer account can add a new payout method and file claims; attackers who take over a seller account can change the bank account that receives payouts. Treat any change to refund or payout destination as high risk: require a fresh face check in the app, notify the existing email and phone before the change applies, and hold the first payment to new details for a short period. Account recovery requests that come by chat or phone should never be enough to change where money goes.

Where Detection Helps Refund Teams and Where It Does Not

Image forensics can show that a damage photo was generated, edited or reused, which shifts a dispute from one party's word against another's to documented evidence. It cannot show that an unedited photo depicts the item from this order, that a genuinely damaged item was damaged in transit rather than by the buyer, or that a buyer is lying about non-delivery. Those questions need order data, carrier records and account history.

False positives are a real cost. Photos taken in poor light, heavily compressed by messaging apps or filtered by phone cameras can look unusual. Use detection to route claims, set thresholds by claim value, and keep a human reviewer between a flag and a denial.

Investigating Disputed Refund Evidence With NeuroDetect

NeuroDetect is Neural Defend's forensic media workspace. A trust and safety investigator uploads the disputed photo or video into a case. The analysis keeps model and version context and shows the original alongside explainability heatmaps and attention views, down to the image regions or video frames that carry signs of generation or editing. An LLM-assisted summary describes the findings in plain language linked to the evidence, and the investigator makes the final call.

Each case exports as a forensic report with observations, uncertainty, provenance, model context and analyst notes, plus structured results the marketplace can attach to the dispute record or share with a payment processor. For high-volume screening at upload, NeuroVerify can check images and selfies in one API call in under 200 ms at p95 on eligible image requests, and Neural Defend ranked first of 36 systems on the public VendorBench-100 benchmark of adversarial still images (MCC 0.876 against 0.646 for the next-best vendor).