One Retailer Beat an AI-Faked Damage Claim With a Single FaceTime Call
A customer told Boll & Branch their bedding had arrived torn, and attached a photo to prove it. The photo had a problem: it still carried a visible AI watermark. Rather than process the refund, Modern Retail reports, the company's customer service team asked the customer to verify the damage over a live FaceTime call.
The customer never responded. The claim died there.
It is a small story with a large implication. As AI-fabricated evidence spreads through retail returns, one of the simplest defences turns out to be one of the oldest tricks in the book: ask for something the fraudster cannot fake in advance.
Why Did a FaceTime Request Beat an AI-Generated Photo?
Because it changed what the fraud had to survive. A photo, however convincing, is prepared offline with no time pressure and no one watching. A live video call has to happen now, in front of a real person, showing a real item, with no chance to touch it up or regenerate it if it looks wrong. The Boll & Branch customer had a photo ready to send. They had nothing ready for a camera pointed at the actual sheets in real time, and they simply disappeared.
That gap, between something a fraudster can manufacture ahead of time and something they have to produce live, is exactly where this kind of check earns its keep. It does not detect the fake. It just asks for evidence that cannot be faked the same way, and lets the fraudster's own absence do the detecting.
Is This an Isolated Case?
No, it is a symptom of something much bigger. Jordan Shamir, CEO of fraud detection company Yofi, described the shift bluntly: merchants went "from seeing it onesie-twosie to seeing it daily." What used to be an occasional convincing fake is now routine, across many merchants at once, using AI tools to fabricate damage photos, shipping receipts, and even police reports.
It has organised, too. Modern Retail describes emerging "returns-as-a-service" networks, coordinated groups that help individuals submit fraudulent claims in exchange for a cut of the proceeds, the same industrialisation of refund fraud we covered in You Can Now Hire a Professional to Commit Your Refund Fraud. The Boll & Branch case was one customer with one bad photo. The trend behind it is a supply chain for fabricated evidence, sold at scale.
Does This Mean Live Verification Solves the Problem?
Not on its own, and that is the important caveat. A FaceTime request works brilliantly on the individual who cannot improvise past it, but no support team can put every claim through a live video call. It does not scale, and even where it does work, it only ever tells you about the one claim in front of you. It says nothing about whether the same person tried the identical fabricated photo at nine other retailers this month.
That is the ceiling on any single-transaction defence, however clever. Catching one fraudster on camera is a win. Catching the pattern of someone doing this repeatedly, across many retailers, at the same time it happens rather than one FaceTime call at a time, is the difference between a good anecdote and an actual fix. We wrote about why AI-faked evidence makes that shift necessary, not optional, in AI Can Fake the Evidence Now.
What's the Real Lesson Here?
That the fight has moved from spotting a fake to demanding something synchronous, and that even a perfect synchronous check only closes one door. The Boll & Branch team did the smart thing in the moment: they refused to let a static, offline artifact settle the question, and asked for something happening live instead.
The retailers who cope with this wave will keep doing exactly that, and they will also stop treating each win as a solved case. One dodged FaceTime call is a fraud attempt stopped. The claimant trying the same photo somewhere else tomorrow is the problem that is still running.
Source: Modern Retail
Frequently asked questions
What happened with the Boll & Branch AI return fraud case?
A customer claimed torn bedding and submitted an AI-generated damage photo that still carried a visible AI watermark. Boll & Branch's customer service team asked the customer to verify the damage over a live FaceTime call. The customer never responded, and the fraudulent claim was shut down.
Why did asking for a FaceTime call work?
Because it demanded something synchronous, in the moment and impossible to prepare in advance, unlike a photo or document, which can be generated or edited ahead of time with no time pressure. A live call cannot be faked the same way a static image can.
Is AI-generated return fraud actually widespread?
Yes. Fraud detection company Yofi's CEO described the shift as going from isolated incidents to systematic abuse, seeing it daily across merchants rather than occasionally. Coordinated networks are also emerging, described as 'returns-as-a-service' operations that assist fraudulent claims for a cut of the profits.
Can retailers rely on live verification for every claim?
Not at scale. A live call works brilliantly against one determined fraudster but does not scale to thousands of daily claims, and it says nothing about whether this claimant is doing the same thing at other retailers. The lasting fix is recognising that broader pattern, not manually verifying every case.
Retail Cache · Fraud Intelligence
Retail Cache builds the shared fraud-intelligence network for retailers, carriers and 3PLs. We write about first-party, refund and delivery fraud, and how the industry can stop treating it as a cost of doing business.
Want to turn the tide on fraud, together?
See how Retail Cache helps retailers and carriers detect, prevent and stop fraud in real time.