A Pizza Burned Black and an Automatic Refund
On the morning of October 23, 2026, Rhonda Reynolds, who runs a pizzeria in Echo Park, Los Angeles, opened the store management screen for the delivery platform 'DoorDash' and ran into an unbelievable photo.
A whole pizza sat in its box, burnt black like a lump of charcoal. The customer account that sent it, 'Dani B,' even left a harsh complaint, saying the pizza was burnt to heck, totally black, and that they didn't wanna eat cancer.
Even more frustrating was how fast DoorDash moved. Instead of asking the shop whether any of it was true, the platform gave the customer an automatic refund and then took that loss out of the restaurant's payout as an 'error charge.'
Under DoorDash's policy, once a food-quality refund is filed, anywhere from 25% to a full 100% of the item price can land on the store.
✦ Order Mix Change at Masa of Echo Park
Share of Total Orders
This pizzeria has been going for 22 years, ever since Rhonda opened it with her husband Rob Rowe on Sunset Boulevard in 2004. It's a mom-and-pop restaurant that bakes Chicago-style 'deep-dish pizza.'
Before the COVID-19 pandemic, takeout and delivery made up only about 6% of all orders, but nowadays DoorDash orders alone make up about 40% of all orders, a sign of how big the platform has become.
When an absurd claim blows up on the channel you lean on for a big chunk of your orders, the stress must've been off the charts.
Something About This Photo Seems Off
Rhonda tilted her head the moment she saw the photo. The deep-dish pizza they make is built nothing like a regular pizza with thin dough and a light layer of cheese.
Thick dough gets stuffed with layers of cheese, sausage and mushroom toppings, and watery tomato sauce, so the thickness alone reaches about 5 centimeters (2 inches).
It takes a slow 40 to 45 minutes in a rotary oven to cook all the way through, and getting it as charred as in the photo would take leaving it in the oven for hours.
The shop records made her suspicion even stronger. A small traditional deep-dish pizza ordered at 4:39 p.m. on October 22, the day before, left the store just 56 minutes after the order, already cooked and packed.
With a normal cook time of 40 to 45 minutes, that leaves only ten-odd minutes. The math didn't add up for a thick pizza topped with moist sauce to burn black in that time.
Zed Cutsinger, a veteran who's watched about 100,000 deep-dish pizzas come out of this one store over 15 years, zoomed in on the photo and noticed something strange.
The crust edge in the photo was black as charcoal, but the sausage topping right inside it was a mouthwatering brown. On the flip side, a jet-black piece was stuck on next to smoothly baked dough, cut off along a curved boundary.
Looking at the photo, Zed pointed out, "The AI couldn't reproduce how flour dough and meat behave and the different temperatures at which each one burns, the way real cooking would, so it left an unnatural boundary."
That was a call as sharp as Google's SynthID Detector, the tool that checks whether an image is AI-generated.
Yep, This Is an AI-Generated Image!
Zed gave an opinion, but to be sure, Rhonda uploaded the photo to an AI image detection app called 'Is This AI.' The screen showed the number '96% chance of AI generation.'
But that percentage is only an estimate, so that number alone couldn't serve as hard evidence to overturn the platform's payout deduction.
✦ Contradictions in the Fake Photo and the Real Evidence
Photo vs. Kitchen Records
The decisive counterattack came from inside the kitchen. Jose Tapa, who handles packing, rewound the kitchen CCTV to the time of the order, and the familiar pizza box showed up clear as day on screen.
The outside of the box had Dani B's name and the order number on it, and the footage recorded the whole scene of a perfectly fine, golden-baked pizza being packed inside and going out the door.
✦ Incident Timeline
October 2026, Los Angeles local time
DoorDash keeps a process open that lets stores file a dispute through its portal within 14 days of an order being completed.
Rhonda bundled the 'Is This AI' result together with the kitchen CCTV footage, the box label records and the order and dispatch timestamps, and submitted it all to DoorDash.
After reviewing the materials, DoorDash reversed the original refund, and the customer account behind the false claim was deactivated for fraud.
Luckily, this case had CCTV as hard proof that it was fake, but if there'd been no CCTV and she'd asked DoorDash to reverse the refund based only on an AI detection app result or a visual check, it wouldn't have been easy to get accepted.
From Chicken Wings to Bedding and Tote Bags, AI Refund Fraud Is Everywhere
✦ AI Photo Faking: 3 Cases Spreading
From Chicken Wings to Bedding and Tote Bags
Cases of AI photo manipulation aimed at delivery apps have been coming to light for months already. In May 2026, there was the TikTok user '@mi5under5t00d.'
This person took a photo of a half-eaten 'chicken wing' and used ChatGPT's image feature to turn it into 'bloody raw chicken.' Then they got a $39.24 credit refund from DoorDash and uploaded the video to TikTok.
The video racked up 4.4 million views and stirred up controversy.
DoorDash said at the time that it was running 'behavioral modeling' and 'manual review of high-risk accounts,' and that AI-powered fraud is still 'fraud,' so accounts caught doing it get permanently banned.
✦ AI Refund Fraud Through Forter's Eyes
Fraud Prevention Company Forter
Similar tricks are spreading fast across retail and e-commerce in general. According to the fraud prevention company Forter, 'refund claims backed by AI-manipulated images' have grown by more than 15% worldwide since early 2025.
Forter blocked organized return fraud attempts during the 2025 year-end shopping season, and those rings had used AI image features to make products look damaged and tried to pull off $1.5 million in fraud.
Cases from the front lines are getting bolder too. Jordan Shamir, CEO of the fraud prevention company Yofi, said, "You used to see it now and then, but AI refund fraud is now something we witness every single day."
Scott Tannen, CEO of the bedding brand Boll & Branch, once got an absurd refund request in which the photo of a torn sheet still had an AI watermark stamped on it. He asked to confirm the damage over a video call, but the customer never responded.
The bag brand Bogg has also been getting a steady stream of tote bag photos with tears faked in, so it changed its process: suspicious defect reports now get refunded only after the item is retrieved.
So Small Business Owners Have No Choice but to Collect Evidence?
In this pizzeria case, what overturned the platform's call was the kitchen CCTV, the order number written on the box, and the cook time of a deep-dish pizza that has to bake for over 40 minutes. I can't say how it would've gone without evidence this perfect.
Rhonda said that in 22 years of running the shop, this was a first. She worried that small businesses can't hang on if, on top of protecting their staff's honor and the shop's reputation, they have to spend all their time sorting out fraudsters.
At least this case was handled by a business owner who understands AI images, but a shop that knew next to nothing about AI technology would've been taken for a ride with no way to fight back.
I figure AI-powered fraud will only keep growing. Like it or not, all of us now need to understand AI technology better and build up the power of knowledge so we don't become targets of crime.
Of course, for small business owners... for now, securing hard evidence like CCTV is the best bet. Even Google's SynthID Detector isn't 100% accurate, so getting an AI image detection app's result accepted doesn't look easy yet.
That's all for this post. See you next time.
A Fake £154 Million Takeover Bid, Made With ChatGPT
AI Scam Gang Posed as CEOs and Stole $129 Million
ChatGPT Lawsuit: "Cut Your Dose, Hide It From Your Doctor"
🌐 Korean version: “완전히 타서 못 먹겠다” AI 가짜 사진으로 피자 환불받은 손님... 피자집이 가짜 이미지를 잡아낸 방법은?
🌐 Japanese version: 「真っ黒で食べられない」AI偽写真で返金?ピザ店が見抜いた方法
No comments:
No comments yet. Want to leave the first one?
Post a Comment