Copy article

Retailers Use AI to Fight Back Against Profit-Wrecking Returns Scams

ended 22. December 2025

UPS isn't the only one using silicon brains to stop shoppers from swapping £300 worth of genuine new boots for a fake pair. With return fraud expected to cost US retailers over £60 billion this year, and Retail Times quoting £11bn in the UK, with £6.6bn of that coming from just 11% of shoppers, a pack of tech firms is deploying AI workflows to gatekeep the returns bar.

​The AI Fraud Fighters

  • ​Loop Returns: Their machine learning classification model automatically flags returns that mirror past "bad actor" behaviour. Merchants see a "high-risk" shield icon before they even process the refund.
  • ​Narvar: Using their IRIS™ AI layer, they process 74 billion interactions to link identities across touchpoints. They claim to block up to 18% of fraudulent return refunds by enforcing strict eligibility in real-time.
  • ​Signifyd: Their ML model scores return requests instantly. "Trusted" customers get immediate refunds, while "riskier" ones face extra verification or manual audits. They even offer a financial guarantee, covering losses if their AI gets it wrong.
  • ​Riskified: Their Policy Protect tool uses "identity clustering" to spot serial abusers. They’ve recently added "agentic" fraud detection to stop AI shopping bots from gaming the system.
  • ​Forter: They leverage a Global Merchant Network to identify abusers across multiple brands. If a fraudster hits one shop, they're flagged across the entire network the moment they try to initiate a return elsewhere.

​The Numbers

​Retailers are desperate because 9 to 10% of all returns are now estimated to be fraudulent, according to the Retail times.

We want your views:

  • Where's the line between fraud prevention and customer humiliation? How many false positives are acceptable before "smart security" becomes "guilty until proven innocent"?
  • How with fraudsters manipulate the system next?

3 responses from the Newspage community

Copy all

Star Quote
Copy

Colette Mason
AI Ethics Consultant at Clever Clogs AI
The free-for-all is over. If you’re planning to "rent" a designer dress for New Year’s Eve, the robots are already onto you.
Copy

Editor
Kate Underwood
Founder & Chief People Strategist at Kate Underwood HR and Training
AI shouldn’t make honest customers feel like criminals at the returns desk.
I get why retailers are doing this. Returns fraud is brutal. It kills margins, clogs up ops, and dumps stress on customer service teams who take the abuse when something’s flagged. But the line is clear: fraud prevention is protecting the business; customer humiliation is punishing someone before you’ve proved anything. If your system blocks a refund with a cold “not eligible” and no explanation or easy appeal, you’ve basically told a decent customer, “We think you’re dodgy.” They won’t forget it. How many false positives are acceptable? Very few. One wrong call can lose a loyal customer and create legal risk if your model disproportionately flags certain groups and you can’t justify it under UK GDPR. What’s next from fraudsters? They’ll “train” your AI by behaving perfectly on small orders, then hit big-ticket returns, swap scams, account takeovers and bot-driven return rings spread across multiple ident
Copy

Rohit Parmar-Mistry
Founder at Pattrn Data
We are sleepwalking into a scenario where legitimate customers are locked out of the online economy by algorithms they can’t see, understand, or challenge. This isn’t just fraud prevention; it is automated discrimination.

AI models rely on historical data, which is inherently biased. If you live in a "high-risk" postcode or your shopping habits deviate slightly from the statistical norm, you get flagged. Meanwhile, professional fraudsters will simply use their own AI to mimic "perfect" behaviour and bypass these checks entirely. The only people getting caught in the net are regular humans behaving like humans.

The real danger lies in shared networks. If a bias flags you at one retailer, you could be silently blacklisted by dozens of others using the same provider. You become "un-shoppable" with zero recourse. That isn’t efficiency; it is lazy corporate overreach that punishes the innocent to save a few percentage points on the bottom line.