It’s not uncommon for a consumer to order the same shirt in three sizes, then return the two that don’t fit. But what happens when the consumer swaps the labels and returns different shirts, or buys a dress for an event, wears it once with the tag tucked in, then returns it?
These are examples of return fraud, a retail-industry risk that’s on the rise “in both scale and sophistication,” said Juan Hernandez-Campos, the chief operating officer at Happy Returns, a software company that manages returns and is part of UPS. Today, about 9% of returns are fraudulent, according to the 2025 Retail Returns Landscape report published by the National Retail Federation and Happy Returns last October.
With fraud a persistent problem amid growing return volumes, merchants can no longer rely on the “highly manual” methods they previously used to process returns, Campos said.
Instead, retailers and their software partners are turning to AI to help manage reverse logistics, the part of the supply chain that involves handling returns and restocking inventory. Narvar, which works with companies including DSW and Newell Brands, touts AI technology that processes billions of consumer data points to detect fraud. Meanwhile, Loop is using AI to automate decisions and unearth fraud for customers like Keen and Princess Polly.
Jackie Swanson, a managing partner at Gartner Consulting, said retailers have shifted over the last 18 months from machine-learning-enhanced fraud detection to AI systems. The goal? Stop fraud in its tracks, reduce losses, and speed up legitimate returns, getting inventory back to sale as quickly as possible.
Among retailers with more than $1 billion in annual revenue, “almost all of them have something in place,” Swanson said, with apparel, beauty, and footwear at the forefront due to their high return rates.
AI can tackle return fraud
Caroline Reppert, the senior director of AI and technology policy at the National Retail Federation, said reverse logistics is “emerging as a key focus area” as retailers consider how to use AI.
Retailers are contending with return fraud issues like overstated quantities — when a consumer claims they returned more items than they actually did — empty-box returns, or sending boxes filled with rocks, label tampering, and counterfeit substitutions. Some consumers engage in wardrobing, in which they use or wear an item for a short time before returning it.
One big advantage of AI, according to Campos, is that it can analyze large volumes of return data to identify patterns that could be tough to find manually. For example, Happy Returns — whose clients include Everlane, Pact, and Under Armour — uses a technology called Return Vision, which detects discrepancies in merchandise like incorrect logos, altered tags, differences in materials, or product swaps.
If Return Vision spots one of these issues, it feeds the data to retail dashboards in real time, allowing retailers to view potential problems quickly, Campos said. Retailers can see why the return was flagged and photos of the returned items. They then have evidence to dispute any consumer claims, even before the item reaches the warehouse.
Happy Returns also uses AI-based behavioral risk scoring, which identifies potentially high-risk returns based on return frequency, timing, geography, and historical behavior. The returns can be flagged for human review before retailers issue a refund.
That said, retailers reveal a mixed picture on AI for return fraud. In NRF’s report, only 45% of companies said they believe AI and machine learning are truly effective at preventing return fraud on their own.
Bully Max, which makes dog food and supplements, uses AI systems from Shopify and Chargeflow to detect return fraud, according to CEO Matthew Kinneman. If fraud occurs or a consumer files a chargeback, the AI automatically generates evidence on Bully Max’s behalf to submit to the credit card company.
That said, Kinneman acknowledged that AI isn’t perfect. The company has to manually send screenshots to the credit card company, for example. “We’ve found that human review is still essential,” he said.
A speedier return
Fraud isn’t the only use case for AI in reverse logistics. Bully Max also uses AI to spot patterns, like recurring return reasons. Then the team can improve a product’s page and description, providing more accurate information before consumers make a purchase.
“In many cases, preventing a return is more valuable than processing one efficiently,” Kinneman said.
Swanson said her clients are also using AI for predictive returns and warehousing decisions. In the former, AI models proactively flag orders that are likely to be returned and suggest alternate sizes or products before the consumer finalizes their purchase. In the latter, when a return arrives at a retailer’s warehouse, AI recommends whether the retailer should resell, liquidate, or even destroy the item based on its potential resale value.
Returns have historically been manual and time-consuming, Reppert told Business Insider. AI “allows retailers to handle higher return volumes with greater speed and consistency.”
Faster processing also benefits consumers by providing a quicker refund for returned merchandise, Campos said. A positive return experience can build brand loyalty and encourage consumers to make another purchase with the retailer, he added.
