How to prevent return fraud
A practical 2026 playbook for online stores: the fraud types that actually matter, the red flags that give them away, and how to stop a bad-faith refund before the money's gone.
Last updated: August 2026
Return fraud is when someone works your return or refund policy to walk off with money or goods they were never owed. A "never arrived" claim on a package that tracking says was delivered. An empty box mailed back for a full refund. A dress worn once and returned with the tags tucked back in. It's quiet, it looks exactly like normal support, and in 2026 it found a brand new door: the AI agent you just put in charge of refunds will approve most of it without blinking.
Below I'll go through the fraud types that actually matter, the red flags that give them away, and how to stop a bad-faith refund before the money's gone. That last part includes a gap almost nobody is talking about yet.
How much does return fraud actually cost?
More than most owners think, and it keeps climbing. In 2024, Appriss Retail's annual returns report put fraudulent returns and claims at around $103 billion, roughly 15% of all returns. Wardrobing on its own, where a shopper buys, uses, and sends it back, hit around 60% of retailers that year. If you're doing real volume, this isn't a rounding error. It's a number you can pull down.
The tricky part is that almost none of it looks like fraud. It shows up as a friendly little message: "Hi, my order never came, can I get a refund?" And the faster you clear those, the more of them you wave through on autopilot.
The main types of return fraud
Most of it boils down to a handful of patterns. Learn the tell for each one and you've already caught the bulk of it.
| Type | What it is | The tell |
|---|---|---|
| Wardrobing / bracketing | Buying to use once, or buying several sizes and returning the rest | High return rate on wear-once categories, or multiple sizes of one item |
| "Item not received" (INR) | Claiming a delivered package never showed up | Tracking says delivered, and the same buyer has done it before |
| Empty-box / wrong-item return | Sending back an empty box, a brick, or a cheaper item | Weight or contents don't match at inspection |
| Serial returning | A customer whose returns dwarf what they keep | Three or more returns in 30 days, or returns from several addresses |
| Price switching / receipt fraud | Swapping tags or returning with a doctored receipt | The refund asked for doesn't match the order record |
| Friendly fraud / chargebacks | Getting the refund, then disputing the charge too | A chargeback on an order you already refunded |
How to prevent return fraud
There's no one rule that stops all of it. What works is a few cheap layers stacked on top of each other.
1. Tighten the policy so it stops being a loophole. Set a clear return window, and shorten it for seasonal or high-abuse categories. Make genuinely final items final sale. Charge a restocking fee where it fits. Ask for the item back in original condition, and say so in plain words. A vague policy is basically an open invitation.
2. Watch the data. Pull the last twelve months and work out your return rate per customer and per product. The outliers jump out fast: the buyer who returns way more than they keep, a sudden run of "never arrived" claims, a product with a damage rate that makes no sense. Set a threshold, say more than three returns in thirty days, that flags an account for a second look instead of an automatic refund.
3. Ask for proof before you pay. For damage or wrong-item claims, get a photo before any label or refund goes out. For "never arrived," check the tracking first. If it says delivered, that's a carrier claim and possibly a police report, not a same-day refund.
4. Verify the payment at checkout. Address Verification (AVS) and 3-D Secure knock down a lot of the card-side fraud that comes back to bite you as a chargeback weeks later.
5. Don't let one order pay out twice. A refund followed by a chargeback on the same order is one of the most common ways money leaves your account twice. Keep a record of what you've already refunded so the second attempt gets caught.
The gap nobody's talking about: your AI support agent
Here's what changed. Every guide ranking above this one was written for a world where a person read the "my order never came" message and used their own judgment. In 2026, more and more of those messages get handled by an AI agent instead. And most AI support agents are built to be agreeable. Ask nicely, they refund you. Whatever safety check they run, if they run one, is usually just a second AI deciding whether the reply "looks right." A confident, well-written scam looks right.
So the front line of return fraud has quietly moved. It isn't your warehouse or your policy page anymore. It's the automated agent you just handed your refund button to. Automating your support without automating the fraud defense is how a store cuts its response time and grows its fraud losses at the same time, and never even notices.
That's the whole reason Resolvas is built the way it is. Every rule up above lives in code, not in the model's head: the return window, the delivered-but-lost check, the photo before a refund, the risk flags, the "don't pay twice" guard. The AI drafts a resolution and suggests it. A separate policy engine decides whether that resolution is actually allowed to move money, and it says no by default. Nobody can sweet-talk it into a refund your policy doesn't allow, because the part that signs off on the money isn't the part having the conversation. You can watch it work through real cases in our playbooks and the worked ticket library.
Automate the boring stuff all day. Just don't automate away the one call that actually costs you money.
Return fraud FAQ
What's the difference between return fraud and refund fraud?
Return fraud abuses the returns process itself: sending back the wrong item, wardrobing, a fake "never arrived." Refund fraud is the wider umbrella for any scheme to pull money you don't owe, chargebacks and doctored receipts included. In real life they blur together, and the same defenses cover both.
Is bracketing illegal?
Usually not, which is exactly why it's so common and so expensive. Buying five sizes and keeping one sits inside most return policies. You push back on it with policy (windows, restocking fees, final sale) and with data, not with lawyers.
How do I spot a serial returner?
Their return rate gives them away. Someone sending back three or more orders in a month, or returning far more than they keep, or shipping returns in from a few different addresses, is worth a longer look before the next refund clears.
Can AI actually stop return fraud, or does it make it worse?
Depends entirely on the AI. An agent built only to keep the customer happy will approve more fraudulent claims, and quicker. An agent whose money actions are gated by real rules (return window, evidence, risk, duplicate-refund checks) stops the bad-faith ones and still clears the honest ones on the spot.
Should I just make everything non-returnable?
Please don't. Punishing every honest customer to catch a handful of bad ones costs you far more in lost sales and goodwill than the fraud ever will. The job is to clear real requests fast and stop only the fake ones, which is what targeted rules do and blanket bans can't.
What's the single fastest thing I can do today?
Pull your return data and find your top five customers by return rate. That one report tends to expose most of the abuse in a store, and it tells you exactly where to tighten first.
Run a Shopify store and sick of choosing between fast support and safe refunds? See how Resolvas closes tickets without giving fraud a free pass.