All articles
Playbook9 min read

In-app purchase fraud with stolen cards is a chargeback your evidence can't fix, so stop it before you deliver

When a stranger buys your in-app purchase with a stolen card, the real cardholder's chargeback is legitimate and usage data won't reverse it. On Google Play that loss is yours from August 3, 2026. Here's how in-app purchase fraud reaches you, what each store tells you, and where to cut it off.

A credit card under a magnifying glass beside a smartphone showing a store page, standing in for in-app purchase fraud with stolen cards

Key takeaways

  • In-app purchase fraud with a stolen card ends in a chargeback filed by the real cardholder, who never made the purchase, so proof that your app delivered the item doesn't make the dispute illegitimate.
  • For Google Play orders placed after August 3, 2026, a lost chargeback costs the developer the purchase price less Play's service fee plus the bank's chargeback fee, while Google keeps covering only the service fee.
  • Google Play's Voided Purchases API separates true fraud from the rest: voidedReason 5 is Fraud, 6 is Friendly_fraud and 7 is Chargeback, and voidedSource tells you whether the user, the developer or Google voided it.
  • Google says it uses the obfuscatedAccountId and obfuscatedProfileId you pass at checkout to detect suspicious behavior and block some fraudulent transactions before they complete, such as many devices buying on one account in a short time.
  • When a Google Play purchase fails your own fraud checks, Google's guidance is to not grant or acknowledge it and to explicitly refund and revoke it, rather than letting the 3-day auto-refund decide.
  • Apple reports that the App Store stopped more than 5.4 million stolen credit cards and over $2.2 billion in potentially fraudulent transactions in 2025, and its revocationReason field has no fraud code, only 0 for other reasons and 1 for an issue in your app.

Stolen-card purchases are the one kind of chargeback where you can be right about everything and still lose. Your app delivered the coins. The player spent them. The logs are clean. But the person who owns the card never touched your app, and when they call their bank, their dispute is real. That's in-app purchase fraud in its plainest form, and since August 3, 2026, Google Play passes the cost of losing it to you.

The useful move isn't a better reply after the dispute. It's spotting the purchase before you hand over anything that costs you money, and both stores give you more to work with there than most teams use.

How in-app purchase fraud reaches your app

A fraudster doesn't attack your billing code. They attack the store's checkout with a card that isn't theirs, then use your app to turn it into something worth having. The pattern looks different from friendly fraud, where your own customer disputes a charge they made, and from refund abuse, where a real buyer gets the store to refund them again and again.

Why stolen cards end up in your store listing

The store's payment screen is the target, and your app is where the value comes out. Anything a buyer can move, trade, gift or resell is attractive: in-game currency, items that can be traded, account upgrades that make an account worth selling, or credits that pay another user. A plain subscription to a note-taking app is a poor prize. A pack of tradable gems is a good one.

What the cardholder's bank sees

The real owner spots a charge they didn't make and disputes it. From the bank's side, that's an unauthorized transaction and the cardholder is telling the truth. Nothing in your server logs changes who held the card.

Dispute typeWho files itDid they make the purchaseCan delivery evidence help
Stolen-card fraudThe real cardholderNoRarely, the dispute is genuine
Friendly fraudYour own customerYesYes, it shows they received and used it
Refund abuseYour own customerYesPartly, through usage data and history
Accidental purchaseYour own customer or a family memberYes, by mistakePartly, it's mostly a store judgment call

What a stolen-card chargeback costs you on Google Play

Google's Play Console Help page is plain about the split. For orders placed after August 3, 2026, Google Play and the developer share chargeback costs. You cover the purchase price less Play's service fee, plus the chargeback fees from the financial institution. Google covers the service fee for that transaction, and nothing beyond it.

Run it on one order to see the shape. Say a fraudster buys a $49.99 gem pack and your service fee on it is 15 percent. Google covers about $7.50. You give back the $42.49 you'd have kept, and you pay the bank's fee on top. Google doesn't publish that fee, because the bank sets it.

The costs that don't come back with the chargeback

That's only the store side. The money you already spent delivering the purchase stays spent, and fraud tends to hit exactly the products where delivery costs real money.

  • Payouts you already sent: if the gems paid a creator, a seller or another player who then cashed out, that payout left your account and the chargeback doesn't return it.
  • Compute you already ran: credits spent on generated images, video or model responses were paid for the moment they ran.
  • API calls you already paid for: third-party requests billed per call, from a maps lookup to a speech engine, fire before any dispute exists.
  • Storage you keep paying for: accounts a fraud ring builds up keep their files and history on your servers until you delete them.

What each store tells you about fraud

Neither store hands you the card details, and neither should. What they do give you is a reason code after the fact, and on Google Play, a way to feed your own account data into Google's fraud checks before the fact.

Google Play labels fraud in the Voided Purchases API

Every voided purchase carries a voidedReason and a voidedSource. The reasons are 0 Other, 1 Remorse, 2 Not_received, 3 Defective, 4 Accidental_purchase, 5 Fraud, 6 Friendly_fraud, 7 Chargeback and 8 Unacknowledged_purchase. The source is 0 User, 1 Developer or 2 Google. Two limits matter for fraud work. The API only returns voided purchases from the past 30 days, and the default quota is 6,000 queries a day and 30 in any 30-second window.

voidedReasonCodeWhat it tells youSensible response
Fraud5The purchase was flagged as fraudClaw back what's left, freeze trades and payouts on the account
Friendly_fraud6A real buyer disputed a purchase they madeClaw back, apply your strike policy
Chargeback7A bank forced the reversalClaw back, check the account for other open orders
Accidental_purchase4The buyer says it was a mistakeClaw back unused items, no penalty
Unacknowledged_purchase8You never acknowledged it within 3 daysFix your acknowledgement path, not the user

Apple doesn't give you a fraud code

Apple's revocationReason has two values. 0 means the App Store refunded the transaction for other reasons, such as an accidental purchase. 1 means it refunded because of an actual or perceived issue within your app. There's no separate value for fraud, so on iOS you can't tell a stolen-card reversal from an accidental one by the code alone.

What Apple does publish is the scale of its own screening. In its May 20, 2026 newsroom report, Apple said the App Store prevented over $2.2 billion in potentially fraudulent transactions in 2025, stopped more than 5.4 million stolen credit cards from being used, and banned nearly 2 million accounts from transacting again. Most stolen cards never reach your app. The ones that do are the ones that got past that.

A developer at a laptop at night reviewing rows of transactions with one row flagged in red, illustrating in-app purchase fraud review

How to stop fraudulent purchases before you deliver

Most of this is Google's own published guidance in its Fight fraud and abuse page. It's free, it runs on your server, and it moves the decision ahead of the money you'd otherwise spend.

Pass an obfuscated account ID on every purchase

Set obfuscatedAccountId, and obfuscatedProfileId if your app has profiles, when you launch the billing flow. Google says it uses this mapping between Google accounts and your in-app accounts to detect suspicious behavior and block some types of fraudulent transactions before they complete. Its own example is many devices making purchases on the same account in a short time. The ID is limited to 64 characters, and putting cleartext personal data like an email in it gets purchases blocked, so hash or encrypt it.

Check the purchase against the account before you grant

Verify every purchase token on your server with the Play Developer API before granting anything, and store each token so it can't be replayed. If you passed an obfuscated ID, confirm it matches the account that's asking for the item. Then run your own risk checks. Google's guidance is direct: if a purchase fails them, don't grant it and don't acknowledge it. Refund and revoke it explicitly, with orders.refund and revoke set to true for one-time products, or subscriptionsv2.revoke with a full refund for subscriptions. Letting it time out into the 3-day auto-refund works too, but Google calls that ambiguous, because it looks the same as a processing failure.

Delay the parts that cost you money

You can grant a purchase and still hold back what's expensive. A brand-new account that buys a large pack and immediately tries to gift it, trade it or cash it out is the pattern worth slowing down. A short hold on transfers and payouts from new buyers costs an honest customer a little patience. It costs a fraud ring the whole point of the purchase.

Claw back fast once the reversal lands

Google suggests clawing back unused items when a purchase is voided, setting a currency balance to negative if it was already spent, and limiting purchases until it's positive again. It also suggests calling the Voided Purchases API more often once you see voided orders, so you can claw back before the rest is spent. On Apple, the same job runs off the REFUND notification from App Store Server Notifications.

Should you answer the chargeback review at all

Sometimes Google sends a PendingRefundReviewNotification for a chargeback that needs developer review. You have 24 hours to answer with orders.reviewrefund, and Google records only your first call. Your refundPreference can be DECLINE, APPROVE or NEUTRAL, and you can attach consumption events with timestamps, the obfuscated account, an IP address and a coarse location.

On a dispute your own data says is real fraud, think before you choose DECLINE. The evidence fields show that someone used the item. They don't show the cardholder was that someone. Where you have strong signs of a stolen card, an honest NEUTRAL keeps your replies credible for the friendly-fraud cases where DECLINE actually belongs. Where usage clearly ties back to the cardholder's own long-standing account, send it and decline. That's the call the 24 hours are for.

Frequently asked questions

Who pays when someone uses a stolen card for an in-app purchase on Google Play?
For orders placed after August 3, 2026, the developer does if the cardholder's chargeback succeeds. Google's Play Console Help says the developer covers the purchase price less Play's service fee plus the bank's chargeback fee, and Google covers the service fee.
Can I win a chargeback filed by the real owner of a stolen card?
Usually not, because the dispute is genuine. Delivery and usage evidence prove your app worked, not that the cardholder made the purchase. Google's Review Refund API is built to help contest illegitimate chargebacks, and a stolen-card dispute isn't one.
How do I know a Google Play refund was fraud and not a normal refund?
Check voidedReason in the Voided Purchases API. Code 5 is Fraud, 6 is Friendly_fraud and 7 is Chargeback, and voidedSource shows whether the user, the developer or Google voided the purchase. The API only returns the past 30 days, so poll it regularly.
Does Apple tell developers when a refund was caused by fraud?
No. Apple's revocationReason has only two values, 0 for other reasons such as an accidental purchase and 1 for an actual or perceived issue within your app. There's no fraud value, so you have to rely on your own account signals on iOS.
Does setting obfuscatedAccountId reduce in-app purchase fraud?
Yes, Google says so directly. It uses the obfuscated account and profile IDs to map Google accounts to your in-app accounts, detect suspicious behavior and block some fraudulent transactions before they complete. Keep the ID under 64 characters and never put cleartext personal data in it.

Sources and further reading

RefundHalt

The refund autopilot for the App Store and Google Play

Keep reading

The next refund request is already on its way.

Set up RefundHalt in the time it takes to read another support email about a refund you didn't get to contest.