AI

AI-Generated Fake Receipts Now Make Up 71% of Expense Fraud

Four in 10 U.S. employees have used artificial intelligence to generate a fake expense receipt, according to a survey released Tuesday (June 23) by expense management platform Emburse.

The survey, conducted by Atomik Research in the U.S. and U.K., found 19% of those employees fabricated a purchase that never happened, 15% inflated the value of a legitimate expense and 6% recreated a lost receipt using AI. Forty percent used company-funded AI tools to do so, according to a Tuesday press release on the survey findings.

The trend extends across the industry. Across AppZen’s enterprise platform, AI-generated receipts accounted for 0% of flagged fraudulent documents in March 2025. By mid-May 2026, that figure had reached 70.8%, based on 1,471 AI-generated fake receipts submitted by 745 employees at 174 companies and claiming a combined $148,143 in fabricated reimbursements, according to AppZen data cited by Accounting Today in a Tuesday report.

Price also sets artificial intelligence fakes apart. AI-generated fake receipts average roughly $100 each, compared to $182 for older template-based fakes, Accounting Today reported. The lower amounts appear intentional, small enough to slip under most companies’ auto-approval thresholds and through manual review without triggering scrutiny.

AI Receipt Generators Have Made Image-Based Fraud Checks Obsolete

Corporate expense management has long treated the receipt as proof of transaction. An employee submits a receipt, a finance team member reviews it and the claim is approved or flagged. That model worked when creating a convincing fake required photo editing skills or a paid online service. It no longer holds when a free AI image generator can produce a receipt with realistic paper texture, accurate itemization and a matching timestamp in under a minute.

“These receipts have become so good, we tell our customers, ‘Do not trust your eyes,’” Chris Juneau, senior vice president and head of product marketing at SAP Concur, told PYMNTS in an October 2025 report. SAP Concur processes more than 80 million compliance checks per month using AI. FinTech company Ramp flagged more than $1 million in fraudulent invoices within 90 days of deploying its AI-powered accounts payable tool.

Financial stress is also a driver. Emburse found that 51% of U.S. respondents incurred overdraft fees, late-payment charges or credit card interest while waiting for expense reimbursement. More than 1 in 4 U.S. employees admitted to passing off personal purchases as business expenses because of their financial situation. “For the most part, employees aren’t intentionally misusing corporate AI tools,” Emburse Chief Revenue Officer Michele Shepard said. “In many cases, they’ve been encouraged to adopt AI quickly and integrate it into their daily lives. The real challenge for employers isn’t bad behavior, it’s a lack of visibility.”

Matching Receipts to Card Records Is the Only Defense Left

The structural problem the Emburse data points to is that receipt-based controls were built to catch forgeries, not generations. A forged receipt has artifacts: inconsistent fonts, mismatched logos, pixelation around edited figures. An AI-generated receipt has none of those. It was never a real document to begin with, so there is nothing to detect at the image level.

That shifts the verification burden from the document to the transaction. Matching a submitted receipt against card transaction data, merchant category codes and payment timestamps can confirm whether a purchase actually occurred, regardless of how realistic the receipt looks.

The only viable response is proactive defense built on AI, Ernest Rolfson, CEO of Finexio, told PYMNTS in a December 2025 report. That means validating identities in real time, confirming payment instructions on every transaction and continuously checking against risk signals.

Emburse’s Shepard said the most effective strategy combines clear policies, faster reimbursements, virtual cards and AI-powered controls. Virtual cards link a specific payment to a specific transaction, making it impossible for a fabricated receipt to represent a purchase that a card record can contradict.