For years, one of the classic employee dishonesty claims has been deceptively simple: an employee submits a phony receipt, the company reimburses it, and the loss disappears into the ordinary flow of business. It might be a receipt for tires supposedly purchased for a company car but actually installed on a personal vehicle. Or it might be invoices from a vendor that looks legitimate but exists only on paper. In a large organization with hundreds of vehicles, thousands of employees, and a constant stream of invoices, these schemes can be difficult to spot.
Artificial intelligence is making that problem harder.
Recent reporting from PYMNTS, citing research from Emburse and AppZen, found that AI-generated fake receipts now account for a significant share of expense fraud. Emburse reported that four in ten U.S. employees surveyed admitted using AI to generate a fake expense receipt. Some fabricated purchases that never happened, others inflated legitimate expenses, and some used AI to recreate “lost” receipts. AppZen data cited in the same reporting found that AI-generated receipts grew from essentially none of its flagged fraudulent documents in March 2025 to about 70.8% by mid-May 2026.
The significance is not just that employees can cheat more easily. It is that traditional review methods are becoming obsolete. Historically, a forged receipt might contain clues: mismatched fonts, odd spacing, blurry logos, or signs of manual editing. AI-generated receipts can avoid many of those red flags. They can include realistic paper texture, plausible line items, merchant names, dates, and totals. In other words, the document may look right even though the transaction never occurred.
That changes the control question for employers. The issue is no longer simply, “Does the receipt look authentic?” The better question is, “Can we independently verify that the transaction happened?”
For companies, especially those with high expense volume, the answer likely requires a combination of policy, process, and technology. Matching receipts to corporate card data, merchant category codes, payment timestamps, approved vendors, and purchase histories is becoming more important than visual inspection. Virtual cards can also help because they tie a specific payment authorization to a specific vendor, amount, or purpose. AI-powered audit tools may become part of the answer as well, not because AI can always “see” a fake receipt, but because it can detect patterns humans may miss: repeat small-dollar claims, unusual merchants, duplicate submissions, weekend or after-hours activity, or employees whose expenses consistently fall just below approval thresholds.
There is also a human element. Emburse’s research suggests that financial pressure may be contributing to the problem, including employees waiting on reimbursements while carrying credit card interest, overdraft fees, or late-payment charges. Faster reimbursement cycles and clearer expense policies may reduce the temptation to misuse the system, while stronger controls reduce the opportunity.
From an insurance and risk-management perspective, this is a familiar crime problem with a new tool. Employee theft, false invoicing, and reimbursement fraud are not new. What is new is the speed, realism, and accessibility of AI-generated documentation. A company that still relies primarily on receipt images and manual review may be defending yesterday’s fraud.
The practical takeaway is straightforward: employers should treat receipts as supporting evidence, not conclusive proof. The real control is independent transaction verification. As AI lowers the cost of creating convincing false documents, companies will need to raise the quality of their internal controls.
And so it goes.
________________________________________
Sources
1. PYMNTS, “AI-Generated Fake Receipts Now Make Up 71% of Expense Fraud,” June 26, 2026:
2. Emburse, “Four in Ten Employees Have Used AI to Generate Fake Expense Receipts,” June 2026:
3. AppZen newsroom / expense auditing resources:
https://www.appzen.com/press-page/auditing-expenses
4. FRPA Fraud Viewer republication of PYMNTS article:
https://www.frpafraudviewer.org/aws/FRPA/pt/sd/news_article/629318/_PARENT/layout_details/false
