The era of fake expense receipts purchased from sketchy template websites is being overtaken by something far more scalable: AI image generators that can produce convincing receipts in seconds.
Data from AppZen, an expense-auditing firm that uses AI to automate accounts payable and expense management for corporate finance departments, reveals a dramatic shift. The share of fake receipts its systems identified as AI-generated rose from 0% in March 2025 to 70.8% by mid-May 2026. During the 12 months ending May 15, 2026, AppZen detected 1,471 AI-generated receipts submitted by 745 employees at 174 companies, representing $148,143 in claimed expenses.
About one-third of those employees submitted AI-generated receipts more than once. At one Fortune 10 company, 142 employees across 22 countries submitted 340 AI-generated receipts representing $34,953 in claims.
The tactic is targeted and deliberate. The median value of AI-generated receipts was just $32 — a figure consistent with attempts to stay below corporate auto-approval thresholds, where no human reviewer ever sees the claim. By contrast, receipts created with older template-based fake services averaged $182.
AI basically flipped the game from one fake big enough to be worth the risk to a pile of tiny ones nobody bothers to review, AppZen CTO Kunal Verma told E-Commerce Times. It is a pretty direct sign that auto-approval thresholds, meant to save reviewers time, have become the thing people are gaming.
The fabricated documents extend beyond restaurant and travel receipts. AppZen detected AI-generated AT&T and Xfinity monthly bills submitted to support telecom expense claims at one US company. At a European industrial company, 29 employees submitted AI-generated receipts across 31 expense reports.
Some fraud attempts have grown more sophisticated. In one case, a submitted restaurant receipt included a forged CamScanner watermark and handwritten signature designed to make the document appear to have been captured by a scanner app.
The shift has forced a corresponding evolution in detection. AppZen says visual inspection alone is no longer reliable, and the company now relies on layered AI models that cross-reference receipts against merchant patterns, transaction histories, peer spending behaviour and internal consistency checks such as whether tax percentages add up correctly.
An image may look believable but contain an incorrect tax percentage or numbers that do not add up properly, Verma said. Looking harder does not help anymore.
The findings suggest that AI-powered fraud tools are creating an arms race within corporate finance, where both the attacks and the defences are now increasingly AI-driven.
