AI-Generated Fake Receipts Surge 70% in Corporate Expense Fraud

The economics of expense fraud have fundamentally shifted

Corporate expense fraud has entered a new phase. According to data from AppZen, a finance automation company, AI-generated fake receipts climbed from zero percent of detected fraudulent submissions in March 2025 to 70.8 percent by mid-May 2026. The shift happened in just 14 months, as reported by Forbes.

The catalyst is simple: generative AI has eliminated the practical barriers that once limited receipt fraud. Employees can now produce convincing fake receipts in seconds using free image generators, requiring no technical skill and leaving little visual evidence of manipulation. Traditional template-based fakes, which previously dominated and cost five to ten dollars from specialized websites, have fallen to roughly 29 percent of detected fraud cases.

AppZen detected 1,471 AI-generated receipts over the 12 months ending May 15, 2026. These submissions came from 745 employees across 174 companies and represented $148,143 in claimed reimbursements. While these figures reflect only detected cases and claimed amounts rather than actual losses, they signal a meaningful break from historical patterns.

Why it matters

This trend reveals how AI democratizes white-collar fraud by removing friction. When deception becomes effortless, more people attempt it. The shift from coordinated schemes to decentralized, individual fraud makes detection harder and threatens the trust-based systems many companies rely on for expense management.

Smaller amounts, higher volume

The dollar amounts tell a strategic story. AI-generated receipts in AppZen's dataset averaged $101, with a median of about $32. Older template-based fakes averaged $182. This represents a deliberate shift toward exploiting auto-approval thresholds that many finance departments use to process small claims quickly without human review.

"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," Kunal Verma, AppZen's CTO, told Forbes.

At one Fortune 10 company, 142 employees across 22 countries submitted 340 AI-generated receipts worth $34,953 in claimed expenses over 12 months. The pattern suggests organic spread rather than coordinated fraud. About one-third of employees caught using AI to fabricate receipts did so multiple times within the same period.

Visual inspection no longer works

Traditional fraud detection relied on visual cues: Does the receipt look authentic? Are the totals correct? Does the date make sense? AI-generated documents now replicate these markers of legitimacy. Some fraudsters add forged scanner watermarks or handwritten signatures to increase credibility. Others fabricate recurring bills from major providers like AT&T or Xfinity.

While OpenAI embeds metadata in generated images that can identify their origin, this information disappears when files are edited, compressed, or converted. The strongest warning signs often appear in numerical inconsistencies rather than visual flaws.

What companies must do differently

Finance teams need to verify transactions against multiple data sources rather than relying on document review alone. This means cross-checking receipts with card records, merchant information, travel details, and historical claim patterns.

Companies should also reassess auto-approval thresholds. While speed matters for operational efficiency, predictable thresholds create exploitable openings. Even automatically approved claims should be sampled regularly, with review rules updated as fraud tactics evolve.

The shift requires combining multiple authenticity checks rather than depending on any single signal. Digital provenance, numerical validation, pattern analysis, and random sampling must work together.

Details of this trend were first reported by James Broughel in Forbes, drawing on proprietary data from AppZen's expense auditing platform.