The art of fraud: An expense audit perspective

In 1976, an unassuming German artist named Wolfgang Beltracchi began painting works “in the style of” long-dead masters like Max Ernst and Heinrich Campendonk. His forgeries were so convincing that collectors, curators, and even experts were fooled—until they weren’t.

For over 30 years, Beltracchi defrauded the art world of more than $45 million (£28.6M), though it’s estimated his total profits may have been well above $100M.

What does this have to do with corporate expense reports?

The spirit of Beltracchi lives on in the world of expense management—with fewer oil paints and more PDFs. Today’s forgers don’t need a studio; they need a smartphone, a template, and increasingly, a little help from AI.

At AppZen, we’ve seen a marked rise in receipt forgery, in the form of digitally manipulated or entirely fake receipts submitted by employees for reimbursement. In many cases, these aren’t just one-offs. They’re systemic. And they’re slipping through manual audits far more often than finance leaders would like to believe.

AI-generated fraud is a growing problem

Let’s start with the data. According to the Association of Certified Fraud Examiners (ACFE) 2024 Occupational Fraud Report:

• Total global losses from occupational fraud were $3.1 billion
• The median loss per case was $150,000
• A whopping 89% of cases involved misuse of company resources, a category that includes T&E spend like—you guessed it—fake receipts.

At AppZen, our Expense Audit product recently uncovered 348 fake receipts, for a total of $24,953 in potentially or likely fraudulent claims. The largest fake receipt in our data, quite brazenly, was $3,860. That’s an average of nearly 4 fake receipts for each company affected, or $274 per receipt. Small individually, but devastating in aggregate.

And that’s just what we caught last week.

The reality is that most finance teams aren’t looking closely enough, and those that are often lack cutting-edge solutions to keep up. And the pace of innovation is accelerating.

Employees can easily access AI tools like ChatGPT and dedicated receipt generators that whip up realistic-looking receipts in seconds. We’ve already seen some of them. Spoiler: We caught them.

Blake Oliver, CPA, and David Leary, hosts of The Accounting Podcast, demonstrate live how easy it is to create this convincing fake receipt with ChatGPT.

The new age of receipt forgery

Counterfeit art is more than a painted lie. It’s a test of authenticity, attribution, and detection. So, too, are fake receipts a test of internal controls.

Today’s receipt forgers can:

• Clone logos from hotels, airlines, or restaurants
• Manipulate amounts and dates using basic image-editing tools
• Use receipt generators or language models to create plausible narratives and layouts
• Submit screenshots that look real on casual review but fail metadata or formatting checks

But unlike museum curators, your average AP analyst or expense auditor doesn’t have time to examine every brushstroke or pixel. And that’s where AI shines.

AppZen Expense Audit uses proprietary AI to detect forgery patterns across receipts and catch metadata inconsistencies, template anomalies, and telltale spend behaviors across time. It can flag a forged Uber receipt in New York that’s suspiciously identical to another from Singapore, submitted three months prior. Or a breakfast at the Hilton that turns out to be completely made up.

How to scale fake receipt detection

So let’s say you’re a CFO or controller and you’ve just implemented automated expense auditing. What now? Who does what when the system detects a fake or suspicious pattern?

We advise our customers to put their fraud response into effect using a three-tiered approach:

1. Detection (automation + AI)

Fake receipts are caught automatically, then flagged with a confidence score and a clear rationale, such as “inconsistent font and logo alignment based on merchant standards.” Example alerts might include:

“Receipt image metadata mismatch”
“Same receipt hash used by multiple employees”
“Known fake template match”

2. Review and investigation (human in the loop)

This is usually where your expense auditor or fraud specialist comes in. If you don’t have one, start with your T&E team or AP analyst. Key questions to ask:

“Is this part of a pattern for this employee?”
“Are there similar issues across the team or department?”
“Has the employee submitted altered receipts before?”

Some companies use a centralized compliance team; others loop in HR, particularly for repeat offenders or high-risk roles.

3. Resolution and reporting

Once confirmed, it’s time to take action. Some options include:

• Reversing the reimbursement
• Employee monitoring
• Escalating to HR for disciplinary action
• Logging the case in your internal fraud tracker
• Feeding learnings back into your audit engine (yes, ours automatically learns from feedback!)

For AppZen customers, this operational flow is already built into the Employee Spend Trend Report, part of the dedicated set of models we call our Fraud Pack Apps, where we track anomalies over time. Not just individual receipts, but patterns of behavior.

AI-generated receipt fraud matters more than you think

The natural question you might ask is: Isn’t this a small problem?

If you look only at the dollars per incident, yes. But the reputation and cultural costs are far greater. When an employee submits a fake receipt and gets away with it, two things happen:

1. They’re more likely to do it again.
2. Others learn from them.

It creates a norm of dishonesty, not just a lapse in expense hygiene.
Additionally, today’s generative AI tools are massively scalable. What used to be a niche act of deception is now mainstream. There are entire Reddit threads and YouTube tutorials dedicated to receipt forging.

What good receipt fraud detection looks like

Our most successful customers treat fake receipt detection as an early warning system for fraud and cultural red flags. Here’s what they do:

Align Finance and HR. Ensure everyone is clear on policy, thresholds, and consequences. Does your expense policy explicitly allow (or prohibit) the submission of fake or manipulated receipts? This is important because some well-intentioned employees might submit “fake” receipts to substitute for a lost one. Intention matters. Pre-empt plausible deniability with clear policy.

Create playbooks. If a fake receipt is caught, who reviews it? Do we automate an email to compliance? Or to HR? Who communicates with the employee? Do we assume best intent? What’s the disciplinary path?

Measure trends. Is forgery increasing in a certain department or region?

Educate employees. Make it clear that digital audit tools are smart, and attempts to defraud the company will be caught (deterrent).

Automate. Most importantly, they don’t rely on humans alone. They use AI and automation to scale, learn, and stay ahead.

Don’t be fooled by forgeries

Just as Wolfgang Beltracchi could fool even the most experienced curators with the stroke of a brush, modern fraudsters can slip through expense systems with a few clicks. But they’re no match for AI that is always on the job, learning and adapting.

At AppZen, we believe financial integrity doesn’t come from tighter rules—it comes from smarter systems. Systems that learn from feedback and spot invisible patterns. So the next time someone tells you fraud is a finance problem, remind them instead that it’s an art form. One AppZen has mastered detecting.

Bonus: A tour of our collection

If you’ve read this far, you’ve earned a peek into our rogue’s gallery of recently exposed AI-generated fake receipts. For a live demo of our Fraud Pack Apps, contact us today.