AI-generated receipts now account for 70.8% of flagged fraudulent expense submissions, up from 0% in March 2025, according to platform data from expense-management vendor AppZen reported by Accounting Today. The figure is based on 1,471 AI-generated fake receipts submitted by 745 employees across 174 companies, claiming $148,143 in fabricated reimbursements. It describes a rapid shift in how already-flagged expense fraud is being committed, not a new estimate of total workplace fraud.
A separate employee survey points in the same direction. Expense platform Emburse, citing a survey of 2,000 U.S. and U.K. workers conducted by Atomik Research between May 5 and May 8, 2026, found that 40% of U.S. respondents and 29% of U.K. respondents admitted to using AI to generate a fake business-expense receipt, according to PYMNTS. Among those workers, 40% said they used company-funded AI tools, while 9% said they built their own tools for the purpose.
For small-business finance teams, the operational consequence is simple: a receipt image is no longer reliable proof of a purchase on its own. That does not mean most expense claims are fraudulent, or that AI is the only source of expense abuse. It means that among the fraudulent receipts already being detected, AI-generated documents have become the dominant format, and controls built around visual review alone are becoming less dependable.
What the 71% Receipt-Fraud Figure Actually Shows
The 70.8% figure is a share of flagged fraudulent receipts, not a share of all expense submissions or a confirmed fraud rate across the economy. AppZen’s data reflects what its own platform caught and classified within a specific detection pipeline covering 174 companies between March 2025 and mid-May 2026. The sample does not represent every expense platform, every employer, or every reimbursement process.
The more important signal is the speed of the change. Accounting Today’s reporting on the AppZen data notes that AI-generated fakes first exceeded template-based fakes in April 2026. Before that crossover, receipts copied or altered from online templates accounted for most flagged fakes. The documented shift suggests the fraud method changed quickly once widely available generative tools made plausible receipt creation cheap and easy.
The Emburse survey adds a different layer because it captures self-reported behavior rather than platform detections. Among workers who admitted using AI to generate a fake receipt, 19% said they fabricated a purchase that never happened, 15% said they inflated the value of a legitimate expense, and 6% said they recreated a genuinely lost receipt. Those are different risk profiles. A finance team that treats all three as identical may miss the distinction between deliberate theft, value inflation, and an employee trying to replace missing documentation.
AI Receipts Undermine the Old Visual-Review Workflow
A traditional reimbursement process assumes the receipt image is a faithful record of a real transaction. An employee uploads a document, a manager or bookkeeper checks the amount and category, the claim is approved, reimbursement is issued, and the receipt is retained for tax and audit purposes. AI-generated fakes weaken that chain at the first step, before a reviewer has any obvious reason to question the document.
Older receipt fraud often left clues, such as mismatched fonts, suspicious templates, or visibly altered totals. Modern AI image tools can produce realistic textures, plausible itemization, and merchant-style layouts without requiring the employee to know how to use photo-editing software. That lowers the skill barrier and makes smaller, more routine claims easier to fabricate.
The pattern also appears designed to avoid attention. AppZen’s data puts AI-generated fakes at roughly $100 on average, with a median of $32, compared with an average of $182 for older template-based fakes. That points toward high-volume, low-dollar claims that can sit below auto-approval thresholds and avoid the scrutiny normally reserved for large reimbursements.
Small Finance Teams Have Less Room for Manual Verification
Large organizations may have dedicated audit staff, tiered approvals, and fraud-detection software sitting on top of their expense platforms. Small businesses often do not. A single owner, office manager, or bookkeeper may review reimbursements while also handling payroll, invoices, scheduling, and vendor payments, leaving little time to investigate a $32 lunch receipt or a $100 travel claim.
Shared corporate cards, outsourced bookkeeping, and informal approval chains can compound the exposure. If no one is reviewing claims across time, repeated low-dollar submissions from the same employee may look harmless in isolation while forming a larger pattern. The AppZen sample’s average fabricated claim size illustrates the kind of claim that may receive limited scrutiny in a small workplace.
The accounting consequences can outlast the reimbursement itself. A fake or misclassified receipt can affect tax records, job-cost reporting, client billing, and financial statements. Correcting the error later requires time that small teams may not have, especially when expense review is handled manually or after the fact.
Small Businesses Should Tighten Receipt Controls Before the Next Reimbursement Cycle
The AppZen and Emburse data point toward a practical conclusion: visual inspection of a receipt image is no longer enough. Small businesses can respond without building an enterprise audit department by adding checks that rely on independent transaction data and repeat-pattern review.
- Match receipts to card transactions. Reconcile submitted receipts against corporate card, debit card, or bank records so the image is checked against an independent payment trail.
- Review repeated low-dollar claims. Because AI-generated fakes in the AppZen data cluster around low dollar amounts, look for patterns of small claims from the same employee rather than focusing only on large outliers.
- Verify merchant details independently. Confirm vendor names, addresses, tax IDs, or merchant records when a claim involves a new vendor, an unusual category, or an amount just below an approval threshold.
- Set a written AI-use policy for company tools. Since 40% of admitted AI receipt generators in the Emburse survey used company-funded tools, employers should explicitly prohibit using licensed AI tools to create or alter expense documentation.
- Route unusual submissions to a second reviewer. Claims involving new vendors, repetitive categories, missing payment records, or odd timing should receive an additional check before reimbursement.
- Keep an audit trail of approvals and exceptions. Document who approved each claim, what evidence was reviewed, and why any exception was allowed, so the business has a record if questions arise later.
The goal is not to treat every employee as a fraud risk. It is to update a control environment built for an older kind of receipt manipulation. Transaction matching, merchant cross-checking, and review of repeated behavior are better suited to AI-native forgeries than visual inspection alone.
Indicators to Watch as Synthetic Expense Fraud Expands
The AppZen and Emburse figures describe conditions through mid-May 2026 within specific samples. Several developments would help show whether the trend is broadening, leveling off, or changing form.
- Disclosures from other expense platforms. AppZen’s 70.8% figure comes from one vendor’s data. Comparable reports from other providers would help show whether the pattern is platform-specific or industry-wide.
- Changes in average claim size. If AI-generated claims move from low-dollar receipts toward larger reimbursements, the risk profile for employers would change.
- Tools that combine metadata, merchant records, and transaction matching. Detection methods that look beyond the receipt image are likely to become more important as synthetic documents improve.
- Employer-funded AI usage in fraud cases. The Emburse survey found that 40% of workers who generated fake receipts used company-funded tools. Future data may show whether workplace AI licenses are becoming part of the fraud pathway.
- Policy updates from expense and accounting vendors. New controls, disclosure language, and AI-detection settings may signal how quickly the industry is adapting to synthetic receipt risk.
Receipt Verification Is Now a Finance-Control Issue
The evidence supports a narrow but significant conclusion: among fraudulent receipts caught by one expense-management platform, AI-generated documents rose from 0% in March 2025 to 70.8% by mid-May 2026. It does not establish that most expense claims are fraudulent, or that AI has replaced every older form of expense manipulation.
For small businesses, the practical takeaway is that receipt images can no longer serve as the primary evidence in an expense-approval workflow. Finance teams should first understand what the underlying fraud flag represents, then adjust controls so approval depends on more than the appearance of a document. Independent transaction records, merchant verification, repeated-claim review, and clear AI-use policies are now part of basic expense governance rather than optional fraud-prevention extras.
