The AI Document Fraud Explosion: $8.9 Billion in Exposure Across U.S. Industries

Every business runs on documents it assumes are real. Invoices, bank statements, pay stubs, IDs, utility bills, expense receipts: the entire machinery of commerce depends on the quiet assumption that the paperwork in front of you is what it claims to be. That assumption is now being industrialised against employers, and a new generation of generative-AI tools has made forging a convincing document faster, cheaper, and harder to catch than at any point in history.

To understand the scale of what that shift is worth, we modelled the total document-fraud loss sitting across US private employers, then isolated the share now driven by AI-generated documents, using fraud-detection network data from Inscribe and Resistant AI, the ACFE and AFP fraud surveys, and BLS and Census employment data. The result is a ranked index any business owner or finance team can use to see where they stand. And while the biggest dollar totals land roughly where you would expect, the data reveals a far more uncomfortable picture of who is actually most at risk.

Key findings

• US businesses lose an estimated $178.0 billion every year to document fraud across all document types, under the combined volume, fraud-rate, and loss-realisation model.
• In just 18 months, AI-generated document fraud has grown from virtually zero to an $8.9 billion problem.
• AI-generated document fraud grew 5x in eight months (April to December 2025) across the Inscribe detection network alone.
• Across all 6.27 million US employer firms, document fraud represents an average annual exposure of $28,374 per business.
• Invoices are the single most-forged document type by dollar loss, driving $126 billion, more than 70% of the national total, followed by bank statements and pay stubs.
• Health care carries the largest AI-generated fraud exposure of any industry at $2.07 billion (23.3% of the national AI-gen total), out of $41.5 billion in total document-fraud exposure, ahead of government and benefits ($1.73 billion AI-gen) and manufacturing ($1.40 billion AI-gen).
• The five most-exposed industries account for $6.85 billion of the AI-generated fraud subset between them, roughly 77% of the national AI-gen total.
• Large enterprises carry the biggest aggregate dollar exposure, but small firms feel each loss far more sharply relative to their size, and AI is eroding the detection advantage big firms once relied on.

The industries most exposed to AI document fraud

1. Health care - $2.07 billion in AI-generated fraud exposure

• Total document-fraud exposure: $41.5 billion
• Estimated AI-generated fraud subset: $2.07 billion

Health care tops the index on the strength of sheer document intensity. It employs more workers than any other private sector in the study, and its operations are saturated with paperwork (insurance claims, billing, patient records) that is routinely processed at speed and under pressure.

Why it ranks here:

• A workforce of roughly 23.7 million, the largest employer base in the study
• Association of Certified Fraud Examiners (ACFE) data shows the median fraud case results in losses of around $100,000
• Rated moderate AI risk, with 61% of fraud schemes being document-driven
• Primary exposure across insurance claims, billing, and patient documents
• 23.3% of both the national total and the AI-generated subset, more than any other industry

2. Government and benefits - $1.73 billion in AI-generated fraud exposure

• Total document-fraud exposure: $34.6 billion
• Estimated AI-generated fraud subset: $1.73 billion

Benefit and entitlement systems run almost entirely on submitted documents (applications, tax records, ID, and address proofs), making them a natural target for fabricated paperwork. The scale of the workforce processing these claims places the sector second.

Why it ranks here:

• A workforce of roughly 23.3 million
• An ACFE median fraud-case loss of $100,000
• Rated moderate AI risk, with 52% of fraud schemes being document-driven
• Heavy reliance on benefit applications, tax filings, ID, and address proofs
• 19.4% of the national total and of the AI-generated subset

3. Manufacturing - $1.40 billion in AI-generated fraud exposure

• Total document-fraud exposure: $28.0 billion
• Estimated AI-generated fraud subset: $1.40 billion

Manufacturing pairs a large frontline workforce with a high-value vendor and invoicing pipeline. The combination of invoices, vendor documents, and expense reports flowing through multi-site operations is exactly the structure that makes consistent verification hardest.

Why it ranks here:

• A workforce of roughly 12.6 million
• The highest ACFE median fraud-case loss in the top five at $170,000
• Rated low AI risk, with 46% of fraud schemes being document-driven
• Primary exposure across invoices, vendor documents, and expense reports
• 15.7% of the national total and of the AI-generated subset

4. Professional services - $0.97 billion in AI-generated fraud exposure

• Total document-fraud exposure: $19.4 billion
• Estimated AI-generated fraud subset: $0.97 billion

The first pure white-collar entry in the top five, professional services lands here on a mix of high case values and document-heavy workflows. Invoices, expense reports, and contracts move constantly between firms, and each is a potential vector.

Why it ranks here:

• A workforce of roughly 10.8 million
• An ACFE median fraud-case loss of $150,000
• Rated low AI risk, with 42% of fraud schemes being document-driven
• Primary exposure across invoices, expense reports, and contracts
• 10.9% of the national total and of the AI-generated subset

5. Retail and e-commerce - $0.67 billion in AI-generated fraud exposure

• Total document-fraud exposure: $13.5 billion
• Estimated AI-generated fraud subset: $0.67 billion

Retail's fifth-place finish is driven by scale and the volume of customer-facing documents it handles. Returns documentation, account openings, and identity checks all create openings for fabricated records, even at a relatively modest median case loss.

Why it ranks here:

• A workforce of roughly 15.4 million
• A lower ACFE median fraud-case loss of $59,000
• Rated moderate AI risk, with 52% of fraud schemes being document-driven
• Primary exposure across returns documents, account openings, and IDs
• 7.6% of the national total and of the AI-generated subset

The sectors facing the highest AI fraud risk

Across all fifteen major industries, the AI-generated subset (a uniform 5% share of each sector's total exposure) tracks the dollar totals, but it makes the absolute scale of the emerging threat concrete sector by sector.

The biggest dollar losses do not always reflect the sectors most exposed to AI-generated fraud. Some industries rely far more heavily on documents, making them easier targets for AI forgeries.

• Corporate expense and travel ranks tenth overall with $3.5 billion in exposure, including an estimated $175 million tied to AI-generated fraud, but it is the most document-driven sector in the study, with 85% of fraud schemes involving paperwork. High-volume, low-scrutiny receipts and corporate card statements make it an ideal target for AI-generated fakes.

• Rental housing and property management ranks eleventh with $2.7 billion in exposure, including an estimated $135 million linked to AI-generated fraud, and is the second most document-driven sector at 65%. Rental applications, pay stubs, and bank statements are often submitted remotely and can now be convincingly generated by AI within minutes.

The same pattern appears across pay stubs, expense receipts, and identity documents, which are highly vulnerable because they are easy to replicate and rarely deeply verified.

The documents most forged

Most of the national loss is concentrated in a handful of document types. Invoices dominate because of sheer B2B volume; the rest of the table is a mix of high-frequency, low-scrutiny documents and low-frequency, high-value ones.

Invoices alone account for more than $126 billion in estimated losses, largely because of the huge volume of business-to-business invoices processed every year, not because they have the highest fraud rate.

By comparison, mortgage application fraud creates much lower total losses because mortgage applications are far less common, but each successful case carries an average loss of $50,000, the highest per-incident loss in the study.

Big companies owe more, but small companies feel it more

Exposure increases with company size in total dollars, but the impact differs sharply by business size. Large enterprises face the highest overall losses, while smaller businesses face lower total exposure but a much heavier financial burden relative to their size.

Large companies with 1,000+ employees account for nearly half of US private employment and therefore carry most aggregate AI fraud exposure. But smaller businesses often face greater operational risk, as even a single fraudulent invoice can have a significant financial impact.

While large firms have historically relied on advanced fraud detection tools, AI-generated documents are becoming increasingly difficult to spot, weakening the advantage traditional detection systems once provided.

What actually changed: the rise of AI-generated documents

Creating convincing fake documents once required skill, time, and specialist tools. That barrier has collapsed. Generative AI can now produce realistic invoices, bank statements, IDs, and receipts in seconds and at scale.

Two major fraud-detection networks, Inscribe and Resistant AI, both found that around 5% of detected document fraud now involves AI-generated content. Applied to the estimated $178 billion US document fraud market, that equates to roughly $8.9 billion in AI-linked losses, despite the category being virtually nonexistent just 18 months ago.

Within Inscribe’s network alone, AI-generated document fraud grew roughly fivefold in just eight months between April and December 2025, showing how quickly the threat is accelerating.

Why this Is already a real problem

AI-generated document fraud is rising as fake documents become harder to detect. Unlike traditional forgeries, AI-generated files are clean, consistent, and free from the obvious mistakes that once exposed fake paperwork.

That weakens manual review, still the first line of defense for many businesses. Highly document-driven sectors like corporate expense, rental housing, and healthcare are especially exposed because AI-generated fakes can often pass with minimal scrutiny.

The divide: verified systems vs. everyone else

At its core, the AI-forgery problem is a verification problem. The businesses most exposed are the ones approving documents without properly verifying them, whether that’s invoices, pay stubs, receipts, or uploaded PDFs.

Many workflows still rely on visual checks alone, exactly the kind of process AI-generated documents are designed to bypass.

The key difference is no longer who receives fake documents, but who can actually verify them before action is taken. A workflow that verifies documents is an asset. One that simply accepts them is a growing liability.

What this means

The $8.9 billion AI-forgery figure is not the ceiling: it is the opening line. It represents a category of loss that was effectively zero 18 months ago and is compounding several times a year. For employers, the businesses that inventory their document workflows and verify before they trust will be the ones that absorb the shift. For everyone still approving documents on a glance, the gap between a real document and a convincing fake is closing fast, and AI is closing it from both ends.

The competition is no longer about who has the cleanest paperwork. It is about who can prove a document is real before acting on it.

“Document fraud has quietly become one of the largest unmanaged costs on the American balance sheet, and AI is changing its shape faster than most businesses can react. The losses are concentrated exactly where verification is weakest: high-volume, low-scrutiny documents moving through pipelines built for an era when faking a document was hard. As generative tools make convincing forgeries trivial, the businesses that thrive will be the ones that treat every document as something to be verified rather than assumed.“ said Jackson Barrett, Director of Product at Xodo Sign.

Methodology

The AI-generated fraud estimate is based on findings from Inscribe and Resistant AI, both of which found that roughly 5% of detected document fraud now involves generative AI. That share was applied to a modeled $178 billion US document-fraud market to estimate $8.9 billion in AI-generated fraud losses.

Industry exposure was weighted using employee counts, Association of Certified Fraud Examiners (ACFE) fraud-loss data, and the share of fraud schemes that are document-driven within each sector. Company-size exposure used US Census and SBA data.

Figures are estimates based on publicly available 2025–2026 data from sources including Inscribe, Resistant AI, ACFE, AFP, Deloitte, IBM, Ramp, and AppZen.