Document fraud in numbers: 2026 data and the cost of checking files

Fraud statistics are easy to quote and hard to use. Headlines mix payment fraud, identity fraud and forged paperwork into one number, and vendors publish estimates alongside official figures without always saying which is which. This report separates the two. It gathers the 2025 and 2026 figures we could open at the source, explains what each one actually measures, and then turns to the question buyers ask next: what does it cost to check files by hand, and when does automation pay for itself?

The headline numbers for 2025 and 2026

Three public sources give the clearest picture of the UK market, which publishes more fraud data than most European countries.

UK Finance. In its Annual Fraud Report 2026, published on 15 June 2026, the banking trade body reports that criminals stole £1.28 billion through payment fraud in 2025, an increase of 4%. Unauthorised fraud losses fell 5% to £703.4 million, but the number of unauthorised cases rose 11% to 3.81 million. Authorised push payment (APP) fraud, where victims are tricked into sending money, reached £576.4 million across 248,070 cases, up 19% in value. The industry also prevented £1.68 billion of unauthorised fraud.

Cifas. The UK fraud prevention service says in Fraudscape 2026 that its members recorded more than 444,000 cases to the National Fraud Database in 2025, the highest number in a single year and 6% more than in 2024. Over 242,000 of those were identity fraud cases, 54% of all fraud-risk filings. Identity fraud and facility (account) takeover together account for 72% of cases. Cifas members prevented £2.4 billion in fraud losses.

The National Crime Agency. Quoted in the same Cifas release, the NCA's Deputy Director of Fraud, Nick Sharp, states that fraud now makes up 45% of all crime in England and Wales.

Where documents fit in

None of these official series isolates "document fraud" as a category. Forged payslips, altered bank statements and fake proof of address sit inside identity fraud, loan fraud and first-party fraud cases. That is why estimates of document fraud specifically come from compilations. CheckFile's 2026 statistics roundup estimates that document fraud costs UK businesses around £1.8 billion a year, drawing on UK Finance, Cifas and NCA data. We treat that figure as an order-of-magnitude estimate rather than an official statistic: it combines several sources, and we could not trace every granular figure in the compilation back to a public primary release. When you brief a board, quote the official series and label any document-specific figure as an estimate.

Key figures at a glance

What changed: forgeries went digital

The most useful document-level data comes from identity verification providers, because they see the documents themselves. Entrust, which acquired Onfido, based its 2026 Identity Fraud Report on data collected from September 2024 to September 2025, drawing on more than 1 billion identity verifications across 195 countries. As reported by Biometric Update, the findings include:

• Digital forgeries made up 35% of document fraud in 2025, up from a 29% average between 2022 and 2024. Physical counterfeits remain more common (47%), but digital forgeries tend to be more sophisticated, driven by the online availability of AI tools.
• National ID cards accounted for 46% of all fraudulent document submissions globally.
• Deepfakes are linked to one in five biometric fraud attempts, and injection attacks rose 40% year on year.

Entrust itself cautions that its data reflects the verification space and may not mirror wider market trends. Still, the direction is consistent with Cifas. In its six-month update for 2026, Cifas notes that false identity filings fell 35%, yet its intelligence continues to point to growing concern about synthetic identities, AI-enabled impersonation and digitally manipulated documentation. Fewer crude fakes, more convincing ones: that is the pattern a manual reviewer faces.

Why this matters for file-based processes

ID verification tools check one identity document and a selfie. Many businesses, however, process a full file: an ID, a payslip, a bank statement, a proof of address, sometimes a company registration. A forged payslip can be internally perfect and still contradict the employer named on the bank statement. The risk sits in the inconsistencies between documents, and that is exactly where a tired reviewer at the end of a queue is least reliable.

What it costs to check files by hand

Public data on the cost of manual document review is thin. The most concrete benchmarks we found come from vendors, so we present them as vendor figures and show you how to test them against your own numbers.

CheckFile's homepage states that manual validation takes about 12 minutes per file on average and cites an 8 to 15% human error rate on verifications. Its document verification cost calculator uses that 12-minute baseline and compares it with an average of €0.30 per file for automated checks, including extraction, cross-validation and external enrichment.

How to run your own cost estimate

1. Count files, not documents. Take last quarter's monthly average of complete files reviewed.
2. Time a sample. Have two reviewers log 20 files each, from opening to decision. Include back-and-forth for missing pieces. Replace the 12-minute default with your median.
3. Use a loaded hourly cost. Salary plus employer charges, management and tooling, not just gross pay.
4. Compute the manual baseline. Files per month × minutes per file ÷ 60 × hourly cost.
5. Add the cost of misses. Estimate how many fraudulent files slip through per year and the average loss per incident in your business. This is often larger than the labour cost, and it is the number your finance team will challenge, so document your assumptions.
6. Price the automated scenario honestly. Per-file fees plus the time people still spend on flagged files, integration work and change management.

A worked example with illustrative inputs: 500 files a month at 12 minutes each is 100 hours of review. At a loaded cost of €35 an hour, that is €3,500 a month in labour. At CheckFile's quoted average of €0.30 per file, the same volume costs €150 in processing fees, before the human time you keep for flagged cases. Your own inputs will move these numbers; the method is what matters.

Tools for automating document checks

Our pick: CheckFile, for teams processing 100+ files a month

CheckFile is built for the full-file problem described above. Rather than reading one document, it cross-checks every document in a file (amounts, identities, dates, references), verifies data against official registries, and flags signs of AI generation or tampering. Each verdict (compliant, non-compliant, attention required) can trigger workflows such as CRM updates or follow-up requests. CheckFile is careful on the AI layer: it says the feature helps surface synthetic media without claiming 100% detection, which is the honest position for any tool in this category.

According to CheckFile, it reduces document validation time by up to 93%, at an average cost of €0.30 per file with 10+ documents, and deploys in 1 to 3 weeks. Data is hosted in France and Germany with AES-256 encryption, and documents are deleted after analysis. The pricing tiers start at 100 files a month (Starter), then 500 (Business) and unlimited volume (Enterprise), with a free pilot on your own documents and results in 48 hours. That pilot is the cheapest way to replace vendor benchmarks with your own. Sector setups cover banking and KYC, insurance, real estate, HR and financing.

Where it fits less well: if you only need to verify a single ID and a selfie at onboarding, a dedicated identity verification provider is the more direct choice.

Alternatives worth considering

• Onfido (Entrust), Veriff, Jumio, Sumsub: identity verification platforms with document and biometric checks, liveness detection and broad global document coverage. Strong for high-volume consumer onboarding. Sumsub also bundles KYC, AML screening and transaction monitoring.
• Ocrolus and Inscribe: focused on lending, with deep analysis of bank statements, payslips and income documents. A natural fit for credit underwriting teams, especially in the US market.
• Resistant AI: document forensics delivered through an API, designed to plug into an existing onboarding or underwriting stack.
• Mindee: developer-first OCR and data extraction. Good for teams that want to build their own validation logic on top of clean extracted data.

What the data says, in short

• Official UK figures show fraud volumes still rising in 2025, with identity fraud the single most common case type recorded by Cifas.
• No official series measures document fraud on its own; treat any document-specific cost as an estimate and say so.
• Verification providers see forgeries shifting from physical counterfeits towards more sophisticated digital ones.
• The cost case for automation depends on your volumes, your review time and the cost of a missed fraud. Measure all three before you buy.