New research shows document fraud is becoming more sophisticated, prompting payment providers to strengthen onboarding controls and fraud detection.
Payment providers face a 5.3% high-risk document rate. That may sound modest next to the most exposed sectors, but in payments, “modest” is a dangerous word.
Payment providers sit close to money movement, merchant onboarding, account opening, payouts, refunds, chargebacks, and business verification. A fraudulent bank statement, ID, or business registration that slips through does not stay a document problem for long.
It becomes an account problem, a transaction problem, a compliance problem, and eventually a trust problem.
The reality is that document fraud in general has industrialised faster than many onboarding controls have adapted.
According to Resistant AI’s Global Document Fraud Report 2026, an analysis of more than 170 million documents, nearly one in 10 documents analysed showed high-risk markers, meaning signs of tampering associated with strong fraudulent intent.
High-risk documents were up 28.5% year over year, while serial fraud increased 7x. Generative AI detections also rose sharply, with a 90x increase between 2024 and 2025.
For payment providers, the 5.3% rate should be read as a warning signal. Fraudsters are not testing documents in isolation. They are testing your defences between onboarding, know your customer (KYC), know your business (KYB), risk operations, and transaction monitoring.
Three forces are driving the pressure:
1) Template farms have made fake documents cheap and accessible. The barrier to entry has never been lower. Fraudsters don’t have to be specialists; anyone can download a template, make a small payment and, after a few minutes of editing, have a fully usable financial document at their disposal. Resistant AI has already catalogued 160 template farm websites, over 350,000 document templates, and an average template price of $28.29.
Payment providers often operate in high-volume environments where friction is the enemy. Fraudsters understand this and use a business’s pressure to onboard quickly as a gateway for documents that look plausible, especially when teams are relying on manual review or basic metadata checks.
2) Serial fraud is changing the shape of the threat. The same base document, or the same underlying fraud infrastructure, can be modified and reused across hundreds or thousands of submissions. Resistant AI found that 23% of all high-risk documents showed signs of serial fraud, and 98.3% of serial fraud detections had no attribution to known template farms.
This introduces another level of sophistication. The most serious operators are not necessarily buying obvious templates from known sources. They are building their own infrastructure, adapting faster, and spreading attacks across document types, entities, and use cases. A payment provider looking only at one submission at a time will miss the network.
3) AI-generated documents are becoming more practical. The current share may still be small, but the growth curve is not. Generative AI lowers the skill required to create convincing documents, especially when combined with template farms and document editing workflows. The future threat is not just a fake bank statement created by AI. It is AI-generated material inserted into a broader fraud operation that already understands onboarding rules, format expectations, and review thresholds.
The payments sector needs to stop treating document checks as a narrow onboarding control.
The right question is no longer, “does this document look real?” Nowadays they’re all pretty convincing. Reviewers need to dive deeper, asking: “What does this document reveal about the applicant, the source, the format, the structure, and the wider pattern of submissions?”
Structural manipulation, format hopping, metadata inconsistencies, repeated document patterns, and cross-submission similarities all require forensic analysis far beyond visual review.
Manual checks still play a role, especially in edge cases, but it can’t be the foundation for detecting industrialised document fraud in 2026.
Payment providers should focus on three changes: clearer document risk policies. Define what counts as unacceptable document risk, especially when the document is suspicious but not obviously fake.
They also need layered detection. No single signal is enough. Metadata alone can be misleading. Visual checks are limited. Template matching misses unknown attacks. Stronger detection comes from combining structural analysis, behavioural signals, document intelligence, and network-level patterning.
Finally, onboarding risk needs to connect to downstream fraud controls. A suspicious document should not disappear once an account is approved. It should inform monitoring, limits, review queues, and future risk decisions.
With a 5.3% high-risk document rate, payment providers are far from the worst-hit sector. But this level of exposure, and the advanced threats it correlates to, means there’s enough to worry about to revisit your document policies, especially those surrounding risk appetite, to make sure you are secure from the level of threats already wreaking havoc on other institutions.
This is the next phase of document fraud: faster, cheaper, and more distributed. Payment providers do not need heavier onboarding for every legitimate customer. They need sharper intelligence at the point where fraud first enters the system.
