For decades, a payslip was a payslip. A PDF bank statement, a P60, a letter from an employer on headed paper: these were treated as reliable proxies for real financial events because forging them convincingly took skill, time and access most applicants didn’t have. That assumption underpins most mortgage underwriting workflows still in use today. It no longer holds.
Generative tools have collapsed the cost and skill required to produce a document that looks entirely ordinary. Not obviously fake. Not badly formatted. Ordinary. The danger in synthetic evidence was never that it would be perfect. It’s that it would be unremarkable enough to pass through a process built to catch clumsy forgeries rather than convincing ones.
This shift is already visible in the market. Reporting on Australian lenders has described a surge in so-called template farms selling editable documents designed to mimic official papers, which borrowers then submit during home loan applications, exposing gaps in existing compliance checks (AFR, Fake document farms target banks amid home loan scam surge. The specifics of that case matter less than what it confirms: the document layer of lending has become a genuine attack surface, not a formality.
The mechanics: where fabricated evidence enters the loan file
Income and asset verification sits early in the mortgage process, usually before a case reaches an underwriter’s full attention. Applicants submit payslips, employer letters, bank statements and sometimes tax documents, typically uploaded as PDFs or photographed images through a broker portal or direct application system. A processing team or junior underwriter checks these against stated income, employment details and loan affordability criteria.
What that check is actually looking for has barely changed in years: correct logos, plausible pay figures, consistent dates, matching employer details across documents. What has changed is what a fabricated document now looks like. Template farms sell base documents that mimic real payroll software output or bank statement formatting, which applicants or brokers then edit with specific figures. Increasingly, generative tools can produce entire statements from scratch, complete with transaction histories that look statistically reasonable rather than obviously invented.
The reviewer checking this document is doing contextual reasoning: does this income match the job title, does this employer exist, does the loan amount make sense against the deposit. That is a genuine skill, and a valuable one. But it was never designed to catch pixel-level inconsistencies in a scanned image, font kerning that shifts mid-document, metadata that doesn’t match the claimed source software, or a bank statement that reuses a transaction pattern seen in a previous unrelated application. Those are pattern-detection problems, not judgement problems, and they sit outside what contextual review can reasonably be expected to catch. This isn’t a competence gap in the reviewer. It’s a mismatch between the skill and the threat.
The operational cost of not knowing what you’re looking at
For a lending team, this creates a genuinely difficult trade-off. Escalate every income document for deeper manual scrutiny and you slow down every application, including the overwhelming majority that are entirely genuine, at exactly the point where speed to decision is a competitive factor in the mortgage market. Escalate nothing beyond the standard visual check and you accept a growing false-negative risk: fabricated documents that look ordinary enough to pass, sitting in loan files that only surface as a problem much later, at far greater cost, if they surface at all.
Volume makes this worse. Processing teams handle high numbers of applications with fixed headcount and service-level expectations. There is no realistic way to apply forensic-level document review to every submission without a corresponding cost or delay that the business, and the borrower, will feel. The instinct to tighten controls uniformly across all applications is understandable, but it treats every file as equally suspicious, which is neither accurate nor sustainable.
What an authenticity layer changes
This is where an authenticity layer earns its place in the workflow, not as a replacement for underwriting judgement but as an input to it. Humanly analyses submitted images and documents, such as payslips, bank statements and financial letters, and surfaces indicators that may warrant closer scrutiny: signals that may indicate manipulation, inconsistency or synthetic generation, presented alongside the document itself inside the existing review process.
The underwriter or processing team member still makes the decision. Humanly doesn’t approve, reject or flag a loan file as fraudulent; it doesn’t automate that call at all. What it does is allow scrutiny to be applied proportionally rather than uniformly, a principle we think of as selective confidence. A document with no notable signals moves through the process at normal speed. A document carrying indicators worth a second look gets that second look, from a human, with context the system has provided rather than replaced.
That is the core of evidence integrity in a lending context: not slowing everything down, and not trusting everything by default, but giving reviewers the information they need to know which files deserve their limited attention. It reflects a wider principle we describe as human first, AI guarded: the reviewer’s judgement stays central, and the system exists to protect the quality of the evidence that judgement is applied to.
As synthetic fraud becomes a normal feature of the application process rather than an edge case, the lenders best placed to manage it will be the ones who treat document authenticity as a distinct layer of the review process, with its own tools and its own signals, rather than something a busy processing team is expected to catch by eye alone.
If your team is reviewing income and asset documents as part of loan origination and wants to understand where an authenticity layer would sit in that workflow, speak to our sales team to discuss your process.
