When Fake Documents Become Fake Realities

Altered payslips, manipulated bank statements and invented references have been a problem for as long as organisations have relied on documents to prove that someone is who they say they are. What artificial intelligence may be changing isn’t the existence of fraud, but how much of the reality surrounding those documents can now be fabricated.

Recent research from lettings platform Goodlord found significant increases in several types of tenancy fraud, including a 226.6% year-on-year rise in fake employment references, alongside increases in referee fraud and identity manipulation. The company estimates that fraudulent tenancies could represent as much as £4.1 billion in annual direct financial exposure across the UK private rented sector.

Those figures are striking in their own right, but the more interesting question is what might be changing underneath them. If AI makes it easier to create not just one convincing fake but several pieces of evidence that support each other, the challenge starts to look quite different.

From Forgery To Fabrication

Traditionally, fraud often involved changing something that already existed. A salary might be increased on a payslip, information altered on a bank statement or details provided for an employer that didn’t exist. Verification was therefore largely about finding the inconsistency:

• the number that didn’t match
• the document that looked unusual
• or the reference that couldn’t quite be confirmed.

AI makes it considerably easier to go further. Instead of manipulating one piece of evidence, fraudsters can potentially create multiple pieces designed to support the same false story, with an invented employer, a convincing referee and supporting documents all appearing to confirm one another.

Individually, each piece may look plausible, but together they can create something much more convincing: an entire version of reality in which everything appears to add up.

That changes the challenge from forgery to fabrication. Rather than only asking whether a document has been altered, organisations increasingly need to establish whether the person, employer or circumstances represented by it are real in the first place.

When Everything Appears To Add Up

Consistency has traditionally been one of the things that helps us establish trust. If a payslip, employment reference and application all tell the same story, we’re more likely to feel confident that the information is genuine, but that becomes a much weaker signal when several pieces of evidence can be created specifically to corroborate each other.

It means looking closely at an individual document may no longer be enough. Goodlord has argued for greater use of independently verified information, including direct sources such as Open Banking, HMRC and payroll providers, rather than relying primarily on evidence supplied by an applicant. The question then shifts from “Does this evidence look real?” to something harder to fake: “Can we independently prove that what this evidence describes is real?”

Brian Feeney from AI-XON, sees that as a significant change for organisations that have traditionally relied on evidence packs to make decisions;

“I’m glad that I don’t have to carry the risk profile that some organisations out there have to carry. Evidence packs that used to be unquestionably safe to make decisions off the back of now shouldn’t be trusted. It’s a huge shift, we’re seeing lots of organisations struggle with leakage, and others that aren’t even aware that it’s happening.” 

Beyond Tenant Referencing

Tenant fraud is one example, but the same challenge potentially appears anywhere organisations rely on digital evidence. Recruiters check employment histories and qualifications, insurers assess evidence supporting claims, financial organisations verify income and identity, while businesses depend on documentation for compliance and due diligence.

Generative AI doesn’t suddenly make those processes ineffective, but it can reduce the effort required to create convincing digital material. A process designed around spotting the suspicious document may therefore struggle when the document is only one part of a much bigger fabrication.

For years, verification has often involved examining the evidence in front of us and looking for the thing that doesn’t add up. AI introduces a rather more difficult possibility: what happens when everything does?