Landlords are being warned to tighten their referencing checks as AI makes it increasingly easy for fraudsters to create convincing fake payslips, employment references and even entire identities.
New research from lettings platform Goodlord, based on more than one million tenant references, reveals suspected fraud increased by almost 40% during 2025.
Fake employment references were the fastest-growing type of fraud, rising by 226.6% year-on-year. Referee fraud was up by 146.4% and fake identity I think we don’t talk about AI, quite enough in there to get these one attribution more at leastby 140.4%.
Although fake references, bogus referees and forged payslips remain above 2024 levels, there are signs that the overall level of fraud is beginning to fall, with all the main categories down slightly this year. Goodlord’s latest figures show 41 applications per 1,000 references were flagged for suspected fraud between July 2025 and June 2026, down from a peak of 46.6 per 1,000 in late 2024..
Changing nature of the fraud
Of greater concern, though, is the changing nature of the fraud being detected.
According to Goodlord, the problem is no longer confined to someone altering the salary on a payslip or changing information on a bank statement. Fraudsters are increasingly using AI to target different stages of the verification process, combining bogus employers, doctored identification and invented referees to create an application in which the false information supports itself.
Goodlord warns: “At first glance, fake documents created with modern techniques can be incredibly convincing.”
This makes fraud increasingly difficult to identify through conventional referencing checks, particularly where they rely heavily on documents supplied by applicants.
Significant costs
Getting those checks right matters because discovering fraud after a tenant has moved in can prove expensive.
Goodlord estimates the average potential financial exposure associated with a fraudulent tenancy is £9,601, including account rent arrears, legal and court costs, bailiff fees, void periods and potential property damage.
Using its suspected fraud rate and an estimated 5.3 million privately rented households, Goodlord calculates that potential financial exposure for the sector could reach as high as £4.1 billion a year.
The figure, however, represents potential exposure rather than actual losses. Not every application flagged during referencing will prove fraudulent and Goodlord’s own records show åa national confirmed fraud rate of 3.3 cases per 1,000 references.
And the level of risk varies widely, too. London recorded the highest rate of confirmed fraud, at almost twice the national average, followed by the West Midlands and North West.
High-value rentals are particularly heavily targeted. Homes renting for more than £10,000 a month recorded confirmed fraud rates approaching 18 cases per 1,000 applications – between three and six times higher than those seen among more typical rental properties.
Direct verification
The growing sophistication of the fraud is changing the way prospective tenants need to be checked. Goodlord advocates moving away from relying heavily on documents supplied by applicants towards information that can be verified directly at source.
This includes connecting directly to a prospective tenant’s bank, payroll provider or HMRC to verify income and employment, alongside technology designed to verify identity documents.
Goodlord describes direct-to-source verification as the “new normal”, because it allows information to be checked against its original source rather than requiring someone to determine whether a convincing-looking document is actually genuine.
Open Banking, for example, allows financial information to be verified directly rather than relying on uploaded bank statements, while payroll and HMRC data provide another way of checking employment and income details.
For landlords, particularly those using an agent or external referencing company, the findings are a good reason to find out how prospective tenants are being checked to see if the referencing service being used can independently verify the information behind them.
