The forgery is free now, and the reading of it is the failure
The fastest-growing form of UK tenancy fraud last year was the forged employment reference, which rose 227% across 2025 according to Goodlord’s analysis of its referencing platform. That is not a story about worse forgeries. It is a story about a verification model that assumed forgery was expensive, and has not been updated now that generative AI has made it free. A payslip used to be evidence because producing a convincing fake one took skill. As the American broker Jeremias Maneiro puts it, a payslip is now “a picture of a claim”, and no amount of careful reading recovers the difference.
Strategic Insight: The failure is not that references are being faked more often. It is that the entire referencing process still terminates in a person examining a document. Once document production costs nothing, every control built on document inspection has already been defeated, whether or not anyone has noticed yet.
That distinction matters commercially, because it determines what an agency should buy. Firms shopping for software that detects better fakes are competing in an arms race they cannot win. Firms that move the check off the applicant’s paperwork and onto the underlying source have stepped out of the race entirely.
Why “is Britain next?” is the wrong question
Maneiro, a RE/MAX associate broker based in Rochester, New York, who runs roughly 600 AI training sessions a year for US agents, set out the American picture in The Negotiator this week: fake rental listings assembled from scraped sale photos, deposits collected on homes the “landlord” has never owned, applicants arriving at properties for viewings that were never booked. In one case he describes, an agent proved a buyer’s identity document was forged only by going away and studying how a genuine one is constructed. She caught it. His observation about the next one is the part worth sitting with.
The framing of the piece asks whether Britain is next. The honest answer from the UK data is that Britain is not next, because Britain is already there. Across the twelve months to June 2026, Goodlord flagged 41 in every 1,000 references as suspected fraud, as reported by Property Industry Eye, below its late-2024 peak but well above historic levels. The genuinely interesting difference is not exposure. It is infrastructure.
The Real Story: Britain is better equipped and less deployed
Maneiro’s own conclusion undercuts his headline. Of the three source-level checks he runs in the United States, he rates the British equivalent stronger on two, and notes that on the third the US has nothing comparable at all. Open Banking gives Britain one national standard where American screening platforms stitch together a patchwork of vendors. Companies House gives Britain one company register where the US has fifty. And Right to Rent, run digitally via a certified identity service provider, gives British agents a real-time identity verification step that American lettings offices simply cannot obtain.
The problem is not capability. It is classification. Each of those three controls sits in most British agencies under compliance rather than under fraud prevention, and gets executed accordingly: at the end of the process, by whoever handles paperwork, to satisfy an auditor rather than to catch anybody.
| Metric | Value | Strategic implication |
|---|---|---|
| Fake employment references, 2025 growth | +227% | The fastest-moving attack targets the check agencies trust most |
| Referee fraud, 2025 growth | +146% | Rental-history verification by phone call is being gamed at both ends |
| Identity manipulation, 2025 growth | +140% | Synthetic identities are engineered to survive manual document review |
| Suspected fraud rate, twelve months to June 2026 | 41 in every 1,000 references | Roughly one in 24 applications carries a fraud signal |
| Average direct exposure per fraudulent tenancy | £9,601 | Arrears, legal and bailiff costs, voids and damage combined |
| Annual PRS exposure | Up to £4.1bn | The aggregate case for treating this as a business risk, not an admin one |
Strategic Reality: The £4.1bn headline is an extrapolation. Goodlord scales its observed fraud rate up across the UK’s roughly 5.3 million rented households, at two references apiece. That makes it a modelled exposure ceiling rather than a measured loss, and it should be quoted as such. The £9,601 per-tenancy figure is the number that actually belongs in an agency’s business case.
What breaks first, and it is not the obvious one
The instinctive fix is to verify the employer. Look the company up, confirm it exists, ring the number. Maneiro recommends exactly this, with the sharp addition that you should call a number you found yourself rather than the one printed on the reference.
Goodlord’s data suggests that control is already being outflanked. Mary Waterfield, its Financial Crimes Manager, describes the shift: “Previously we saw a lot of fraudulent income documents, provided by individuals declaring fake companies… Recently, we have seen an increase in false employment references being provided by individuals within large and well-known companies.” Over the past two years the firm has twice had to warn a real employer directly, once the NHS and once a large telecommunications business, on discovering that staff inside those organisations were writing false references for applicants.
Read that carefully, because it dismantles the check. The organisation is genuine. The email domain is genuine. The person answering the phone genuinely works there. Every signal an agent is trained to look for returns clean, and the reference is still false. Pair that with a payslip a generative model has formatted to match the employer’s real template, and the file survives any scrutiny short of asking the payroll system itself.
⚠️ Warning: An insider-issued reference from a real employer defeats company-register verification completely. If your fraud control is “confirm the employer exists”, you are validating the one element of the file that was never in dispute.
Success factors that get skipped
- Sequence over substance: The Right to Rent check is your strongest identity control and it usually runs last, after the applicant has already consumed hours of negotiator time. Moving it to the front changes what it is for. At the end it is evidence for a file; at the start it is a filter.
- Independently sourced contact details: Calling the number on the reference is not verification, it is a conversation with the person who wrote it. This is the cheapest control available and, per Maneiro, the one most consistently skipped.
- Source data over supplied data: Through direct-to-payroll, HMRC and Open Banking checks, Goodlord reports catching more than 97% of forged employment references, and 83% of referee cases, on its own platform. Those rates come from bypassing applicant-supplied contacts, not from reading their documents more attentively.
- Behavioural signals as a triage layer: An applicant based well outside the area. One who is never available for an in-person meeting, wants the whole process handled electronically, and keeps pressing on timing. None of these prove anything individually. Together they identify the files that deserve the slow treatment.
The implementation reality
Open Banking income verification is available to any agency whose referencing provider supports it, which most now do. The genuine obstacle is not procurement. It is that switching from PDF uploads to a source-connected check introduces a consent step, and a consent step introduces drop-off, and drop-off looks in the weekly numbers exactly like lost business.
That is the real trade-off, and it should be stated plainly rather than dressed up. Source verification costs you a measurable amount of applicant friction in exchange for an unmeasurable amount of avoided fraud. Every incentive in a lettings business points the wrong way, which is precisely why the infrastructure Britain already has sits so widely unused.
Hidden Cost: The applicant most likely to abandon an Open Banking consent step is not the fraudster. Fraudsters expect friction and are equipped for it. The drop-off lands disproportionately on legitimate applicants who are privacy-cautious or digitally uncomfortable, which means source verification has an access cost that agencies need to manage rather than deny.
The people problem underneath the technology problem
Every control described here asks a negotiator to slow down at the exact moment their compensation, their pipeline and their branch targets are asking them to speed up. Maneiro’s closing advice is “Slow is smooth and smooth is accurate”, and he is right about the mechanics. He is also asking front-line staff to absorb a commercial penalty personally.
Fraud arrives through whichever door has the least friction, and in a lettings agency the least-friction door is usually a person under time pressure who does not want to be the reason a let fell through. No verification technology fixes that. It is fixed, if it is fixed at all, by removing the discretion: making the source check mandatory and automatic rather than something an individual can decide to skip when the diary is full.
| Stakeholder group | Primary impacts | Support needs | Success metrics |
|---|---|---|---|
| Lettings negotiators | Asked to add friction against their own targets | Verification made automatic, not discretionary; targets adjusted for check time | Zero manually bypassed source checks |
| Referencing and compliance staff | Role shifts from document review to exception handling | Training on synthetic identity patterns and insider-reference risk | Reduction in files cleared on applicant-supplied evidence alone |
| Landlords | Bear the £9,601 average exposure directly | Clear explanation of which checks were run at source versus on paper | Arrears and void incidence by verification method |
| Applicants | Face new consent steps and digital identity checks | Accessible non-digital route; plain-English explanation of data use | Application completion rate held steady after rollout |
🎯 Success Factor: The measure of a working verification process is not how many frauds it catches. It is what proportion of approved tenancies were approved on source-verified data rather than on documents the applicant handed over. That number is knowable today, and most agencies have never calculated it.
What to actually do about it
💡 Implementation Framework: Move the check to the source
Phase 1: Audit what you currently trust (2 weeks)
- Calculate what proportion of last quarter’s tenancies were approved on applicant-supplied documents
- Identify every point in the process where a decision rests on someone reading a PDF
- Confirm whether your referencing provider offers Open Banking and direct-to-payroll verification, and whether you have it switched on
Phase 2: Reorder and harden (1 quarter)
- Move the digital Right to Rent check to the start of the application, not the end
- Stop accepting uploaded payslips where a source-connected alternative exists
- Mandate that referee contact details are independently sourced, and log the number actually dialled
Phase 3: Instrument and review (ongoing)
- Track source-verified approval rate as a standing management metric
- Review declined and abandoned applications for behavioural clustering
- Reassess controls annually against emerging attack methods, voice impersonation in particular
Priority actions by agency maturity
For agencies still on document review:
- Switch on what you already pay for: Most referencing contracts include source-connected options that are simply not enabled. This is a configuration change, not a procurement cycle.
- Fix the referee call: Independently sourcing the number costs nothing and closes the second-fastest-growing fraud category.
- Reposition the Right to Rent check: Run it digitally, run it early, and stop thinking of it as immigration paperwork.
For agencies partway through:
- Close the insider gap: Company-register verification does not survive an insider-issued reference. Direct-to-payroll or HMRC income verification is the only control that does.
- Measure the drop-off honestly: Track completion rates before and after introducing consent steps so the friction cost is a known number rather than a fear.
- Write down the escalation rule: Define in advance which behavioural signals trigger the slow process, so it is not left to individual judgement under pressure.
For agencies with source verification embedded:
- Prepare for voice impersonation: Goodlord’s Nicola Harding warns that fraudulent applicants could soon “provide someone else’s contact details and then use AI voice technology to impersonate them.” Any control resting on recognising a human voice needs a replacement plan now.
- Extend the model beyond lettings: The same reasoning applies to proof of funds in sales, where the document-inspection assumption is equally obsolete.
- Share intelligence: Attack patterns migrate between agencies faster than defences do.
Resource Reality: For most agencies this is roughly two weeks of process work plus a configuration change, not a technology programme. The expensive part is not the tooling. It is the management attention required to stop staff quietly reverting to the faster path.
Four things that will complicate this
The Companies House check is mid-upgrade
Maneiro rightly rates Companies House above the American state registers. But identity verification for directors and people with significant control only began on 18 November 2025, with a 12-month transition for existing companies filing their next confirmation statement. Until that period closes in November 2026, an established company on the register may still have entirely unverified officers.
Mitigation: Treat the register as a source of contradictory evidence rather than confirmatory evidence. A company incorporated recently at a mail-forwarding address with an officer sharing the applicant’s surname tells you something. A long-established entry does not yet tell you its officers are who they claim.
Voice cloning removes your cheapest control
The instruction to independently source a phone number assumes that the voice answering it is evidence of something. Voice cloning removes that assumption, and Goodlord’s own operations team is already flagging it as the next step.
Mitigation: Reclassify the referee call as a supporting signal rather than a verification step, and make sure no approval decision depends solely on it. Where rental history matters, verify the payments in the applicant’s bank data rather than in a conversation.
Detection tooling creates false confidence
Agencies that buy AI-powered document analysis often relax the source checks, on the reasonable assumption that the software is handling it. The failure mode is that detection and generation improve together, so the tool’s accuracy quietly degrades while confidence in it does not.
Mitigation: Never allow a detection tool to substitute for a source check. It can prioritise files for attention. It should not clear them. Our reporting on conveyancers using AI on the front line found practitioners drawing the same boundary: helpful on volume, never trusted on judgement.
The disclosure question is unresolved
Agents increasingly receive applications and correspondence they suspect were AI-generated, without any established basis for asking. The Property Ombudsman ruled this month that consumers will not have to declare AI use, on the grounds that its codes bind agents rather than complainants.
Mitigation: Do not build any control around applicants disclosing AI use, because they are under no obligation to and the honest ones are not the risk. Build around what can be verified at source regardless of how the submission was produced.
Reality Check: None of this eliminates fraud. A determined, well-funded attacker targeting a high-value tenancy will get through a proportion of the time, and Goodlord’s data shows premium properties attract fraud at multiples of the mainstream rate. The realistic goal is to make your agency the higher-friction door.
The strategic takeaway
British lettings agencies hold a verification advantage over their American counterparts that most of them have never recognised as an advantage. One national Open Banking standard, one company register, and a statutory digital identity check are exactly the tools US brokers are asking for and cannot get. The gap is not in what Britain has. It is in whether those tools are run as fraud controls at the point they would catch something, or as paperwork completed once the decision has already been made.
Three factors that decide the outcome
- Position, not possession: Owning a digital Right to Rent capability and running it at the end of the process are not the same thing. Sequence determines whether a control filters or merely records.
- Sources over documents: Every check that terminates in an applicant-supplied artefact is now compromised by default. The question to ask of any control is where its evidence originates.
- Removing discretion: Controls that individuals can skip under commercial pressure will be skipped under commercial pressure. Automation here is a governance decision, not a technology one.
Reframing what good looks like
The industry’s instinct is to measure fraud prevention by frauds caught, which rewards agencies with the worst applicant pools and punishes those whose reputation deters attempts. It also cannot distinguish between a control that works and one that has been silently defeated, since both produce few catches.
The more useful measure is coverage: what fraction of tenancies were approved on evidence that came from a bank, a payroll system, or a real-time identity check rather than from the applicant. That number is auditable, it moves when you change the process, and it does not depend on knowing what you missed.
Strategic Insight: Maneiro ends his piece with the whole argument compressed into two sentences: “Our industry keeps shopping for tools that spot better fakes. I’d start asking sources.” The tooling question is a distraction. The sourcing question is the strategy.
Your next steps
Immediate actions (this week):
- Calculate what proportion of last quarter’s approvals rested on applicant-supplied documents
- Confirm whether Open Banking and direct-to-payroll verification are available on your existing referencing contract, and whether they are enabled
- Check where in your application sequence the digital Right to Rent check currently sits
Strategic priorities (this quarter):
- Move identity verification to the front of the application process
- Stop accepting uploaded payslips where a source-connected alternative exists
- Introduce a logged, independently sourced referee contact requirement
Long-term considerations (this year):
- Adopt source-verified approval rate as a standing management metric
- Build a replacement plan for any control that depends on recognising a human voice
- Extend source-first verification from lettings into sales proof of funds
Source: BLOG: AI is creating havoc in US for estate agents so is Britain next?, a guest post by Jeremias “JMan” Maneiro for The Negotiator, 2026. Fraud statistics from Goodlord and Property Industry Eye’s report on Goodlord’s analysis of over one million tenant references.
This strategic analysis was written by Resultsense, a UK-focused AI news and analysis publication. We will be watching whether the Companies House verification transition, which closes in November 2026, measurably changes how British agencies treat register checks, and whether voice impersonation arrives in referencing before the sector has retired the phone call. Read more analysis at Insights, or get in touch.