Loan application LN-30417Document 3 of 5
A monthly payslip from Morvan Engineering Ltd, marked specimen
payslip_march.pdf, 14/04/2026File data: no producer
AI tracesManipulation traces
Loan application LN-30417Document 3 of 5
payslip_march.pdf, 14/04/2026File data: no producer
What we readWhat the eye sees
Logged to the recordAnalysis kept, reason written
ConclusionSuspect
Justifications
AI tracesHigh
Manipulation traces, net pay line
No producer in the file data
No generator signature found

A full queue,
and a decision to write down for every file.

Credit applications, new accounts, supplier invoices, expense claims, sick notes, tenant files: whoever handles the file has to decide whether the document can move forward. They also have to write down why, often with the next file already waiting.

What it does

Checks every document you receive and tells you whether AI produced or edited it, with the reason written out for the record.

Why it matters

A generated payslip is consistent by design, so the checks built to catch forgeries let it through.

Who it’s for

Teams that clear payslips, invoices and statements in credit, insurance and outsourced back offices, and the integrators who serve them.

The problem

Controls built for forged documents have little to catch in a generated one

Your controls check that a file is consistent, and AI now produces fakes that are perfectly consistent, a payslip produced from scratch or a genuine one with a single line regenerated. Label4.ai looks elsewhere, in the file itself.

The file with no original

A payslip that no payroll system ever produced. Your current controls compare a document with an original, and this one has none to compare it with.

The single rebuilt line

A genuine document where one figure has been regenerated rather than retouched, so everything else holds up, including the parts your rules check.

Supplier invoiceWhat your eyes see
What we see

A genuine supplier invoice in which only the bank details were regenerated.

The series

One actor sends dozens of applications from as many employers, with a single model producing all of them. Each file passes on its own, and the pattern shows once you can read how they were produced.

Payslip 1 of 3Payslip 1 of 3
Payslip 2 of 3Payslip 2 of 3
Payslip 3 of 3Payslip 3 of 3

Three employers, three applicants, one model behind all three payslips.

In figures

Document fraud keeps climbing, and AI is speeding it up

444,000

cases were filed to the UK National Fraud Database in 2025, a record and 6% more than the year before.

Source: Cifas, Fraudscape 2026, March 2026
30%

of false application filings in 2024 came down to false documentation, the most common reason of all.

Source: Cifas, Fraudscape 2025, April 2025
nearly5×

more AI-generated document fraud each month in December 2025 than in May of the same year.

Source: Inscribe, 2026 Document Fraud Report

Our answer

Where Label4.ai fits in your process

Label4.ai adds one question to the checks you already run, and leaves the rest of your process as it is.

Document received

Upload, e-mail, portal or scan, whatever your intake already is.

Your existing checks

Template matching, rules, the engine you already licence. Unchanged.

Label4.ai analysis

Is this document what it claims to be? Checked automatically as it arrives, in seconds, with its reasons.

Decision

Accept, ask for another document, or escalate, with the reason written down.

A result anyone on the team can read

The operator sees the conclusion first, then three sentences written in the words of the job, like these.

  • This document carries traces of a generative model across the whole page.
  • The net pay line was rebuilt separately from the rest of the page.
  • No generator signature was found, so the conclusion rests on forensic analysis alone.

How we do it

What sits behind each conclusion

  • An analysis shaped to your process Each project is set up for the way you work, with the checks that matter for your content, your threshold and the action we recommend. Every analysis looks first for a signature left by the generator, then for the traces a model leaves and for signs of editing.
  • Detection that keeps the process running Turning away an honest applicant is costly, so before the contract we run your own content and agree with you on the false alarm rate that suits your process. That is why we quote no generic detection rate: we set up a project and tune it to give you the best detection rate on your own use case.
  • A conclusion that holds up The same content, analysed twice, gives the same report twice, which turns a series of applications from suspicious into provable.

It runs in our cloud or on your own servers, through the Anomaly Engine console or the API.

See the technology in detail

Your content

See what your own files reveal

You send a batch of documents whose answer you already know, including the frauds you caught and the ones you found out about too late. We come back with a result for each file, your false alarm and missed fraud rates, and a plain statement of what we did not see.

We reply to every request within two working days, under NDA if you prefer.