Verify income documents before you lend
The problem
Editing a payslip used to need skill; now anyone can generate a convincing fake in seconds. Lenders, landlords, and platforms make daily decisions on PDFs nobody forensically examines.
Manual review catches the sloppy fakes. The good ones — consistent fonts, plausible numbers, clean metadata at a glance — sail through.
How Stipple helps
Stipple runs deep forensic inspection on every document: file-structure and metadata analysis, font and layout integrity, image-tamper signals, and arithmetic reconciliation (does gross − tax − deductions actually equal net?).
You get a risk band (low / medium / high), the specific findings that fired, a recommended action, and an audit-ready PDF report — a signal for your reviewer, not a black-box verdict.
The call
View Full Recipe →curl -X POST https://www.stipple.sh/v1/warrants -F "file=@payslip.pdf"
# → { "risk_band": "high", "recommended_action": "review_before_action",
# "signals": [ { "id": "domain.payslip.reconciliation", ... } ] }Integrate at intake: low-risk documents flow through automatically, elevated ones route to a human with the findings highlighted.
How the corpus was built, what was caught, and what the method does not prove.
Questions
- Can I reject a loan application because a payslip was flagged?
- No — treat a flag as a reason to look closer, not a decision. The check reports which signals fired and what each one is standing on, so a reviewer can see whether the net pay genuinely fails to reconcile or whether the document simply could not be read properly. A clean result means these checks found nothing, which is coverage, not a guarantee the document is real.
- Which super rate do you check against?
- The rate in force for the document’s pay period — 12% for periods starting on or after 1 July 2025, 11.5% for 2024–25 periods, and so on back through the legislated schedule. A payslip is judged against the rules that applied when it was issued, not against today’s rate, so a correct 2024–25 payslip showing 11.5% is not flagged. This only applies when the payslip states no percentage of its own; where it does, that figure is what gets reconciled.
- What kinds of payslip forgery does this actually catch?
- Careless ones, reliably: broken gross-to-net arithmetic, a superannuation figure that is not the statutory percentage, reversed pay-period dates, an employer identifier that fails its checksum. On a 210-document benchmark all four of those families were caught without a miss. A forgery that adjusts every figure in step still reconciles and will pass — which is why anything consequential needs one figure confirmed from outside the document.
- Does a photographed or scanned payslip still work?
- It is read, but coverage drops and the result says so. Scans and photos lose the structural signals a born-digital PDF carries, so they come back with a lower inspection quality rather than a confident verdict. That distinction is deliberate: a document we could not fully read is reported as exactly that, not as clean.
- What if the forgery adjusts every number in step?
- A consistent forgery still reconciles internally, and that family is exactly why the result recommends confirming one figure — an employer identifier, a bank credit — from outside the document. Coverage is stated on every result so you can see what was and was not tested.