Home/Guides/Bank Statement Fraud Detection
Document Fraud
Bank statement fraud detection that finds the three numbers someone changed.
Edited balances, inserted salary credits, and fully AI-generated statements are the backbone of income and affordability fraud. TamperCheck analyses every statement across structural, visual, and metadata layers - digital PDF, scan, or phone photo - and returns a verdict with the specific evidence your team can act on.
- No document storage
- Digital, scanned or photographed
- ~1 minute per verdict
Last updated:
The problem
Why bank statement fraud gets through
A fraudulent applicant rarely fabricates a whole statement. They take a genuine one and change three numbers: the opening balance, a deposit or two, and the closing balance. Done carefully in a free PDF editor, the result is 95% authentic - which is exactly why visual review fails.
The template-matching tools many teams rely on fail for the opposite reason: a tampered statement still matches the bank's template perfectly, because it started life as a real statement from that bank.
And the newest wave skips editing entirely: AI-generated statements built from a text prompt, with plausible transaction histories and correct branding, that have never touched a bank's systems at all.
Manual checklist
How to spot a fake bank statement
Start with these six checks. They catch careless forgeries in a few minutes - but a statement that passes them is not proven genuine, because a careful edit is invisible on screen.
Recompute the running balance
Take the opening balance, apply each transaction in order, and confirm every running balance and the closing balance. One edited deposit breaks the chain on every row after it.
Compare the digits in edited fields
Zoom in on balances and salary credits. Re-typed figures often differ from neighbouring rows in weight, spacing, baseline or anti-aliasing.
Check the column alignment
Inserted or deleted rows leave uneven line spacing, amounts that don't sit on the same right edge, and dates that drift out of the column.
Read the PDF properties
Banks generate statements with a small set of known tools. A Producer or Creator of Word, Canva, Acrobat or an online PDF editor - or a modified date after the statement period - is a red flag.
Look at the transaction pattern
Round-number deposits just before the application, a salary credit with no matching employer name, or a balance that sits exactly at the affordability threshold all deserve a closer look.
Cross-check against other documents
Net pay on the payslip should credit the account on the stated pay date, and the name and address should match the rest of the application.
What we detect
Forensic signals that expose fraud
Running balance arithmetic
Every transaction row's running balance is recomputed against the previous row. A single edited amount breaks the chain - even when the edit is visually perfect.
Font metrics and substitution
Edited figures are usually re-typed in a near-match font. TamperCheck compares character-level font metrics across the document and flags fields that don't match their neighbours.
Editing tool fingerprints
Banks produce statements with a small set of known PDF generators. Traces of Acrobat, LibreOffice, or online PDF editors in the file's metadata and object structure are a strong tampering signal.
Transaction row insertion artifacts
Inserted or deleted rows leave structural residue - inconsistent line spacing, misaligned columns, orphaned formatting objects - that forensic parsing exposes.
Cross-document consistency
When statements are submitted alongside payslips or employment letters, salary credits, employer names, and dates are checked against each other. Fabricated bundles rarely stay consistent.
AI-generation signatures
Fully synthetic statements carry noise patterns, compression artifacts, and layout regularities that differ from genuine bank output. TamperCheck's computer-vision layer detects these signatures.
Manual vs automated
Can software detect altered bank statements?
Yes - and it catches the edits manual review can't. A reviewer won't notice that the running balance is off by $12.40 in row 47, or that the PDF's internal font table lists a typeface that doesn't match what's rendered on screen. Software checks every row and every object in the file.
| Manual review | Automated detection | |
|---|---|---|
| A single edited balance | Missed unless the edit is clumsy | Flagged from arithmetic, font and text-layer evidence |
| Inserted transaction rows | Sometimes spotted from alignment | Flagged from structural residue in the PDF |
| AI-generated statements | Built to pass a visual check | Detected from generation signatures |
| Scans and phone photos | Same visual check as PDFs | Image forensics replace PDF-structure checks |
| Coverage | A sample, when the queue allows | Every statement, at submission |
| Time per statement | Minutes, and it varies by reviewer | Seconds, and it is consistent |
For developers
Bank statement fraud detection API
Lenders, landlords and fintechs run the check inside their own application flow: the statement is uploaded once, checked automatically, and the verdict decides whether it goes to underwriting or to a reviewer.
One REST endpoint
POST the statement to /api/v1/documents/ with your API key and get a JSON verdict - authentic, suspicious or tampered - with a 0-100 risk score and findings tied to regions of the document.
Webhooks for batch workloads
Send a whole application bundle - statements, payslips, ID - and receive each verdict as it completes, without polling.
Route on the result
Clean statements skip manual review; suspicious ones go to a reviewer with the specific rows and fields already highlighted.
Self-serve
Create an account, generate an API key and test on your own statements with matched trial credits. The full reference is in the API docs - no sales call required.
The solution
Every statement treated as a forensic object, not a set of numbers to read
Every statement gets the same treatment: structural analysis, computer vision, AI adjudication in parallel. Verdict comes back with findings tied to specific regions of the document.
How it works
- Upload via API or dashboard - one endpoint, one file, verdict fast
- 200+ forensic checks run automatically on every statement
- Works on any bank - no template library to maintain, institutions covered globally
- Findings your team can act on directly - not just a risk score
- per-document pricing - no subscriptions, no minimums
- Zero document storage - files are processed ephemerally and never retained
Why us
What you won't get elsewhere
Findings, not a score
We name the field that changed and where it sits on the page.
Others hand you a number to justify.
One endpoint, every type
100+ document types, 190+ countries, no per-type setup.
Elsewhere, each new form is a project.
Catches what was never real
A dedicated pass for AI-generated and deepfake documents.
Template matching only catches edits.
Nothing is retained
Analysed in memory, then discarded. Never trained on.
Most services keep documents for QA.
Live in minutes
Published pricing, no minimum. Test it before you talk to us.
Platforms want a contract first.
Sits alongside your stack
Keep Onfido or Jumio for face match. We add the forensics.
Not a rip-and-replace.
FAQ
Common questions
What is bank statement fraud detection?
Automated forensic analysis of a submitted statement to determine whether it was genuinely issued by the bank in its current form - catching edited balances, inserted transactions, and fully AI-generated statements before they enter a lending, tenancy, or onboarding decision.
Can TamperCheck detect statements edited in a PDF editor?
This is the core use case. Edits leave traces in running-balance arithmetic, font metrics, the PDF's internal structure, and its metadata - even when the change is invisible on screen.
Does it work on scanned statements and phone photos?
Image-based checks (noise analysis, lighting consistency, AI-generation detection) run on scans and photos; PDF-structure checks run on digital files. The right check set is applied for each file type automatically.
How is this different from bank transaction data via open banking?
Open banking verifies income when the applicant connects their account - but many workflows still receive statements as documents, and applicants can decline to connect. Document-level detection covers the submissions open banking never sees.
How do lenders integrate bank statement fraud detection?
One REST API call per document: upload the file, receive a JSON verdict with findings. Most teams wire it into their loan origination or application workflow in under an hour. There's also a dashboard for manual checks, and trial credits to evaluate accuracy on your own documents.
Can software detect altered bank statements?
Yes. An edited statement keeps the bank's layout but breaks things a screen can't show: the running balance no longer reconciles, the edited digits use slightly different font metrics, the PDF's text layer disagrees with the rendered pixels, and the metadata names an editing tool the bank never uses. Forensic software checks all of these on every page in seconds.
What are the signs of a fraudulent bank statement?
Running balances that don't add up, figures whose font weight or spacing differs from the rest of the row, misaligned columns around a transaction, round-number deposits that appear just before the application, a creation or modification date after the statement period, and PDF metadata naming Word, Acrobat, Canva or an online editor instead of the bank's statement generator.
Is there a free fake bank statement detector online?
Free tools can show some of the evidence. TamperCheck's PDF Metadata Extractor reveals the tool that created or last saved a PDF, and the Document Tamper Checklist walks through the visual signs. A full forensic verdict - balance arithmetic, font metrics, text-layer and AI-generation checks - runs on paid credits, and new accounts get their first top-up matched with free credits to test on their own statements.
How accurate is automated bank statement fraud detection?
No system catches every edge case. Layering structural, visual, and metadata analysis catches the sophisticated fakes that pass manual review - partial edits, AI-generated statements, template forgeries. Every verdict comes with a risk score and the evidence behind it, so your team sets the threshold and can see why a statement was flagged.
What types of bank statements can be checked?
Any format: digital PDFs from online banking, scanned paper statements, and phone photos of printed statements, from banks globally. There is no template library to maintain, so a statement from a bank TamperCheck has never seen is checked the same way.
Or skip the call entirely
We'll match your first top-up in credits, so you can run your own documents through it now.