Detection is an arms race. Verification is not.
AI-generated fakes will only get better. Stipple focuses on whether documents are true, not how they were made.
Documents, citations, prose — AI fakes them all. Check before you trust.
Trust nothing.Verify everything.
Stipple helps teams catch fake documents, broken citations, incomplete onboarding packs, and AI-written submissions. Run evidence-backed checks on the documents you receive — over one API and MCP.
Run a free check
No account · 20 documents/day free
Who it’s for
Same attack.Different victims.
Any business that makes a decision based on a document a stranger submits is now exposed. Find your team below and see exactly how it hits — and where to start.
Non-Bank Lenders
Fabricated income statements and bank transactions
Direct loan losses
Insurance
AI-generated repair invoices from fake businesses
Fraudulent claims paid
Property Management
Forged payslips and employment letters
Tenants who cannot pay
Finance Teams
Fake supplier invoices and doctored bank details
Payments to non-existent vendors
Mortgage Lenders
Doctored bank statements and inconsistent deposits
High-risk loans approved
HR and Recruitment
Falsified credentials, references, and employment records
Unqualified hires
How it works
Send it.See why.
No integration required to start. Every check returns an explainable signal with the evidence behind it — never an opaque score — so a person makes the final call.
- 01
Send a document
Upload a PDF or image, paste prose, or call the API directly. No account needed to start.
- 02
Stipple inspects it
Forensic and AI-generation signals, citation and arithmetic checks, field and identity reads — whichever the check needs.
- 03
You get an explainable signal
Evidence and reasoning, not an opaque score — with a shareable permalink and report. You make the final call.
For developers & agents
One call toverify anything.
Fact-check a report, detect AI-written text, or run document forensics over a plain REST API — no key required to start. AI agents get the same tools natively over MCP.
bash
curl -X POST https://www.stipple.sh/v1/verify-references \
-H "Content-Type: application/json" \
-d '{"url": "https://example.com/report.pdf"}'Free while in preview · 20 documents/day · no signup
Capabilities
One platform.Every document check.
Nine checks for the documents, identities, citations, and prose you receive — plus labelled datasets to test your own pipeline. One REST API, one MCP server.
Check a submission
Decide whether a document a stranger sent you is real.
Verify a document
Upload a PDF or image. Tampering, AI-generation signals, math reconciliation, provenance.
Verify nowVerify identity
Do these documents add up to identity? The AU 100-point check across passports, licences, Medicare and more.
Check identityCheck a document pack
Do these documents satisfy a checklist? Identify each, match the required slots, and see what’s present and what’s missing.
Check a packScreen adverse media
Screen a name for negative news and sanctions/PEP exposure. Corroboration-gated, with same-name false matches filtered out. A signal to review, not a verdict.
Screen a nameUnderstand a document
Read the type, the fields, and the personal data out of any document.
Classify a document
Payslip, invoice, bank statement, receipt? Financial document types with country and confidence.
Classify nowExtract fields
Pull structured values out of any document — by field, by template, or auto-detected. Reads checkboxes and tables.
Extract fieldsRedact PII
Find and mask personal data — names, addresses, emails, payment cards, SSN/PAN/ABN — and get a redacted copy of the text.
Find PIICheck generated content
Catch AI-written prose and citations that don’t hold up.
Fact-check sources
Feed it an AI research report. Citations resolved and matched, internal math recomputed.
Check sourcesDetect AI text
Paste prose and get the probability it was LLM-written, with the tells. Style, not authenticity.
Check textTest your system
Benchmark your pipeline before real submissions arrive.
Data Labs
Labelled financial-document packs — de-identified originals and synthetic sets with field-level ground truth — to benchmark OCR, extraction, and fraud checks.
Browse datasetsHow we think about verification
Our principles.
A flag without an explanation is useless.
Every flagged inconsistency comes with clear reasoning your team can stand behind.
The smaller you are, the less you should pay to be protected.
Document verification should not be a luxury only large institutions can afford.
No integration required to start.
Upload documents and get results. Build deeper integrations when you are ready.
Ready to trustyour documents?
Run evidence-backed checks on the documents you receive, and test your systems with labelled packs before real submissions arrive. Start free — results in days, not months.
No integration required to start