Detect forged documents in under 10 seconds.
Upload a document, get a clear forensic verdict in <10s. Works straight from your browser — no technical setup needed.
<10s
Median time to verdict
from upload to result
10
Forensic signals
run per document
190+
Issuing countries
document coverage
0
Documents stored
zero retention policy
The problem
Document fraud is bigger than most teams realise
1 in 5
fraudulent documents bypass manual review
Industry average, lending & KYC workflows
How it works
A verdict in three steps
Submit a document via API or dashboard. Get a specific AUTHENTIC / TAMPERED verdict with every forensic finding listed — in under 10 seconds.
Submit your document
Upload via the dashboard or POST to the API. PDFs, images, and scanned copies all work.
10 forensic signals run
Pixel analysis, arithmetic validation, font metrics, MRZ checks, AI-generation detection — all in parallel.
Verdict in <10s
AUTHENTIC or TAMPERED with every specific finding listed. Your team acts on facts, not a black-box score.
The product
10 forensic checks. One clear verdict.
Every failed check is explained specifically — not a black-box score your team has to guess at.
bank_statement_oct2024.pdf
Bank Statement · Meridian Bank · synthetic test document
78/100
Risk Score
10 / 10
Checks Run
3
Checks Failed
Forensic findings
Arithmetic Integrity
Running balance on page 2 is off by $4,320. Row 14 total does not match itemised sum.
Error Level Analysis
Pixel compression artefacts detected around transaction amount fields — consistent with copy-paste editing.
Font Metrics
Character spacing in rows 14–18 differs from the rest of the document. Possible text replacement.
Document Metadata
Creation timestamp and software metadata are internally consistent.
Visual Template Match
Header layout matches known bank statement template.
Verdict summary
Three independent forensic checks failed. This bank statement shows clear signs of digital manipulation targeting transaction amounts and running balances. Do not accept as evidence of income or financial position.
What each signal catches
Each check runs independently across pixel analysis, arithmetic validation, typography, metadata, and AI-generation detection. Failing one is enough to flag — nothing slips through.
Error Level Analysis
Detects pixel-level edits invisible to the naked eye
Arithmetic Integrity
Verifies totals, balances, and running sums add up correctly
Font Metrics
Spots character-substitution and copy-paste edits
MRZ Validation
Machine-readable zone checksum for passports and IDs
Template Matching
Compares layout against genuine issuer formats
AI Generation Detection
Identifies synthetically created or AI-generated documents
Metadata Analysis
Examines creation dates, software, and internal structure
Signature & Seal Integrity
Validates stamps, wet signatures, and security features
Plus dozens of issuer-specific rules, geolocation validation, QR code integrity, and cross-field consistency checks.
Tampered vs authentic
What tampering looks like — and what our checks catch
The same bank statement. One altered, one genuine. Forensic analysis tells them apart in under 10 seconds.
Altered bank statement
Balance off by $4,320 — arithmetic check failed. Pixel manipulation detected around row 2.
Genuine bank statement
All balances consistent. No pixel anomalies. Font metrics uniform throughout.
Coverage
Works on every document your team receives
100+ document types across 190+ issuing countries — all through a single API endpoint.
For developers
One API call. Every document type.
POST a file, get a structured JSON verdict. Integrate fraud detection into your onboarding, lending, or claims workflow in an afternoon.
curl -X POST https://api.tampercheck.ai/v1/verify \
-H "Authorization: Bearer tc_live_key_..." \
-F "file=@bank_statement.pdf" \
-F "document_type=bank_statement"{
"verdict": "TAMPERED",
"risk_score": 78,
"checks_run": 10,
"findings": [ ... ]
}Industries
Compliance & deployment
Built for regulated industries
Zero-retention architecture, BYOK inference, and API-first design means TamperCheck fits inside your existing compliance posture.
SOC 2 readiness
Architecture supports SOC 2 Type II. Audit-ready logging, access controls, and incident response procedures.
GDPR aligned
Zero document storage eliminates retention risk. No personal data written to disk or databases after analysis.
REST + Webhook
Single endpoint for all document types. Webhooks deliver async results to your own infrastructure.
OpenAPI spec
Full OpenAPI 3.0 specification. Auto-generate type-safe clients in any language in minutes.
Regional residency
BYOK routes inference to your provider's regional endpoint. Data stays in your chosen jurisdiction.
99.9% SLA
Uptime commitment across all API endpoints with status monitoring at status.tampercheck.ai.
Privacy
Your customers' documents are never stored
Most verification services retain documents for QA and model training. TamperCheck doesn't — not one byte, not one second.
Zero retention
Every document is analysed in memory and immediately discarded. Nothing is written to disk or stored in any database.
TLS 1.3 in transit
All uploads travel over TLS 1.3. No document content is ever logged or retained in request logs.
You own the outcome
Submitted documents are never used to train our models. Your data remains entirely yours.
Your key. Your inference.
BYOK routes LLM inference directly to your provider — OpenAI, Anthropic, Azure, AWS Bedrock, or Google Vertex. Token spend never passes through TamperCheck.
How much could you save?
Compare manual document review against automated fraud detection at any volume.
Manual review
$15k
/ month
With TamperCheck
$500
/ month
You save
$14.5k
$174k / year
Manual review estimate: 15 min analyst time at $60/hr ($15/doc). TamperCheck: $0.50/doc.
See full pricingFAQ
Common questions
What is TamperCheck.ai?
An AI-powered document fraud detection API. Submit any document — bank statement, payslip, passport, degree certificate — and receive a specific AUTHENTIC or TAMPERED verdict with every forensic finding listed, in under 10 seconds.
How does TamperCheck detect tampering?
Each document is run through 10 independent forensic checks in parallel: Error Level Analysis for pixel manipulation, font metric analysis for typography inconsistencies, arithmetic integrity for number fields, metadata examination, MRZ validation for identity documents, template matching against genuine issuer formats, and AI-generation detection. Failing any one check triggers a flag.
How much does it cost?
$0.50 per document check. Billing is page-based: 1–6 pages = 1 check, 7–12 pages = 2 checks (formula: ceil(pages ÷ 6)). New accounts get $5 in free trial credits. No seat fees, no volume tiers, no minimum commitment.
Are submitted documents stored?
No. Every document is processed in memory and immediately discarded after analysis. Nothing is written to disk or retained in any database. Submitted documents are never used to train our models.
Can I use my own OpenAI or Anthropic API key?
Yes. TamperCheck supports BYOK for OpenAI, Anthropic, Google Vertex AI, AWS Bedrock, and Azure OpenAI. Inference costs go directly to your provider at your negotiated rate — no markup.
What document types are supported?
100+ document types: passports, national IDs, driver's licences, bank statements, payslips, tax returns, invoices, insurance claims, degree certificates, academic transcripts, employment contracts, tenancy agreements, medical reports, utility bills — across 190+ issuing countries.
Get started
Your first 10 checks are on us
Sign up, receive $5 in free credits, and run real documents through TamperCheck today. No sales call. No enterprise contract.
$0.50 / document · $5 free · zero document storage