Read our white paper on AI forgeries in 2026
New AI fake detection model. Try it now
4.6 / 5 on Trustpilot

Detect a fake document
in seconds

Upload a PDF, a scan or a photo. 150 checks on the file, the image and the content, and a fraud score from 0 to 100.

  • 22k+ users
  • SOC 2 and GDPR
  • Try for free
  • SOC 2 Type II, GDPR
  • Deleted after analysis
  • Never used to train an AI

For document volumes or an API, talk to an expert.

Check your own document

Analyzed securely by Koncile. Never used to train a model, deleted as soon as the analysis is complete.

  • SOC 2 Type II, GDPR
  • Secure servers
  • Deleted after analysis

For document volumes or an API, talk to an expert.

The document checker used by insurers, banks and large companies

4.8 / 5
Definition

What is a fake document detector, and what does it check?

Specimen bank statements on a table
01

Fakes multiply in the age of AI

Editing an amount, changing a name or producing a statement from scratch now takes a few minutes with an AI tool, with no trace visible to the eye. Digital fakes, files modified after they were issued, make up 5% of the 1.6 million documents analysed by Koncile.

€946 Mof fraud identified by French insurers in 2025, according to ALFA, the French insurance anti fraud agency.
  • Native PDF, original file: issuer library, serialiser, edit traces.
  • Image, scan or photo: classic editing, AI modification, file anomalies.
  • Internal consistency, business rules: calculations and amounts, dates, mandatory mentions.
  • External consistency, official databases: HistoVec, company registry, IBAN and VAT.
02

Four analyses in parallel

Two analyses look at the file itself, depending on whether it is a native PDF or an image. Two more look at what the document says: its internal consistency, then its match against external databases. Their signals add up to a single score from 0 to 100.

150+checkpoints on every document, from the PDF producer software to recalculated totals.
03

A score from 0 to 100, with evidence

The checker returns a fraud score from 0 to 100 and the list of anomalies, test by test, with the area concerned on the document. A high score remains a signal to verify, never an accusation, and a clean document is authenticated.

89 %of fakes caught on 1,200 scanned bank statements, versus 51% for the best language model.
Features

What the checker verifies on every document

Six families of checks run at the same time on every document you upload: image editing, the PDF file itself, the issuer's template, business rules, the consistency of amounts, and cross checks against external databases.

Edits and AI

K-Detect, our AI detection model, spots edited areas and images generated in whole or in part.

PDF file forensics

The checker reads the structure of the file, not only what it displays: metadata, successive versions, fonts and producer software give away an edit made after issue.

Issuer template

Every document is compared with the known templates of its issuer: format, layout, metadata.

Business rules per document type

Rules written with accountants, lawyers and compliance specialists check what a forger forgets: the deductions on a pay stub, the VAT on an invoice, the dates on a statement.

More than 30 databases

More than 30 databases are queried: company registries, banking references, data leaks.

Comparison

Why not simply ask an LLM?

On 1,200 bank statements including 400 forgeries, Koncile detects 89% of the forgeries, against 51% for the best LLM. Same test set and same threshold for every model.

LLMs are not built to detect forgeries

These general purpose models ignore fraud patterns

Customer feedback

What results do our customers get on OCR and fake documents?

Insurers, fintechs, audit firms, manufacturers and healthcare players: published case studies, video interviews and verified reviews, in Europe and the United States.

Barely three months after going live, we saved about 70% of the time spent checking documents. Between 95 and 99% of the data is read automatically by Koncile.

Ghislain AvertyChief Operating Officer, Toyota Assurances

It takes three minutes to create the Koncile button on the scanner, and the team is up and running right after. Hardware alone is no longer enough: hardware plus solution is what creates the added value.

Jean-Michel Pavie-HoudryEpson France

Koncile secures the processing of our insurance documents: extraction and fraud detection make us more reliable on the statements and documents we receive.

Louis ReignierHead of production, CMAM

We receive documents from all our rental partners in very different formats. Koncile extracts and structures them down to the detail: the data arrives ready to export, with no retyping.

Pierre-Henri VernayDeputy Director, Next & Go

Before Koncile, consolidating certificates took several days. Now it takes a few hours, with 100% traceability and simpler audit preparation.

Hugo DepondDeputy Director, Bilfinger LTM Industrie

Across several million documents a year, extraction and fraud detection keep the promises we make to our customers.

Michel BrancaleoniChairman and co founder, ADN Solutions

The success rate is close to 100%, even on an exam code that is not present on the document.

Sean HickeyHealthNet Homecare

What I appreciate with Koncile is its ability to extract any piece of information from a document.

Benjamin KasprzakChief Operating Officer, Meelo

Koncile integrates in under an hour.

Vincent SattlerYomoni

Thanks to Koncile, processing receipts no longer takes a whole team: one person is enough, and document checking is 3 times faster.

Charles LacrouxHead of Venture, Reward Pulse

We connected Koncile through the API in a few days. It processes tens of thousands of invoices a year, at a third of the cost of the tool we used before.

Lucas Biguet-MermetTech Lead, Ecotable

Koncile read our invoices, student files and teacher contracts from day one. We cleared a backlog of several thousand documents in a few days.

Melinda MagyarCFO, Luxembourg School of Business
See more testimonials
Security

Your documents are safe

The uploaded file is encrypted, analyzed on secure servers and deleted as soon as the analysis is complete, unless you keep it in your history. It is never shared or used to train a model. Koncile is SOC 2 audited and compliant with GDPR, HDS and HIPAA.

Read our security documentation
  • HIPAA Compliant certification badge
  • SOC 2 certification badge (AICPA Service Organization Control)
  • GDPR Compliant certification badge
  • HDS certification badge (French health data hosting)
Use cases

A fraud detection tool for risk, compliance and operations teams

The same checker, online for one document or by API for your whole flow.

  • KYC

    Customer onboarding, identity verification and income validation.

  • Procurement

    Supplier onboarding, invoice fraud detection.

  • Lending

    Loan origination, mortgage verification and BNPL risk analysis.

  • Insurance

    Claims validation and new customer onboarding.

  • Fintech

    Fraud detection in onboarding and transactions.

  • Marketplaces

    Seller verification and document risk control.

Blog

Understanding document fraud, and how to beat it

Analysis

Detecting a fake bank statement: our 9 methods

Study of 150,000 scanned or native statements: the nine checks that bring forgeries down, from the creation software to the balances.

Read the article
Comparison

Top 10 document fraud detection software in 2026

The ten tools on the market put to the test: what they see, what they miss, and for which document type.

Read the article
Analysis

Document processing and fraud detection: build or buy?

With AI, building your own pipeline has never looked so easy. What in house fraud detection really costs.

Read the article
Comparison

In healthcare, an OCR is judged on what it flags, not on what it reads

Every vendor claims high accuracy rates. On a prescription, that is not the right question.

Read the article
Analysis

Developers launch MacDoc, an illegal fake passport generator

An investigation into MacDoc, the fake passport generator built by the same engineers behind OnlyFake, the tool already shut down by US authorities.

Read the article
Feature

Document deepfakes: catch the fakes slipping past your checks (2026)

Document deepfakes pass classic checks: how to detect AI-generated fake documents in 2026 with semantic coherence analysis.

Read the article
Comparison

OpenCV in Python: Detecting Document Fraud Through Image Analysis

Document fraud detection with OpenCV in Python: real tests and limitations.

Read the article
Feature

Invoice fraud : how to detect fake invoices and suspicious suppliers

How invoice fraud works, the most common red flags, and why basic controls are no longer enough.

Read the article
Feature

Fraud detection on documents: weak signals that matter

How weak technical signals reveal document fraud risks.

Read the article

Your questions about the document checker

Short answers, with figures when we have them.

Talk to us
  1. Upload a PDF, a scan or a photo, up to 20 MB. The checker analyses it in a few seconds and shows a score from 0 to 100 with the list of anomalies, test by test. The first document is free, no credit card needed.

  2. From 0 to 39 the document is authenticated, from 40 to 79 it needs a closer look, from 80 the risk is high. Every score comes with the tests that explain it and the area concerned on the document. A high score is a signal to verify, not an accusation.

  3. Across 1.6 million documents processed, bank statements, payslips, invoices, ID documents or proofs of address, Koncile flagged 5% as manipulated or transformed files. We measure detection document type by document type, on test sets where the fakes are known: on scanned bank statements, for example, 89% of fakes are caught, versus 51% for the best language model. Each rate is published together with what we do not measure yet.

  4. Koncile's false positive rate is 1.5%. A poorly calibrated metadata test flags 10 to 20%, which is why Koncile calibrates every check by issuer and by channel. A verdict remains a signal to verify, not an accusation.

  5. The Advanced plan is €199 per month, about €0.20 per page, with API, SDK and MCP access. The Enterprise plan is quote based. The online checker stays free for a first document. See pricing.

  6. Yes. It is encrypted and deleted as soon as the analysis is complete unless you keep it in your history. It is never shared, sold or used to train a third party model. Koncile is SOC 2 certified. Security policy.

  7. Bank statements, payslips, invoices, identity documents, proofs of address, tax notices, company registers, prescriptions, driving licences, in around thirty countries and several languages. All documents.

  8. REST API, Python and JavaScript SDKs, webhooks and MCP server: one call returns the extraction and the score. Ecotable connected Koncile through the API in a few days. See the API.

  9. Yes, for PDFs as well as images, thanks to the known traces left by generators: 5.2% of the PDFs in our corpus are AI generated. Video deepfakes are out of scope, and we say so plainly.

  10. Yes: Koncile includes an OCR that extracts precise data, then cross checks it between documents. Identity, income, dates, addresses or bank transactions that do not match. It can also easily identify several documents within a single file.

  11. Koncile is a tool built for professionals. Banks and fintechs, insurers, KYC teams and healthcare providers: Meelo, Toyota Assurances, ADN Solutions, HealthNet Homecare, 1stKYC.

  12. Because it does not only look at the image: it cross checks the file, the image and the content, then compares each document with the known documents of its issuer, bank by bank, software by software. On 1,200 scanned statements, it detects 89% of the forgeries, against 51% for the best language model.

  13. A language model reads the document like a human in a hurry: it sees neither the file metadata, nor pixel level editing traces, nor the issuer's usual template. A clean forgery, rewritten with the right software, goes through without an alert. Purely visual tools, for their part, do not check that the amounts add up.

  14. Because they apply the same threshold to every document: a statement exported from a bank's app has unusual metadata without being fake. Without a reference by issuer and by channel, these tests produce 10 to 20% false positives. Koncile calibrates every check on what each issuer actually produces.

Try Koncile's fraud detection on your own documents

Upload your bank statements, pay stubs or IDs: Koncile extracts the data and flags fakes in seconds.

Talk to an expert