Detect AI-generated and fake documents

Upload a PDF and the free checker scans it for the evidence AI tools leave behind — C2PA Content Credentials, machine-readable "trainedAlgorithmicMedia" declarations, AI-tool metadata and Stable Diffusion traces — even inside compressed streams. Paired with tamper forensics and SHA-256 sealed-record verification. In your browser, no account.

Check a document now →

What the AI-origin scan finds

How to check a document, step by step

  1. Upload the document. Drop the PDF into the free checker at pdfverified.com/verify. It is analysed entirely in your browser — the file is never uploaded to a server.
  2. Read the AI-origin signals. The checker scans the file — including its compressed internal streams — for C2PA / Content Credentials manifests, IPTC "trainedAlgorithmicMedia" declarations, AI-tool metadata (ChatGPT/DALL·E, Midjourney, Stable Diffusion, Firefly) and generation-parameter blocks.
  3. Weigh the forensic report. Alongside AI markers you get structural forensics: edits after signing, changed file IDs, editing-software metadata, a producer classification (scanner vs HTML-to-PDF generator), and a conservative risk band.
  4. Confirm against the sealed record. If the document was sealed with PDF Verified, its SHA-256 fingerprint is checked against the signing record — the definitive answer, independent of any metadata.

The honest limits

A found marker is hard evidence baked into the file itself. But finding nothing proves nothing — metadata can be stripped, and many AI tools leave none. The reliable end-state is provenance: documents sealed at signing with a SHA-256 fingerprint and a public verification record can be confirmed genuine forever. That is what PDF Verified issues with every signed or stamped document.

Frequently asked questions

How can I tell if a document is AI-generated?

Upload it to the free PDF Verified checker. Many AI tools leave concrete, machine-readable evidence inside the files they produce: C2PA / Content Credentials provenance manifests (DALL·E via ChatGPT, Adobe Firefly and others), IPTC digitalSourceType declarations that literally state "trainedAlgorithmicMedia", tool names in embedded metadata, and Stable Diffusion prompt blocks. The checker finds these — even inside compressed PDF streams — and shows exactly what it found.

Can you detect a fake bank statement or payslip?

The checker surfaces the signals that expose most fakes: AI-generation markers, editing-software metadata, edits made after the original save, and a producer classification — genuine statements come from banking systems or scanners, while fakes are overwhelmingly rendered from HTML by tools like wkhtmltopdf or headless Chrome. No tool can promise 100% detection; for high-stakes decisions the issuer records remain the authority, and documents sealed with PDF Verified can be confirmed cryptographically.

Is AI-document detection 100% accurate?

No — and be wary of tools that claim it is. A found marker is hard evidence baked into the file itself. But finding nothing proves nothing: metadata can be stripped, and many AI tools leave none. That is why PDF Verified pairs the AI scan with structural forensics and, for sealed documents, a SHA-256 fingerprint match that is definitive regardless of origin.

What are C2PA Content Credentials?

C2PA (Coalition for Content Provenance and Authenticity) is an open standard for cryptographic provenance manifests embedded in images and documents. AI generators like DALL·E and Adobe Firefly attach one that records the tool that produced the content. If a C2PA manifest is present, the checker reports it and whether it declares AI generation.

Is the AI document checker free?

Yes. Upload and analysis are free with no account, and the file is processed locally in your browser — it never leaves your device. Verifying against the PDF Verified sealed-record database is also free and anonymous.

How do lenders and insurers use this?

Intake teams run uploaded statements, payslips and supporting documents through the checker to triage: explicit AI markers or post-signing edits escalate to manual review, while documents sealed with PDF Verified are confirmed cryptographically in seconds. The REST API supports bulk verification for KYC and onboarding pipelines.