Read between the bytes: what metadata reveals about document and image authenticity.
Upload a file to start verification.
Drag & drop your document
or click to browse
PNG, JPG, WebP, PDF · Max 10MB
Your file stays in your browser until you create an account and confirm your email. Originals are deleted after analysis.
Estimated savings based on replacing a 10–15 minute manual document review with automated TrueDoc analysis.
Proprietary detection scans template variance, metadata drift, pixel-level retouching, and structural anomalies the human eye misses.
Inspects EXIF, software signatures, edit history, and structural fingerprints.
Originals are processed in encrypted memory and removed after analysis. Reports stay redacted by default.
Most fake documents carry metadata that contradicts the story they tell — creator tools that shouldn't have produced the file, edit timestamps after the claimed date, or edit traces from image editors.
TrueDoc surfaces these signals as part of every verdict, and exposes the raw metadata for analysts who want to dig in.
PDF creator tools that don't match the claimed source (e.g., screenshot editor producing a 'bank statement')
Timestamps that post-date the document's claimed date
Image editor signatures embedded in EXIF / XMP
Multiple incremental updates indicating later edits
Stripped or scrubbed metadata used to hide edits
PDF or image.
Creator tools, timestamps, edit traces, and embedded fingerprints.
Cross-check metadata against the document's claimed origin and date.
Use metadata findings alongside forensic and AI signals — never in isolation.
Metadata can be stripped or rewritten by determined actors. A clean metadata block does not prove authenticity.
Use metadata as a soft signal alongside visual forensics, structural analysis, and AI-generation detection — not as the sole basis for a decision.
Most "AI detector" tools look at one signal — usually a perplexity score on extracted text. The metadata fraud detection runs that as one layer of many. It also evaluates metadata lineage (software, edit history, geo), pixel-level forensics (ELA, font kerning, retouching regions), and structural anomalies in the underlying PDF or image container.
The reason: pdf creator tools that don't match the claimed source (e.g., screenshot editor producing a 'bank statement') rarely leaves only one fingerprint. A convincing forgery usually fails on two or three of those layers, even when one of them looks clean.
A high-risk verdict on PDFs (creator, producer, modification history, incremental updates), Images (EXIF, XMP, software fingerprints), Office documents where metadata is present returns per-field evidence — not just a score. You see the suspicious regions highlighted on the page, the specific metadata fields that triggered the flag (for example, "Timestamps that post-date the document's claimed date"), and the layer each finding came from.
That structure is what makes the verdict actionable: fraud and risk analysts can read why a document was flagged before deciding to reject, request a reupload, or escalate.
Scanned originals, mobile-camera shots, and re-exported PDFs are the three most common sources of benign anomalies. The metadata fraud detection scores those differently from the patterns associated with deliberate forgery — for example, a recompressed JPEG from a phone is not treated the same as a recompressed JPEG with a font substitution.
When a document is flagged, the report tells you which signal triggered it. If the only signal is a low-confidence compression artifact, the verdict is downgraded rather than counted as fraud.
No credit card. Redacted report in under a minute.