AI ScreenerIndependent content forensics
Forensic evidence, not a magic score.
Upload an image, video, audio recording, or document photo. AI Screener runs the relevant checks and explains what supports—or weakens—an authenticity concern.
Check suspicious evidence
No account required · one report in a few minutes
Drop files here
One file for a quick check, or up to four related images for one evidence case.
Image evidence
JPEG, PNG, WEBP · 1–4 related images · 25 MB each
Audio & video
Research previewWAV, FLAC, MP3, OGG, M4A, MP4, WEBM, AVI · one file
Public-preview uploads and derived evidence are currently stored indefinitely. Only upload files you are authorized to submit. Read privacy details.
Illustrative sample report
A result a human can act on.
Concern, confidence, and coverage answer different questions. The report keeps them separate, shows which evidence families contributed, and recommends a next step without declaring a file “definitely fake” or “guaranteed authentic.”
Case assessment · example
Review recommended
Elevated concernTwo localized noise inconsistencies agree with an unusual recompression pattern. Provenance is absent, while document fields remain internally consistent.
Most useful next step
Ask for the original, full-resolution file. Compression limits two pixel checks and may explain part of the raised signal.
Concern
64 / 100Several independent signals deserve attention. This is not a probability that the file is fake.
Assessment confidence
HighEnough usable evidence was available for the methods to support a stable assessment.
Evidence coverage
7 of 9Two checks could not run on this compressed copy. Missing evidence is not counted as a pass.
What gets checked
Independent signals, kept separate.
Every applicable family reports its finding, confidence, limitations, and available visual evidence. Disagreement is preserved instead of averaged into a verdict.
See the full method registry- 01
AI-generation signals
Learned synthetic-pattern detectors report model-specific signals. A negative result stays inconclusive when coverage is weak.
- 02
Pixel & compression
Noise, resampling, copy-move, frequency, and recompression checks surface regions that may have a different history.
- 03
Provenance & integrity
Metadata, C2PA credentials, capture history, and native file structure show what can—or cannot—be verified about origin.
- 04
Document & identity logic
OCR, MRZ, barcode, portrait, geometry, date, and cross-file consistency checks help a human review related evidence.
- 05
Audio & voice
Research previewWaveform, frequency, watermark, and speech anti-spoof signals are presented with timelines and explicit limitations.
- 06
Video
Research previewRepresentative frames, model-score timelines, temporal motion, encoding, audio tracks, and provenance remain separately visible.
For fraud and trust teams
Related files become one evidence case.
Add an ID front, reverse, selfie, or proof photo. When relevant, AI Screener automatically adds face, OCR, MRZ, barcode, date, and metadata consistency checks without hiding each file’s own analysis.
Join the workspace betaInvite-only preview. The production workspace is not publicly enabled.
Marketplace review · Case 1842
ID front
Document and portrait detected
ID reverse
MRZ and barcode extracted
Selfie
Presenting face detected