Signal forensics check
What the pixels say when nobody drew them
Beyond what the eye sees, generated images carry signal-level tells: frequency patterns from the generator, upscaling traces, compression that doesn’t match the claimed source. Chekr’s forensics pass reads those signals — and reads the image’s declared provenance too.
What forensics finds
- Signal fingerprints characteristic of generative models
- Upscaling and re-compression traces that contradict the stated source
- Resolution too low for fine-detail checks (flagged, not guessed at)
- C2PA Content Credentials: declared generator and AI-generated status
- Missing provenance on images claimed to be credentialed
How Chekr checks it
The forensics pass analyses the image signal itself, independent of content, and folds the result into the integrity score alongside the visual checks.
Chekr also reads embedded C2PA Content Credentials and shows a provenance badge per creative: whether a credential is present, whether it declares AI generation and which generator signed it. Declared, not signature-verified — and absence of a credential never means "not AI".
Frequently asked
Is this an "AI detector"?
No — Chekr assumes your creatives are AI-generated and checks whether they’re fit to publish. Forensics exists to catch provenance mismatches and processing artifacts, not to play cat-and-mouse about whether an image is synthetic.
What is C2PA?
C2PA (Content Credentials) is an industry standard for embedding signed provenance in media files — which tool made the image, and whether AI was involved. Chekr reads and displays these credentials on every upload.