AICheck365 Blog: AI Detection and Provenance
Evidence-first articles about AI content marking, C2PA, metadata, platform signals, and the limits of automated detection.
Updated featured analysis
Claude Watermark Explained: How It Works, EU Exceptions and Detection Limits
Updated after Anthropic’s revised explanation: learn why the mark is a statistical word-choice pattern—not hidden Unicode—what the planned detector API could prove, and how standard editing, code, short text, SynthID, and C2PA differ.
More articles
SynthID Detector Explained: What Google’s Watermark Checker Can and Cannot Tell You
Google opened synthid.com to everyone on October 7, 2026. Supported formats, the companies it covers, how to read a result, and how it complements local metadata checks.
Read the article →How to Tell If an Image Is AI-Generated
Read C2PA Content Credentials, EXIF/XMP, and PNG generation parameters from the original file instead of relying on visual guesses.
Read the article →How to Check If a Video Is AI-Generated
Check an original MP4/MOV for C2PA, container metadata, SEI markers, file-name clues, and sampled frame watermarks.
Read the article →How to Check If a Gemini Image Is AI-Generated
Inspect Gemini and Google AI files for C2PA, digitalSourceType, claim generator strings, XMP fields, and the limits of SynthID attribution.
Read the article →AI Image Metadata Explained: EXIF, XMP, C2PA and PNG
Where AI platforms write provenance data, how each field should be interpreted, and what disappears after sharing or re-exporting.
Read the article →What Is C2PA? Content Credentials Explained
How signed provenance manifests work, what validity and trust mean, and why missing credentials do not prove human origin.
Read the article →