This is a marker check, not a verdict. It looks for known AI-provenance signals: Content Credentials (C2PA), the IPTC digital source type, Stable Diffusion and ComfyUI metadata and invisible Unicode. It cannot prove that an image or text is AI-generated or human-made, and a clean result only means no marker was found; most social platforms strip this metadata on upload, and pixel or token watermarks such as SynthID need the vendor’s own detector.

Related: /exif shows every metadata tag in a photo and strips it, /docmeta does the same for documents, and /injscan checks text for prompt-injection payloads hidden in the same invisible characters.