Purpose
AI Watermark Detection (PixPix) is used to first check exactly what hidden information a file contains before removal processing. It can analyze images and documents, presenting in a centralized view the identifiable metadata, C2PA content credentials, and AI-generation-related information they contain, helping users decide whether further cleanup is needed.
Main Capabilities
- Supports images and documents, and multiple files can be dragged in at once.
- Labels detection results by type: C2PA content credentials, AI traces, ordinary metadata, and items requiring attention.
- Distinguishes ordinary attributes, privacy fields, and generation source clues.
- Only analyzes files and generates a report; it does not rewrite the original files.
- Reports are generated locally in the browser; the detection process does not send files to a server.
How to Use
Drag in files, review the report, and decide the next step. The entire process is completed in the browser. If detected content needs to be removed, go to the corresponding image or document cleanup page for processing. Plain text files can be dragged in directly; to paste a piece of text for checking, the official recommendation is to use the text removal page, which includes a visual preview.
Use Cases
Suitable for scenarios where you need to first confirm file risks and then decide whether to clean up, such as doing a read-only check before delivering a resume, processing documents, or checking image metadata. After detection is complete, you can then enter PixPix's related tools to continue handling follow-up needs such as product image sets, detail pages, and marketing videos.
Limitations
- It can only provide clues, not determine whether an image is AI-generated. If the file contains C2PA content credentials or a generator field stating a certain AI tool, that is strong evidence; but finding no traces does not mean it is not AI-generated, because metadata is easily stripped (screenshots, format conversion, uploading to social platforms and downloading again can all make traces disappear).
- It cannot detect SynthID. SynthID's image watermark requires Google's detection model to read, and the text watermark requires a key; neither has a publicly available interface.
- This page is completely read-only and has no removal button.
- This tool handles identifiable hidden characters, metadata, and content credentials in files; it does not change the true creative origin of the content, nor can it guarantee bypassing any AI detection system.
- Only process files you own or have been authorized to use; before removing source credentials, consider copyright, audit, and platform disclosure requirements.
