Detecting Amazon AI-Generated Content

Upload an image, audio, or video file to check whether it contains C2PA Content Credentials or watermarks indicating that one of our generative AI systems generated it.

Upload a file

Drag and drop or select a file.

Supported formats: PNG, JPEG, GIF, WEBP, MP4, MOV, AVI, MP3, WAV

By uploading content, you agree to our Conditions of Use and acknowledge that you have reviewed our Privacy Notice.

Analyzing your file...

File Details

Filename
Type
Size

Detection Signals

Validation Status Details

Active Asset
Signed by
Outcome Code Details
No validation results were reported for the active manifest.

Signature Information

Common Name
Issuer
Signed
Certificate Validity

Active Manifest

Claim Generator

Assertions

Assertion Details
No assertions were reported for the active manifest.

Review C2PA manifest


    




        

    


    

        

            
                
            
        

    




                

    

        
How does this work?

        Upload an image, audio, or video file. The detector analyzes the file for embedded provenance signals, including C2PA Content Credentials and Amazon watermarks. Once analysis is complete, the detector will return a result indicating whether supported signals were identified. This detector identifies Amazon-applied signals only. Files are used for the purpose of providing this detector and will be deleted following completion of analysis.
    

    

        
What types of files are supported?

        This detector accepts image, audio, and video files in the following formats: PNG, JPEG, GIF, WEBP, MP4, MOV, AVI, MP3, WAV. Image files must be under 23 MB. Video and audio files must be under 50 MB. Files must be uploaded one at a time.
    

    

        
What signals does the detector look for?

        The detector identifies two categories of signals: C2PA Content Credentials, which attach cryptographically signed metadata to media files, and Amazon watermarks, which are imperceptible signals embedded directly into generated content.
    

    

        
Does this detector work on content from other AI providers?

        No. This detector identifies signals that our generative AI systems applied to content. If another provider applied C2PA Content Credentials or watermarks to content, this detector will not detect those signals.
    






    
Tips for best results

    For the most accurate results, follow these guidelines when preparing files for upload.

    
Images

    Where possible, upload the original file rather than a screenshot or re-saved version. Screenshots will remove C2PA Content Credentials (which are attached to the original file as metadata), though watermark signals embedded in the image itself may still be detectable. For images embedded in other content, crop closely around the single image before uploading. Avoid uploading files that contain multiple images composited together, as this may interfere with signal detection.

    
Audio

    Upload the original audio file where possible. Recordings of audio playback (for example, recording a speaker playing a clip) may degrade the watermark signal. Avoid uploading files that have been heavily edited, spliced together from multiple sources, or converted through multiple format changes.

    
Video

    Upload the original video file rather than a screen recording or re-capture. Trimming or light compression is generally acceptable, but significant re-encoding, overlaying additional content, or stitching multiple clips together may reduce detection accuracy. If the video contains AI-generated content within a larger non-AI-generated production, consider isolating the relevant segment before uploading.





    
Frequently Asked Questions


    
        
        
        

            
                What signals does the detector recognize?
            
            

                The detector recognizes two categories of provenance signals: C2PA Content Credentials and Amazon watermarks. C2PA Content Credentials are based on an open technical standard that attaches cryptographically signed provenance metadata to media files. When present, C2PA Content Credentials can indicate that content was created using one of our generative AI systems. Amazon watermarks are imperceptible signals embedded directly into generated content at the point of creation. The watermark signal is part of the content itself rather than attached as metadata.