Where the keyword is stored
The value is written as an item in the Dublin Core dc:subject RDF Bag inside an XMP packet. For JPEG files, XMP normally lives in an APP1 segment. For PNG files, XMP can be stored in an iTXt chunk.
A valid result keeps any existing dc:subject keywords and appends the disclosure keyword only when it is missing.
<dc:subject><rdf:Bag><rdf:li>contains-synthetic-performer</rdf:li></rdf:Bag></dc:subject>
What the tag does and does not do
The tag provides a machine-readable disclosure signal. Amazon may use that signal to inform audiences that an image includes AI-generated people where applicable.
The keyword does not prove how an image was made, detect AI content, change visible pixels, or guarantee that content will be accepted. The person uploading the file remains responsible for deciding whether the rule applies.
How to verify the keyword
Use an IPTC/XMP-compatible metadata editor or the browser-based tagger on this site. A reliable workflow writes the keyword, reads the file again, and confirms that dc:subject contains the exact value.
- Check the exact spelling, including both hyphens.
- Confirm existing keywords are still present.
- Confirm the compressed JPEG scan data or PNG IDAT data did not change.
Common mistakes
Adding the phrase to alt text, a filename, or visible copy does not place it in XMP dc:subject. Re-exporting an image after tagging can also remove metadata, so verify the final file that will be uploaded.