PicBarn

Glossary · Image standards

XMP metadata

XMP (Extensible Metadata Platform) is an ISO standard, created by Adobe, for storing descriptive data such as keywords, creator and rights inside an image file, where editing tools and upload systems can read it.

XMP stores metadata as structured text embedded in JPEG, PNG, TIFF and other formats. It sits alongside the pixels and does not change how the image looks. The IPTC Photo Metadata Standard, the vocabulary photographers and news agencies use for captions, keywords and rights, is commonly carried in XMP, and keywords live in the Dublin Core field dc:subject.

Sellers meet XMP mainly through AI disclosure. Amazon asks for the keyword contains-synthetic-performer in the IPTC/XMP dc:subject field of any image that shows a photorealistic AI-generated person, and its tagging guide uses exiftool to write it. The IPTC also publishes a Digital Source Type vocabulary that includes a value for media created by a trained algorithm, trainedAlgorithmicMedia, used by some tools to mark generated images.

The common mistake is assuming metadata survives. Many steps strip it: screenshots, some compression tools, messaging apps, and exports with metadata switched off. A tag added early in the workflow can be gone by the time the file is uploaded. Read it back with exiftool on the final file.

Metadata is also not a substitute for correct pixels. Amazon's technical file requirements tell sellers not to rely on embedded ICC colour profiles, another kind of embedded data, so colour should already be right in RGB.

PicBarn's model exports carry the Amazon contains-synthetic-performer XMP tag.