contains-synthetic-performer

Q: Do I have to go back and tag images that are already live?

Amazon states compliance is mandatory going forward for new uploads; older images fall under a best-practice recommendation to update them while editing, with no retroactive deadline. Editing the local file does not change a live listing — you have to re-upload it.

Q: Does this cover A+ content and product video?

Yes. Listing images, images and video inside A+ content, and product video are all in scope. Amazon has so far only specified the metadata field for images; the field for video has not been published.

This is the metadata keyword Amazon asks sellers to use. If a listing image or video contains a photorealistic person generated entirely by AI, the keyword has to be embedded in the file's XMP metadata before it is uploaded to Seller Central. Amazon reads it and adds a disclosure on the detail page telling shoppers the image includes AI-generated people.

This is why re-compressing an image can lose the tag while the picture itself looks unchanged: the two live in different parts of the file.

Which field it goes in

The XMP dc:subject field. That is the standard IPTC/XMP subject-and-keywords field — in most image tools it simply shows up as “Keywords” or “Tags”.

Adding it to a single file (three free options)

Windows File Explorer

Right-click the image → Properties → Details → type contains-synthetic-performer into Tags → Apply.

macOS Preview

Open the image → Tools → Show Inspector → Keywords tab → click + → type the keyword → Enter.

exiftool

exiftool -XMP-dc:Subject+=contains-synthetic-performer yourfile.jpg

Use += rather than = so the keyword is appended to the existing list instead of replacing it.

What about a few thousand files?

All three methods above handle one file at a time. For a whole catalogue, drop the folder into the workspace: it tags everything in one pass and exports a ZIP. Writing happens in your own browser, so nothing is uploaded and there is no file-count limit.

Open the workspace →

Verify after tagging. Retouching software, compression tools, asset managers and agency portals can drop embedded metadata when exporting — a tag you wrote is not necessarily a tag that survives to upload time. Drop the file into the checker to be sure.

Tagging many files from the command line

exiftool takes a folder as readily as a file, which covers most bulk cases without any other tool:

# every JPEG in a folder, keeping existing keywords
exiftool -XMP-dc:Subject+=contains-synthetic-performer -ext jpg DIR

# recurse into subfolders, and skip exiftool's _original backups
exiftool -r -overwrite_original -XMP-dc:Subject+=contains-synthetic-performer DIR

Two flags worth understanding before running that on a real catalogue. -overwrite_original edits in place instead of leaving a _original copy beside every file — convenient, and irreversible, so keep a backup the first time. And += rather than = appends, which matters if your images already carry keywords for search or asset management.

Confirming the tag actually landed

Writing a tag and having a tag are different claims. To read it back with exiftool:

exiftool -XMP-dc:Subject yourfile.jpg

Or drop the file into the checker, which reads the same field and needs nothing installed. Do this on the file you are about to upload, not the one you tagged — those are often not the same file, and the steps in between routinely drop metadata.

Mistakes that quietly produce nothing

Writing to the wrong field. IPTC Keywords and XMP dc:subject are related but not identical, and some tools write only the legacy IPTC block. Amazon specifies the XMP field.
Tagging the master, exporting a copy. The export is a new file. Unless the export step preserves metadata — and most do not — the tag stays behind on the master.
Trusting a visible label. A burned-in “AI-generated” caption is pixels. It survives everything and proves nothing about the metadata Amazon reads.
Assuming PNG behaves like JPEG. It does not. In testing, every tool that preserved the tag in JPEG still lost it in PNG.

What needs the tag, and what does not

Content	Tag needed?
Photorealistic people generated entirely by AI (AI models, AI lifestyle figures)	Yes
AI-generated photorealistic bystanders in the background	Yes
Photos of real people, even when retouched or background-swapped with AI	No
Cartoons, illustrations and other clearly non-photorealistic figures	No
Characters from film, television or games	No
Product-only images with no people in frame	No

Common questions

Do I have to go back and tag images that are already live?

In its seller forum reply, Amazon stated that compliance is mandatory going forward for new uploads; for older images it is a best-practice recommendation to update them while editing, with no retroactive deadline set. Worth knowing: editing the local file does not change a live listing — to get a tag onto an existing image you have to re-upload it.

Does this cover A+ content and product video?

Yes — listing images, images and video inside A+ content, and product video are all in scope. That said, Amazon has so far only specified the field for images; which metadata field video should use has not been published, which is why video handling here is marked experimental.

Will the disclosure hurt conversion?

Research suggests AI-labelled content takes a hit on consumer trust and purchase intent. If that matters a lot for your category, the more fundamental fix is a creative one — use AI for backgrounds and scenes and keep human figures real, so the requirement never triggers.