Amazon Product Video Requirements: the AI avatar tag, and the 10-day lead time

The specification, as published

The AI condition is identical to the image rule

Amazon uses the same mechanism for video as for images: a photorealistic AI-generated person must be declared by writing the keyword contains-synthetic-performer into the dc:subject field of the file's XMP metadata, using an IPTC-compatible editor. Amazon then adds a disclosure to the video where the law requires one.

The exclusions are the same list as well. Real people need no tag, even when altered with AI tools. Characters from films or games need no tag. Videos with no people need no tag, and neither do people who are not photorealistic.

This is the part that matters for anyone generating video rather than filming it. A talking-avatar tool produces exactly the case the rule is written about. The output is not just a video file — it is a video file plus a metadata field, and the tool is not going to add that field for you.

What gets a video rejected

• External website links or URLs anywhere in the video.
• Promotional calls to action, or promotional content of any kind.
• Unauthorised music, artwork, or any audiovisual work you do not have permission to use.
• Visible third-party branding, or external branding of any kind.
• Misleading claims or incorrect product information.
• Blurry or low-quality footage, or a poor-quality thumbnail.

The timing nobody plans for

Amazon's own figures: initial review takes 1 to 7 business days, and publication to the detail page happens within three days after approval. That is up to ten business days between upload and a live video, and Amazon notes the timelines can run longer before and during holiday shopping seasons.

For a seller launching a product on a fixed date this is the constraint that decides the schedule, not the editing. A generated video does not shorten the review queue — the queue is the queue regardless of how the file was produced.

The softlines trap

On softlines products, a listing with six or more images suppresses the video in the desktop view. Amazon's own troubleshooting section lists removing excess images as the fix. Amazon elsewhere recommends six additional images plus one video, so on softlines those two pieces of guidance pull in opposite directions and the video is what loses.

This is the kind of rule that is invisible until someone asks why the video is not showing. It is not a rejection and nothing in the dashboard says the video is being suppressed.

What this means if the video was generated

Three of the rejection triggers above are ordinary editorial mistakes that AI generation makes more likely rather than less: promotional calls to action written into a script, a brand logo left in frame from a reference image, and music lifted from the model's training output without a licence.

The fourth is specific to generation. A synthetic presenter is a photorealistic AI-generated person, so the file needs the tag before upload. A generated voiceover over footage of the real product does not. Generated scenery with no people in it does not.

The practical rule: decide whether a synthetic person appears on screen, and if one does, treat the metadata step as part of the export — not as something to fix after Amazon reviews it.

Frequently asked

Sources

Every figure above comes from the pages below, read on the date shown. Marketplace policies change without announcement, so if you are about to rebuild a catalogue on one of these numbers, open the source and confirm it still says what it said.

• Amazon Seller Central — Shoppable video guide — read Oct 28, 2026
• Amazon Seller Central — Product image guide (the same AI-person keyword) — read Oct 28, 2026