In July 2026, Amazon began requiring third-party sellers to add a compliance tag — contains-synthetic-performer — to any product image, video or A+ content that features AI-generated people, and to show shoppers an on-page indicator that the content contains AI-generated people. If you sell on Amazon and you've used AI-generated virtual models, this rule applies to you directly.

This guide lays it out: where the rule came from, what exactly it asks for, which images need tagging and which are exempt, where the tag actually lives inside the file, and how to process a whole batch quickly.

Where the rule came from: New York's "synthetic performer" law

Amazon's AI Image Rules (2026): The contains-synthetic-performer Tag Explained — visual 1
Where the rule came from: New York's "synthetic performer" law

Amazon's move follows a newly effective New York law. It requires that when an advertisement uses a "synthetic performer" — a digitally generated figure made to look like a real person — in place of a real human, that must be disclosed to consumers. The law targets "digitally-created media that appear as a real person."

Amazon, as the platform, has passed that obligation to the seller side: you declare it at upload, and Amazon surfaces the indicator to shoppers. In other words this isn't a one-off marketplace whim — there's actual legislation behind it, so it isn't going away soon.

The requirement: write one keyword into the metadata

Amazon's AI Image Rules (2026): The contains-synthetic-performer Tag Explained — visual 2
The requirement: write one keyword into the metadata

In practice the ask is specific: before you upload, write the keyword contains-synthetic-performer into the image's metadata.

"Metadata" here isn't the filename — it's descriptive information embedded inside the image file. The two standard fields are XMP dc:subject and IPTC Keywords — the same "Tags" row you see under Windows' Properties → Details. Different systems read different fields, so the safe move is to write both.

Not a single character can be off. contains-synthetic-performer must be written verbatim; a stray capital, a missing hyphen, or synthetic-performer alone and the platform may not read it.

What to tag and what not to tag

The test is clear: is the person in the frame conjured by AI?

Needs tagging:

  • Virtual models generated with tools like Midjourney or PixPix
  • AI face-swap or AI outfit-swap model shots
  • Fully synthetic human product images

Doesn't need tagging (Amazon explicitly exempts):

  • Photos of real human models, even if AI-retouched
  • TV, video-game and movie characters

When unsure: if the person was generated by AI, tag it; if a real person was photographed, don't. For finer edge cases — AI-swapped backgrounds only, AI touching just part of the frame — see Which images need an AI label.

What happens if you don't

An untagged image later flagged as using an AI model can trigger manual review, image rejection, or search suppression. For cross-border sellers this is a hard compliance step, not a nice-to-have. Tagging on the way out beats a forced clean-up later.

How to comply fast

Editing metadata one file at a time isn't realistic. The right approach is batch: drop a whole folder into a tool, write the tag in one pass, verify each file, then download and upload. For the exact steps and how to verify afterward, see the walkthrough: How to batch-tag AI images for Amazon.

If you don't have AI model shots yet, or want a fresh set, you can generate product model images with AI first, then come back and tag them — two steps to both get the shots and list compliantly.

In short

  • Amazon's rule stems from New York's "synthetic performer" disclosure law — it has legislative backing;
  • It requires writing contains-synthetic-performer into image metadata (XMP + IPTC) before upload;
  • AI-generated virtual people need tagging; real-person photos (incl. AI-retouched) and TV/game/movie characters are exempt;
  • Skipping it risks rejection and search suppression; a batch tool clears a whole set in minutes.