Amazon's New AI Model Disclosure Rule: Why Every Photorealistic AI-Generated Person in Your Listings Now Needs a Metadata Tag

Amazon told third-party sellers on July 22, 2026 that any product image, A+ Content module, or video containing a photorealistic AI-generated person now needs a specific metadata tag before it can be uploaded. The rule reaches every one of Amazon's worldwide stores, not just the US, and it traces directly back to a New York disclosure law that took effect June 9. If your listings use AI-generated models instead of real photography, this is not a suggestion you can quietly skip.

What Actually Changed

The trigger is New York General Business Law Section 396-b, which requires any commercial advertisement to clearly disclose when it uses a "synthetic performer," defined as a digital asset created or modified by generative AI to give the impression of a real human in a visual performance. Because Amazon's marketplace reaches New York buyers regardless of where a seller is based, Amazon rolled the requirement out platform-wide rather than try to geofence it.

The mechanism is a metadata tag, not a visible badge on the listing page. Sellers add the keyword contains-synthetic-performer to the dc:subject field in an image or video file's XMP metadata using an IPTC-compatible editor, then upload the file as usual. Amazon reads that tag on ingestion and shows its own disclosure indicator to shoppers when applicable. You are not writing customer-facing copy here. You are stamping the file before it ever reaches Seller Central.

What Does and Doesn't Require the Tag

Requires it: any photorealistic person generated entirely by AI who appears in a product image, A+ Content module, or brand video.
Does not require it: real people, even if AI tools were used to edit the photo (background removal, lighting correction, retouching); illustrated or non-photorealistic characters; AI-generated backgrounds with no AI-generated person in frame; and characters from film, television, or games.

That last line matters more than it looks. A lifestyle shot with a real model photographed against an AI-generated kitchen background does not need the tag. The same shot with an AI-generated model against a real photographed kitchen does. The test is whether the person is synthetic, not whether AI touched the file at all.

Why This Is Bigger Than One State's Law

A meaningful share of sellers, particularly in apparel, beauty, and supplements, have leaned hard on AI-generated models over the last eighteen months. It is cheaper than a photo shoot, faster to iterate across sizes and skin tones, and doesn't require booking talent. None of that changes under the new rule. What changes is that the practice now has to be disclosed at the file level, and Amazon is the one enforcing it, not New York's Attorney General directly.

The penalty structure sits with the underlying law: a $1,000 civil fine for a first violation and $5,000 for each one after that, and the obligation to disclose falls on the seller or brand owner, not on Amazon. Amazon's own enforcement lever is separate and more immediate, since a photorealistic AI model image uploaded without the tag is simply non-compliant content sitting on your listing, the kind of thing that draws a content violation notice the same way a missing safety warning would.

There's also a discovery angle worth naming plainly. Once Amazon can identify which listings run AI-generated people at scale, that data doesn't disappear after the disclosure requirement is satisfied. It's a reasonable bet that listing quality signals, return rates, and content-authenticity scoring start factoring in whether the imagery is synthetic, even before Amazon says so explicitly.

Who Should Be Paying Attention Right Now

Brands using AI model-generation tools (the kind that put a "model" in different poses or body types without a real photo shoot) for any current or recently uploaded listing images.
Agencies and freelance designers producing A+ Content or brand story modules on a seller's behalf, since the disclosure obligation follows the seller even when someone else built the file.
Sellers running video ads or Sponsored Brands video creative that use AI-generated spokespeople rather than hired talent or founder-recorded footage.
Anyone who batch-uploaded a large catalog refresh in the last few months using an AI content pipeline and hasn't gone back to check which files include synthetic people.

What to Do This Week

Audit your live catalog for AI-generated model imagery. Don't rely on memory, actually open the product image folders and A+ Content modules and look, since this often accumulates across multiple contractors and tools over time.
Get an IPTC-compatible metadata editor into your content workflow now, whether that's a standalone tool or a plugin for whatever DAM or editing software your team already uses, and make tagging a required step before any file gets uploaded, not an afterthought.
Re-tag and re-upload existing non-compliant images rather than waiting for a violation notice to force the issue. Amazon's ingestion reads the tag at upload time, so an untagged file already live doesn't get grandfathered in just because it predates the announcement.
If you outsource content production, put the tagging requirement in writing with every agency or freelancer touching your listings. A vendor who doesn't know the rule exists will keep shipping you non-compliant files.

How TKL Helps

We audited every client catalog for AI-generated model imagery as soon as Amazon's notification went out, flagging which SKUs needed the metadata tag and which ones were already clear because the imagery used real photography. For clients whose content pipeline includes AI-generated models, we've built the tagging step directly into our image production workflow so nothing goes live without it.

If you're not sure whether your current listings or A+ Content carry AI-generated people that need disclosure, send us your catalog. We'll tell you exactly which files are exposed and get them tagged before it turns into a compliance notice.