Amazon's New AI Image Rule: The Metadata Sellers Have to Add in 2026

Amazon now wants sellers to tag photorealistic AI people in their listing images before upload. Here is what the rule says, how to do it, and how to keep it from becoming busywork on every listing.

The 30-second version

•  What changed: If a product image or video shows a photorealistic, fully AI-generated person, you have to write the keyword contains-synthetic-performer into the file's metadata before you upload it to Amazon.

•  Why: A New York disclosure law, the first of its kind, took effect in June 2026. Amazon turned the rule on across its whole marketplace.

•  The catch: Seller Central cannot do this for you. The tag goes into the file when the image is made, so it is a production job, not a quick fix at upload.

•  The easy path: Use an imagery partner that writes this metadata for you, so the listing is already compliant before anyone on your team touches it.

What Amazon actually changed

On July 22, 2026, Amazon updated its Policy and Compliance guidance. Sellers now have to disclose when a product image, video, or A+ Content asset shows a photorealistic AI person. The rule is live across Amazon's stores worldwide.

Amazon is specific about how. You use an IPTC-compatible metadata editor to add the keyword contains-synthetic-performer to the dc:subject (XMP) field of the file, and you do it before the asset goes on your listing. Once the tag is there, Amazon adds a note on the listing telling shoppers the content includes AI-generated people, where it applies.

One detail gets lost in most write-ups. The tag lives in the file's metadata, not in the picture. Shoppers never see it in the image. Amazon reads it at upload and decides whether to show a disclosure. So the whole job comes down to how the asset gets made. Nothing in Seller Central will fix it after the fact.

Why Amazon is doing this now

The rule comes from a New York State law, the first of its kind in the country. It says businesses have to disclose when an ad uses a realistic AI-generated performer instead of a real actor. It took effect in June 2026, and it carries penalties that climb for repeat violations.

Amazon did not build a version that only applies in New York. It turned the requirement on for the whole marketplace. That is the part worth watching. One state law became a marketplace-wide rule in a matter of days. Other retailers will probably do the same, so it is cheaper to solve this cleanly once than to patch it platform by platform later.

What needs the tag, and what does not

The rule is narrower than “all AI imagery.” It only covers photorealistic people who are fully AI-generated. Here is the line Amazon draws:

Needs the tag	Does not need the tag
•  Photorealistic, fully AI-generated people in product images •  AI-generated models wearing your product •  AI-generated people in lifestyle shots •  AI-generated people in product and shoppable videos •  AI-generated people in A+ Content	•  Real people, even if you edited them with AI •  Characters from movies, TV, streaming, or video games •  Product-only images with no people •  Illustrated or clearly non-photorealistic people

Quick test: if a shopper could look at the image and think they are seeing a real human model, and that person was generated by AI instead of photographed, it needs the tag.

How to comply, step by step

Doing this in house? Here is the workflow. If you would rather never open a metadata editor, jump to “The simplest way to stay compliant” below.

1. Audit your catalog. Find every listing with images or videos that use photorealistic AI people. Start with your high-traffic ASINs and A+ Content, since those carry the most risk.

2. Get an IPTC-compatible metadata editor. You need a tool that can write XMP metadata. Seller Central will not add the tag for you.

3. Write the exact keyword. Put contains-synthetic-performer in the dc:subject (XMP) field. Spelling and field matter, because Amazon reads the string automatically.

4. Do it during production, not after. Bake the tag into the file while the asset is being made. Going back to tag a live catalog one image at a time is how teams burn days.

5. Re-upload the fixed assets. Swap the affected images and videos in your listings and A+ Content for the tagged versions.

6. Keep a record. Note which assets were AI-generated and how you tagged them. If the rules grow, you will want the paper trail.

One caveat: this is guidance, not legal advice. The New York law has real penalties, and the rules are moving fast. Have your legal or compliance team confirm how it applies to your listings.

What this really means for brands and agencies

On paper, it is “add one keyword.” In real life, if you run AI lifestyle imagery at any scale, it drops a new recurring step into your listing process. A few things make it more annoying than it sounds:

•  It repeats on every asset. Each qualifying image and video needs the tag, every single time.

•  Amazon cannot automate it for you. The work sits in your image pipeline, before upload, outside Seller Central.

•  Your back catalog may need fixing. Live AI imagery might have to be re-tagged and re-uploaded.

•  It probably will not stop at Amazon. State laws are driving this, so expect other retailers to follow, each with its own quirks.

The teams who feel this the most built their setup out of spare parts. One tool makes the image, another edits it, and someone has to remember the metadata step. Every new rule turns into a scramble.

The simplest way to stay compliant: make it part of the pipeline

Here is the shift that saves the most time. This is a production question, not an upload question. If the metadata goes into the file the moment the image is created, nobody has to remember it later. It just ships that way.

That is how we built Flock. We are a full-service AI imagery partner for brands and retailers, and writing XMP metadata, including the contains-synthetic-performer tag, is already part of how we produce assets. For the brands we work with, “are we compliant with Amazon's new rule?” is already answered.

We own the whole production chain: the imagery, the metadata, the brand standards, the disclosures. So a change like this gets handled before it ever reaches your team. No new software to buy. No listing by listing scramble. No hand-tagging a year of old assets. And when the next retailer ships its own version of the rule, you are already set.

Why a full-service partner beats a DIY generator here

•  Compliance happens at the source. The metadata is built in, not bolted on.

•  One partner you can hold accountable, instead of five tools and a checklist.

•  Photorealistic, on-brand imagery produced like real work, not quick exports.

•  We watch the rule changes so your team does not have to.

Frequently asked questions

Does this apply to every AI image?

No. It only covers photorealistic people who are fully AI-generated. Product-only shots, illustrations, and images without people are fine.

What about real models edited with AI?

Those are fine. The rule is about people who are fully AI-generated, not real people you retouched or enhanced.

Can I add the metadata in Seller Central?

No. There is no field for it. The keyword has to go into the file's XMP metadata with an

IPTC-compatible editor before upload, which is why it is really a production task.

Will shoppers see a label?

Where it applies, yes. When Amazon spots the metadata, it adds a note that the content includes AI people. The tag itself stays invisible in the image.

What if I ignore it?

You would be out of step with Amazon's policy and the New York law behind it, which carries fines. And undisclosed AI imagery is a trust problem with shoppers if it comes out.

Will other retailers want this too?

Probably. It comes from state law, and platforms tend to match the strictest rule they have to meet. Building metadata into your workflow now saves you from redoing it for the next one.

Does Flock do this automatically?

Yes. Writing the required XMP metadata is already part of how we produce imagery, so what we deliver is tagged before it reaches your listings.

Let Flock handle the compliance part

See how we deliver photorealistic, on-brand AI imagery with the right metadata already inside, so rules like this get handled before they reach your team.

Book a Demo >> Flockshop.ai