New York law drives Amazon's AI people tagging rule

By Diana Trent · 2026-07-29

One metadata keyword, every store worldwide

Amazon notified third-party sellers in July 2026 that any product listing image or video containing photorealistic AI-generated people must carry a metadata tag before upload.^[1] The requirement applies immediately to all new uploads across worldwide stores, meaning sellers in Australia, Europe and elsewhere face the same obligation as those in the United States.^[2]

Sellers must add the keyword "contains-synthetic-performer" to the dc:subject (XMP) metadata field of each affected file using an IPTC-compatible editor before upload.^[1] The tag must be embedded in the file itself, not applied through Amazon's listing interface after the fact.

The one state law that set the standard

New York's S8420A, which took effect on 9 June 2026, was the first law in the United States to require that advertisements disclose the use of AI-generated synthetic performers.^[4] The legislation targets the entertainment and media advertising sector, where generative AI tools have made it increasingly practical to replace human talent with photorealistic digital stand-ins.

Governor Kathy Hochul said the law balances commercial flexibility with honesty toward audiences. "Requiring simple, honest disclosure when an ad uses synthetic performers protects consumers, respects our creative workforce and keeps New York at the forefront of responsible innovation," Hochul said.^[4]

New York State Senator Michael Gianaris framed the law as a worker-protection measure. "Film and television production remains an integral part of New York's economy, and that is why New York will continue to lead in protecting the workers who power this industry. As this legislation goes into effect, New York's performers will be better protected from their likenesses being deceptively replaced by artificial intelligence, and we can hold accountable those who would take advantage of these tools to profit improperly at the expense of the real talent," Gianaris said.^[4]

How the tagging process works in practice

Sellers working with AI-generated imagery need to open each affected file in an IPTC-compatible metadata editor, locate the dc:subject field under the XMP schema, and type the exact string "contains-synthetic-performer" before saving.^[1] Widely used tools such as Adobe Bridge and ExifTool support this workflow natively. The tag travels with the file on upload and signals to Amazon's systems that the asset contains a photorealistic AI-generated person.

The requirement covers listing images, listing videos and A+ content, making it one of the broader metadata obligations Amazon has imposed on third-party sellers.^[2] Sellers who miss the tag on a qualifying asset risk having the upload rejected or the listing penalised once enforcement mechanisms are in place.

What is and is not covered

The rule is narrower than it might first appear. It triggers only when a listing image or video contains a photorealistic AI-generated person, meaning a fully synthetic human likeness created by a generative model.^[3]

Images or videos featuring only real people, even where those people have been AI-altered, as well as fictional characters from films, television or video games, and assets containing no people at all, do not require the disclosure metadata.^[1] A seller who uses a real model and then smooths skin tone with an AI retouching tool is not caught by the rule. A seller who generates an entirely synthetic model in Midjourney or a comparable tool is.

Sellers outside the US are equally bound

Because the obligation extends to all worldwide Amazon stores, non-US sellers cannot treat this as an American compliance matter.^[2] An Australian apparel brand using a fully synthetic AI model in its listing photography must tag those files before uploading, even though S8420A is a New York statute with no direct legal reach outside that state.

New York's S8420A took effect less than seven weeks before Amazon issued its worldwide seller notification in July 2026, making the timeline from state law to global platform policy unusually short.^[4] A single US state disclosure law has, via Amazon's global policy, set the effective standard for how AI-generated people must be disclosed in product listings sold anywhere on the platform.

Diana Trent

Diana Trent writes about regulation, competition and the law as it meets technology. She reads the judgments and the regulator filings that most people skip, and finds the story in them.