Amazon's AI Image Rule: You Must Now Label Synthetic People

In late July 2026, Amazon told third-party sellers they must disclose photorealistic AI-generated people appearing in product images, videos, and A+ Content. As reported on July 23, 2026 by CNBC and covered subsequently by Forbes, eWeek, and Quartz, the requirement is implemented through image metadata rather than a visible watermark, and Amazon says it will surface an indicator to customers where it applies.

This is not an Amazon-invented content rule. It follows New York General Business Law Section 396-b, which took effect June 9, 2026 and requires businesses that create commercial advertisements to disclose when those ads knowingly contain a synthetic performer — digitally created media that appears to depict a real person. As reported, the New York statute carries a civil penalty of $1,000 for a first violation and $5,000 for subsequent ones. Amazon's move pushes that compliance obligation down to the sellers who upload the creative.

If you have leaned on AI image generation for lifestyle shots — a model holding your product, a person using it in a kitchen, a face in a hero image — this now touches your listing workflow directly. Everything below reflects the policy as publicly reported as of August 2026; the authoritative version is the content compliance and image requirements documentation inside your own Seller Central account, so confirm there before you change your process.

"The compliance step here is thirty seconds of metadata work. The expensive part is not knowing which of the last two years of AI lifestyle images in your catalog actually contain a fully synthetic person."

— Tom Reiter, Product Research & Operations Specialist

So the real work is an audit, not a workflow change. Start with what does and does not qualify.

The scope is narrower than most sellers assume on first read. As reported, the requirement targets photorealistic people created entirely by AI and not based on a real person. Several categories of imagery that feel like they should be covered are explicitly outside it. The table is a plain-language summary of the reported guidance — verify each row against Amazon's own current documentation before you apply it at scale.

The mechanic itself is simple: as reported, you add the keyword contains-synthetic-performer to the dc:subject (XMP) field of the image file using an IPTC-compatible metadata editor, then upload the file as normal. Amazon reads the metadata on ingest. What takes real time is deciding which existing assets qualify and rebuilding your creative pipeline so the tag survives every export.

Sellers who never used generative imagery for people have nothing to do here. Sellers who used it heavily for lifestyle shots have a catalog audit — and possibly a creative-strategy question about whether synthetic models are still worth it once they are labeled as such.

Amazon spent the last two years shipping generative tools into Seller Central — AI listing copy, AI-assisted image generation, AI ad creative. This is the first widely reported instance of the platform adding a disclosure obligation back onto seller-supplied AI output, and the trigger was state law rather than platform preference. That direction of travel matters more than the specific keyword.

The practical read for 2026: keep provenance records for AI-generated creative. Knowing which assets were machine-generated, by which tool, and when, is now an operational asset rather than trivia. If a second jurisdiction adopts a comparable disclosure rule with a different mechanism, the sellers who can answer that question in an afternoon will comply in an afternoon.

It is also a reminder that AI listing tools are inputs to a compliance process, not a replacement for one. The tools accelerate production; the seller remains accountable for what ends up on the detail page. Choose tooling that tells you what it generated and preserves file metadata on export — that is a genuine differentiator now, not a nice-to-have.