Amazon makes sellers tag AI-generated people in images
Amazon now requires third-party sellers to embed a metadata keyword on product images with AI-generated people. What to tag, how, and why it spreads.
If you've been generating product photos with an AI model standing in for a real one, Amazon just made that a metadata problem. The company told third-party sellers this week that any product image or A+ content containing a photorealistic AI-generated person has to carry a specific tag before it's uploaded. This isn't a suggestion buried in a style guide — it's a listing requirement, and it lands on the sellers who move more than 60% of Amazon's units.
What actually happened
CNBC reported that Amazon notified sellers Wednesday of a policy change requiring disclosure of AI-generated people in product images and in A+ content — the enhanced graphics and video on listing pages. The mechanism is embedded metadata: sellers add the keyword contains-synthetic-performer to qualifying assets using an IPTC-compatible editor before upload. Amazon then uses that signal to surface a disclosure on the listing "where applicable."
The scope has edges. It covers photorealistic synthetic humans — the kind indistinguishable from a real person. It excludes TV, movie, and video game characters, and it doesn't cover real people whose appearance was touched up with AI tools.
The driver is New York General Business Law § 396-b, which took effect June 9 and requires advertisers who knowingly use a "synthetic performer" to disclose it. We wrote about that law when it landed — fines start at $1,000. What's new is the enforcement path: one state statute has become a platform-wide obligation for every seller, in every state, on the largest marketplace in the country.
Why it matters for your business
This is how compliance actually reaches small operators. Not via a regulator knocking, but via a marketplace changing its upload rules on a Wednesday. If you sell on Amazon and your creative pipeline includes a generated model holding your product, you now have a tagging step — and if you outsource photography or hire a freelancer to produce AI creative, that obligation flows to you, not them.
The practical trap is metadata stripping. Most image workflows — resizing scripts, CDN optimizers, Canva exports, a designer saving "for web" — quietly discard IPTC fields. You can tag an asset correctly and still ship an untagged file. Verify the tag survives your last export step, not your first.
And treat this as the template rather than the exception. Amazon moved first because New York moved first. Walmart, Etsy, Shopify's channels, and the ad platforms are looking at the same statute. The cheap version of this is a provenance field in your product data — one column that says whether an asset contains a synthetic person — so the next platform's rule is a query, not a re-audit of 400 SKUs.
Key takeaways
- Amazon now requires the metadata keyword
contains-synthetic-performeron product images and A+ content depicting photorealistic AI-generated people - Sellers apply it with an IPTC-compatible editor before upload; Amazon renders a disclosure on the listing from that signal
- Excluded: TV/film/game characters, and real people whose appearance was AI-altered
- It stems from New York's synthetic performer law (effective June 9), but Amazon's rule applies to sellers everywhere
- Check that your export pipeline preserves IPTC metadata — most resize and "save for web" steps strip it
Don't audit 400 SKUs by hand. We build product-data pipelines that carry provenance fields through every export, so the next platform rule is a query instead of a fire drill. See how we build it or send us your catalog problem.
- #amazon
- #ai-marketing
- #compliance
- #synthetic-media
- #product-listings
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
Agentic AI spending: $944B now, $3.35T by 2030 — read it right
A new PHD/WARC forecast puts agent-facilitated consumer spending at $944B in 2026, rising to $3.35T by 2030. That's 3.8% of consumer spend. Here's the operator read.
Read itInstacart buys Arpalus: who owns your shelf data?
Instacart acquired computer-vision startup Arpalus to turn 600,000 shoppers into a shelf-intelligence sensor network. If they run in your stores, they know your inventory better than you do.
Read it