Amazon starts labelling AI generated People
2 August 2026
Paul Adams – Chief Growth Officer
Amazon has told third-party sellers that any product image or A+ content featuring a photorealistic AI-generated person must be tagged before upload. The seller needs to add a keyword ‘contains-synthetic-performer” to the image’s XMP metadata. Amazon then uses that tag to surface a disclosure to shoppers “where applicable” though it hasn’t yet said what triggers the visible label or how prominent it will be.
The scope is narrower than the headlines suggest. It covers fully synthetic humans that aren’t based on a real person. Real models edited with AI tools are allowed. So are non-photorealistic illustrations and TV, film or game characters. Background figures in lifestyle scenes are in.
This move has been triggered by New York General Business Law effective 9 June 2026, requiring advertisers to disclose “synthetic performers” with penalties of $1,000 for a first breach and $5,000 thereafter. Amazon isn’t alone. TikTok Shop, Meta, Pinterest and YouTube all run AI labelling of some kind, and California’s AI Transparency Act phases in from August, with similar platform obligations .
What Amazon doing here is compliance plumbing, not a crackdown, and for now isolated to a few US states. It is to early to say whether this regulation will become more widespread. The more interesting question is where AI images make commercial sense. Early research suggests AI-labelled creative sees lower trust, authenticity and purchase intent, particularly on considered purchases. Brands leaning on synthetic models to cut shoot costs should be testing labelled versus unlabelled creative now, auditing what’s already live, and pushing tagging upstream into their DAM rather than fixing it at upload.