A distinction for judging whether AI-generated imagery containing synthetic human figures requires disclosure: the same technique is compliant when it demonstrates product function (scale, fit, assembly — pure information) and deceptive when it's arranged to imply a real satisfied customer (testimony/social proof that doesn't actually exist). The dividing line isn't "AI vs. real," it's the implied claim behind the image.
First observed in Amazon's 2026 AI-generated-image labeling rule (via Incrementum Digital), which requires a metadata tag and shopper-facing indicator only on photorealistic AI-generated PEOPLE, following New York's synthetic-performer disclosure law. Imagery without people, and real people digitally retouched (lighting, background cleanup), are exempt; a real person placed into a scene they never shot counts as fabrication requiring consent regardless of AI involvement.
Apply: when generating AI imagery for listings, Amazon A+ Content, or marketing, confine synthetic human figures to demonstration roles (scale, fit, assembly) and never assemble synthetic "customers" to simulate social proof — that crosses from information into testimony and triggers disclosure obligations under this and similar rules.