Amazon AI image labels will flag synthetic people in seller ads

Amazon AI image labels are coming to some marketplace listings after New York enacted a law requiring disclosure when ads use digitally created people instead of human actors. For investors, the change shows how fast AI rules are reaching Amazon’s seller base, which drives more than 60% of goods sold on its marketplace.

Amazon told third-party sellers Wednesday that product photos, videos and other listing graphics must be marked when they include “AI-generated people,” according to an announcement CNBC said it reviewed. The company said sellers will need to add specific metadata keywords before uploading those files.

Metadata is behind-the-scenes information attached to a file. In this case, Amazon is using it to identify listing content that may need a consumer-facing disclosure.

What does Amazon's AI image label rule require?

Amazon’s rule applies to seller-uploaded product images and “A+ content,” a term Amazon uses for enhanced listing material such as videos and other graphics on product pages. If that content shows photorealistic AI-generated people, sellers must tag it before upload, according to the company’s notice cited by CNBC.

Amazon told sellers that recent legislation requires disclosure when advertising images or videos include realistic AI-created people. The company also said it will add an indicator to listings “where applicable” to tell shoppers when content includes AI-generated people, according to CNBC.

The exact standard Amazon will use to decide when shoppers see that indicator is not clear. Amazon did not immediately provide a comment to CNBC.

Why New York's law matters

New York’s law took effect last month and requires companies to disclose when a “synthetic performer” appears in an advertisement in place of a human actor. The law applies to digitally created media that looks like a real person, according to the office of Gov. Kathy Hochul.

Hochul’s office called the measure a first-of-its-kind law and said unlabeled synthetic performers and manipulated media can make it harder for the public to tell what is real. The rule targets advertising, a field where AI-generated people can appear as models, presenters or influencers without a real person ever being filmed.

Amazon’s seller notice also drew boundaries around the policy. The company said the requirement does not apply to content featuring characters from TV, movies or video games. It also does not apply to content featuring real people, even if AI was used to alter the image or video, according to CNBC.

How AI is changing Amazon listings

Amazon has been adding AI across its shopping business. CNBC reported that the company has used AI to improve listing titles and details, invested in its rebranded Alexa for Shopping assistant, and introduced a feature that can place AI-generated products into the search bar in real time based on shopper queries.

Third-party sellers are also using AI more often to create listing text, images and other content, including through Amazon’s own seller tools, according to CNBC. That makes disclosure rules more relevant to the day-to-day mechanics of Amazon’s marketplace.

There is no federal U.S. law requiring companies to disclose AI-created advertising content, CNBC reported. States have started setting their own rules. California, for example, began requiring large AI providers earlier this year to embed watermarks in AI-generated images, video and other content.

Major platforms have also moved toward AI labels. Meta, TikTok, Pinterest and Google’s YouTube have added labels for AI-generated images and videos uploaded to their services, according to CNBC. TikTok and Meta have faced criticism over ads featuring AI-generated influencers promoting questionable products, with TikTok saying it has acted against accounts making misleading health claims and Meta saying it labels AI videos.

This story draws on original reporting from CNBC.