Amazon's New Rule: Sellers Must Label 'AI-Generated Human Images' in Visual Content

[Ebrun Exclusive] July 26th, Amazon has recently issued a notification to third-party sellers, requiring them to disclose the presence of "realistic AI-generated human images" within product visuals and embed specific metadata tags in qualifying content. This move is directly triggered by New York State's newly enacted "Synthetic Performer" disclosure law, which officially took effect on June 9th.

According to the notification, sellers must provide clear disclosure when advertisements contain images or videos featuring realistic AI-generated human likenesses.

Specifically, using AI to generate a non-existent human model or creating a virtual character to showcase products falls under the regulatory focus. The platform also requires sellers to embed metadata keywords compliant with the IPTC standard to label "Synthetic Performers."

The new rule's applicability has several clear exceptions.

In the realm of creative production, two categories are exempt from disclosure requirements: first, fictional characters in movies, television, or video games; second, content that uses AI tools to modify the appearance of real individuals.

Consequently, operations such as AI beautification of real-person photos, background modifications, or adjustments to appearance details typically do not trigger the disclosure obligation for synthetic performers.

The legal basis for the above rules stems from New York State's "Synthetic Performer" disclosure law.

This law defines a "Synthetic Performer" as a digital asset created, replicated, or modified through generative AI or software algorithms, with the purpose of leading consumers to believe it involves a human performance by a person, where that person is not an identifiable real individual.

The regulation further stipulates that if the advertiser or publisher knows the advertisement contains a synthetic performer, they must disclose this fact to consumers in a prominent manner. Regarding penalties for violations, the fine for a first offense is $1000, escalating to $5000 for each subsequent offense.

In addition to requiring seller-side disclosure, Amazon also plans to add consumer-facing prompt labels on relevant product pages.

Some analysts suggest this approach may require weighing potential market reactions. Existing research indicates that content bearing AI labels often receives lower user engagement rates and may have certain impacts on consumer trust and purchase intent.

It is noteworthy that the trend of strengthening oversight of AI visual content is not unique to Amazon.

Over the past six months, major platforms like TikTok and YouTube have also been advancing AI content labeling policies, reflecting a broader global trend towards increased regulation of synthetic content.

For brands, balancing the efficiency gains brought by AI tools with potential shifts in consumer perception due to AI disclosure labels is becoming an ongoing challenge.

In response to these policy changes, industry insiders recommend that companies prepare from several angles.

In the review process, priority should be given to checking core best-selling products, main images, and brand flagship store pages, identifying which ASINs use AI-generated human images.

In terms of content management, metadata management should be integrated into the content production workflow, rather than being addressed post-upload as a remedy. If a company collaborates with advertising agencies or freelance designers, delivery requirements should clearly specify whether AI-generated elements are included, who is responsible for labeling, and how responsibility is allocated.

For record-keeping, as legal enforcement depends to some extent on subjective intent and knowledge, companies should systematically preserve documents such as creative briefs, AI prompts, and production instructions.

For example, if the design goal for the content explicitly aims for mannequins, cartoon characters, or abstract avatars, related records can support the claim that there was no intent to create the impression of a real human performer.

Regarding creative pathway choices, different approaches correspond to different costs and disclosure outcomes. Adopting a hybrid production strategy, such as "real models combined with AI-generated backgrounds," typically does not trigger the synthetic performer rule, which may become a viable option for some brands to balance efficiency within the compliance framework.

Ebrun continues to track this development. To learn more about information related to this article, please scan the QR code to follow the author on WeChat.