Amazon Now Requires You to Label AI-Generated People—Is Your Brand Ready?

Amazon’s policy update introduces a formal requirement for sellers to disclose when product images, videos, or A+ Content include photorealistic people generated entirely by AI. Instead of restricting the use of AI-generated visuals, Amazon now requires specific metadata tagging so the platform can automatically apply a customer-facing disclosure. This shift means AI-generated human likenesses are no longer treated as standard creative assets. They must be explicitly identified at the file level before upload.

What Changed in Amazon’s AI Image Policy?

Amazon announced that sellers must tag media containing photorealistic AI-generated people before uploading it to listings or A+ Content in any of Amazon’s worldwide stores.

Sellers must:

• Use an IPTC-compatible metadata editor

• Add the keyword contains-synthetic-performer

• Place it in the dc:subject field within the file’s XMP metadata

• Ensure the metadata remains embedded when uploading to Amazon

Amazon will then add an indicator informing customers that the content includes AI-generated people where applicable.

This means the Amazon AI labelling requirement is handled primarily through information embedded within the image or video file, not through:

• Product titles

• File names

• Image alt text

• Bullet points

• Product descriptions

The update follows the implementation of New York’s synthetic-performer disclosure law, which took effect on June 9, 2026. Signed by Governor Kathy Hochul in December 2025, the legislation amended New York General Business Law Section 396-b and requires covered advertisements to conspicuously disclose the use of a synthetic performer.

Key points of the law include:

• A synthetic performer is defined as a digitally created asset intended to give the impression of a human performance without representing an identifiable natural performer

• A first violation may carry a $1,000 civil penalty

• Subsequent violations may carry a $5,000 civil penalty

Which Amazon Images Must Be Labeled?

Amazon’s requirement applies when a product image or video includes a person who is both:

• Photorealistic

• Created entirely through artificial intelligence

Examples may include an AI-generated model holding a skincare product, wearing an item of clothing, using a household appliance, or appearing beside a product in a lifestyle scene.

Background figures can also require attention. A synthetic person does not necessarily need to be the main subject of an image to fall within the policy. Creative teams should review background shoppers, family members, stylists, healthcare workers, or other realistic human figures added through generative tools.

The AI-generated people label Amazon requires is therefore narrower than a general “made with AI” disclosure. It does not apply to every image that has passed through an AI-powered editing tool.

Amazon says sellers do not need to provide the metadata when an asset:

• Only features real people, even when their appearance has been altered with AI tools

• Does not contain people

• Contains people who are not photorealistic

• Only features characters from movies, television programs, streaming content, documentaries, video games, or similar expressive works

These exclusions mean that basic retouching, background removal, color correction, lighting adjustments, and other AI-assisted edits do not automatically trigger the synthetic-performer requirement.

Sellers should still ensure that every asset accurately represents the product and complies with Amazon’s broader product image requirements.

How Sellers Must Tag AI-Generated People

Source: Pexels

For standard listing images and videos, sellers need an IPTC-compatible metadata editor capable of modifying XMP metadata.

The required workflow is:

1. Identify whether the asset contains a photorealistic person created entirely through AI.

2. Open the asset in a compatible metadata editor.

3. Add contains-synthetic-performer to the dc:subject XMP field.

4. Save the metadata within the original asset file.

5. Upload the tagged file to Amazon.

6. Confirm that the correct version was added to the listing or A+ Content.

The dc:subject property is commonly displayed as a “Keywords” field in photo-management and metadata applications. Sellers should verify that the keyword remains embedded after exporting, resizing, compressing, or transferring the file because some creative tools can remove metadata during processing.

Amazon’s instructions also apply to qualifying A+ Content assets. Sellers should therefore review enhanced product-description modules, Brand Stories, comparison charts, lifestyle banners, and videos, not only the main image gallery.

Amazon says certain tools may provide an upload option for identifying AI-generated content, while assets created through some Amazon AI tools may be tagged automatically. Sellers should not assume, however, that every generative platform or export workflow will add the required metadata.

Do Existing Amazon Listings Need to Be Updated?

Source: Pexels

Amazon representatives have clarified that compliance is mandatory going forward for new uploads. They have described updating older images as a best practice, particularly when a seller is already editing the relevant listing.

That does not mean brands should ignore their existing catalogs.

A sensible risk-based review should prioritize:

• High-traffic ASINs

• Recently redesigned listing images

• A+ Content created with generative AI

• Lifestyle images supplied by freelancers or agencies

• Product videos containing virtual models or presenters

• Assets reused across Amazon Stores, advertising, social commerce, and other channels

Sellers should also preserve the original creative files. If an uploaded image is later compressed or downloaded from Amazon, its original metadata may not be available for verification.

Why This Matters Beyond Compliance

AI-generated lifestyle content has made it easier for sellers to create multiple visual concepts without arranging a conventional photoshoot for every variation.

That efficiency remains valuable. The issue is that lower production costs do not remove the need for creative governance.

A customer-facing AI indicator may affect how shoppers interpret an image, particularly when the person appears to demonstrate product performance, results, fit, scale, safety, or suitability. The implications may be more significant in trust-sensitive categories such as beauty, health, baby products, apparel, and personal care.

For example, an AI-generated skincare model could imply a product outcome that was never photographed. A synthetic parent using a baby product might unintentionally suggest a level of safety validation. A virtual healthcare professional could create an appearance of authority that the brand cannot substantiate.

The central question is no longer simply whether a brand can create an AI model. It is whether that model communicates something accurate, useful, and appropriate for the product.

Important Considerations for Amazon Sellers

• Disclosure does not excuse misleading images: Metadata does not override Amazon’s core listing policies. Imagery must accurately represent the item without implying unsupported benefits or endorsements.

• AI-assisted is not AI-generated: Editing a real person with an AI tool is treated differently than generating a synthetic person from scratch. Document your creative processes carefully.

• Assign metadata ownership: When working with agencies, freelancers, or AI vendors, explicitly state who is responsible for identifying assets, adding the metadata, and preserving the compliant files.

• Scrutinize high-risk categories: A synthetic background figure in a household scene carries low risk. However, a photorealistic person demonstrating a supplement, medical device, or cosmetic product requires strict scrutiny.

• Seek legal guidance when necessary: Amazon’s technical labels are not a substitute for legal advice regarding advertising laws, intellectual property rights, or consumer protection requirements.
  

beBOLD Digital’s Take: AI Creative Now Needs Operational Controls

Amazon’s decision should not be interpreted as a signal that sellers must stop using generative AI.

It should be interpreted as a signal that AI creative can no longer be managed as an informal shortcut.

A seller may have images produced by an in-house designer, a listing agency, a freelance contractor, an AI image platform, and Amazon’s own creative tools. Without a centralized record, the seller may not know which assets contain fully synthetic people or whether the correct metadata survived the final export.

Brands should add several questions to their creative-approval process:

• Was any person in this asset generated entirely through AI?

• Is the person photorealistic?

• Does the image imply a product benefit, result, endorsement, or use case?

• Has the required metadata been embedded and verified?

• Who approved the asset?

• Where else is the asset being used?

• Is a real model, product-only image, illustration, or user-generated asset more appropriate?

This process turns AI disclosure from an emergency catalog cleanup into a routine quality-control step.

It also gives brands a stronger foundation for experimentation. Rather than assuming synthetic models will improve performance because they are faster or less expensive to produce, sellers can compare product-only photography, real-model photography, illustration, and AI-assisted lifestyle creative based on actual engagement and conversion data.

A Five-Step Amazon AI Image Audit

Sellers can begin with the following process:

1. Inventory the catalog

Create a list of listing images, product videos, A+ Content modules, Brand Stories, Amazon Store assets, and advertising creative.

2. Identify synthetic people

Flag photorealistic faces, hands, bodies, presenters, models, professionals, families, and background figures created entirely through AI.

3. Confirm the production method

Ask internal teams and external vendors whether each person was photographed, AI-edited, composited, illustrated, or fully generated.

4. Add and verify the metadata

Embed contains-synthetic-performer in the required XMP field and confirm that the metadata survives the final export.

5. Update the approval workflow

Add AI provenance and disclosure checks to every future creative brief, vendor agreement, asset-delivery checklist, and Amazon listing optimization review.

What Amazon’s Decision Signals for Ecommerce Brands

Amazon’s response reflects a broader movement toward provenance and disclosure standards for synthetic media.

For ecommerce brands, the long-term operational change maaiy be more important than the individual label. Product content teams will increasingly need to know how an image was created, which elements are synthetic, which disclosures apply, and whether embedded metadata remains intact as an asset moves between platforms.

Brands that establish those controls now will be better positioned to adapt as marketplaces, advertising networks, and state regulators introduce additional AI transparency requirements.

Amazon’s AI Image Rule Signals Mandatory Transparency for Listing Images

Amazon’s requirement doesn’t eliminate AI-generated creativity. It simply raises the standard for how it is managed. AI imagery now requires a structured process built around compliance, accuracy, and customer trust.

If your brand needs help navigating these new rules, auditing your current Amazon listings, or building compliant AI creative workflows, beBOLD Digital can help you implement these operational controls seamlessly. Contact beBOLD Digital today.

Frequently Asked Questions

Does Amazon ban AI-generated product images?

No. Amazon’s current requirement does not prohibit AI-generated product imagery. It requires qualifying images and videos containing photorealistic people created entirely by AI to carry specific metadata before upload.

What is the required Amazon metadata keyword?

The required keyword is contains-synthetic-performer. It must be added to the dc:subject field in the file’s XMP metadata using an IPTC-compatible editor.

Do AI-edited photos of real people require the label?

Amazon says the requirement does not apply when an image only features real people, even if their appearance has been altered using AI-powered tools.

Does the rule apply to A+ Content?

Yes. Amazon’s announcement specifically includes qualifying images and videos used in A+ Content as well as regular product-listing media.

Do AI-generated backgrounds need to be labeled?

An AI-generated background without photorealistic synthetic people does not fall within the stated requirement. Sellers should review the entire scene, however, because realistic background figures may qualify.

Must sellers update previously published images?

Amazon has said the requirement is mandatory for new uploads. Updating older qualifying assets is a prudent best practice, especially when a listing is already being revised.

Will every tagged image display a visible AI disclosure?

Amazon says it will add a customer-facing indicator where applicable. The company has not publicly detailed every condition that determines when or how that indicator appears.