Product Research · Aug 7, 2026 · 7 min

Amazon AI-Generated Image Labeling: What Sellers Must Do Before Upload (2026 Compliance Checklist)

Sean Travis

Founder · Kaldon

TLDR

As of June 9, 2026, Amazon requires sellers to label photorealistic AI-generated people in all listing images, A+ Content, and videos using the metadata keyword `contains-synthetic-performer`. The rule applies to fully synthetic humans, not real models edited with AI or product-only shots. Sellers must embed the tag in IPTC/XMP metadata before upload or use the in-product checkbox for A+ modules. Retroactive updates to existing catalogs are mandatory, and missing the tag can trigger listing rejection or account-health penalties.

TLDR. As of June 9, 2026, Amazon requires sellers to label photorealistic AI-generated people in all listing images, A+ Content, and videos using the metadata keyword contains-synthetic-performer. The rule applies to fully synthetic humans, not real models edited with AI or product-only shots. Sellers must embed the tag in IPTC/XMP metadata before upload or use the in-product checkbox for A+ modules. Retroactive updates to existing catalogs are mandatory, and missing the tag can trigger listing rejection or account-health penalties.

What Amazon’s AI-Generated Image Labeling Requirement Means for Your Catalog

As of June 9, 2026, Amazon requires sellers to label photorealistic AI-generated people in all listing images, A+ Content, and videos using the metadata keyword contains-synthetic-performer. The rule applies to fully synthetic humans, not real models edited with AI or product-only shots. Sellers must embed the tag in IPTC/XMP metadata before upload or use the in-product checkbox for A+ modules. Retroactive updates to existing catalogs are mandatory, and missing the tag can trigger listing rejection or account-health penalties.

The policy was triggered by New York’s synthetic-performer disclosure law, effective June 9, 2026. Amazon notified sellers on July 23, 2026, and enforcement is already live across main images, secondary images, A+ Content, Brand Story, Store images, product videos, A+ videos, Store videos, and advertising videos. Consumer surveys cited in July 2026 coverage show 74% of shoppers think retailers should disclose AI-generated content, and 68% trust brands more when AI product photos carry disclosure labels. The market is framing this as a transparency issue, not a creative limitation.

If you are evaluating new products using Amazon’s demand validation framework, the labeling requirement now affects visual asset planning. Categories that rely on lifestyle imagery with human models (apparel, beauty, home fitness, pet care) will need either real photography or clearly labeled synthetic images. That choice should be part of your pre-inventory research, not a post-launch fix.

This checklist covers the operational steps sellers must complete before upload, how to handle existing content, and what the rule means for product research decisions.

What Counts as AI-Generated for Amazon’s Labeling Rule

Amazon’s rule applies to photorealistic AI-generated people. It does not apply to real people edited with AI, fictional characters from movies or games, stylized art, or images without people.

Here is the line Amazon is drawing:

Requires labeling: A fully synthetic human created by AI (e.g., a model generated by Midjourney, Stable Diffusion, or Amazon’s own generative tools) that appears in a product photo, lifestyle shot, or video.
Does not require labeling: A real model photographed normally and then edited with AI (background swap, color correction, object removal). The person must be real, not synthetic.
Does not require labeling: Product-only shots with no people, even if the background or product itself was AI-enhanced.
Does not require labeling: Fictional characters from licensed IP (e.g., a Marvel character, a video game avatar). Amazon treats these as creative works, not synthetic performers.
Does not require labeling: Stylized or cartoon-like representations that are not photorealistic.

The confusion point: if a partial human figure (hand, arm, background person) appears in a lifestyle shot and that figure is AI-generated and photorealistic, the image needs the tag. Amazon’s guidance does not set a threshold for “how much of a person” triggers the rule. If the asset contains a photorealistic synthetic human element, label it.

Seller forum chatter from late July 2026 shows brands are asking whether AI-edited real photos need tagging. The answer is no, as long as the underlying subject is a real person. The synthetic-performer rule targets fully generated humans, not photo retouching. If you shoot a real model and use AI to swap the background or enhance lighting, the image is exempt. If you generate the model from scratch using AI, the image needs the tag.

If you are building a new product and planning visual assets, this distinction affects your creative budget. Real photography with AI editing is compliant without labeling. Fully synthetic lifestyle imagery is compliant with labeling, but may reduce trust or conversion depending on category. For product research, weigh the visual asset requirement as part of total launch cost. A category that demands lifestyle imagery with human models now has two cost tiers: real photography (higher upfront, no label) or synthetic models (lower upfront, mandatory label).

Step-by-Step Metadata Workflow for Listing Images

Amazon’s implementation is file-level, not visible. You embed the disclosure in IPTC or XMP metadata before upload. Amazon surfaces a shopper-facing indicator, but you do not need a watermark or in-image label.

Here is the workflow:

Identify images containing photorealistic AI-generated people. Audit hero images, secondary images, infographics with lifestyle shots, and any creative assets with synthetic human elements.
Embed the keyword contains-synthetic-performer in IPTC or XMP metadata. Most image editors support IPTC metadata fields. The exact field is not specified by Amazon, but sellers are using the IPTC Keywords field or XMP Subject field. The critical piece is that the keyword appears in machine-readable metadata before upload.
Verify the metadata persists through export. Some image export workflows strip metadata. Re-check the file after export to confirm the keyword is present.
Upload the image to Seller Central. Amazon reads the metadata on upload and applies the disclosure tag automatically. You do not manually toggle a checkbox for main images or secondary images. The metadata itself is the disclosure.
Confirm the listing reflects the indicator. After upload, check the live listing. Amazon will display a shopper-facing indicator (exact wording not yet confirmed) on images tagged with contains-synthetic-performer.

For sellers using Amazon’s Creative Studio or generative AI tools inside Seller Central, Amazon tags the output automatically. You do not need to add metadata manually for assets created in-platform. This applies only to Amazon-native generative tools, not third-party AI image generators.

If you are launching a product and planning to use AI-generated lifestyle imagery, add metadata tagging to your pre-upload checklist. Missing the tag can trigger listing rejection or delay launch. The window between upload and rejection is short. Amazon’s automated review appears to flag untagged synthetic images within hours, based on seller reports from late July 2026.

A+ Content and Brand Story Updates: In-Product Checkbox vs. File-Level Tag

A+ Content and Brand Story have a different workflow. Amazon added an in-product checkbox for synthetic-performer disclosure.

Here is how it works:

For new A+ Content modules or Brand Story panels: When you upload an image containing photorealistic AI-generated people, you check the “contains synthetic performer” box in the upload modal. You do not need to embed metadata in the file. The checkbox replaces the file-level workflow.
For existing A+ Content or Brand Story assets: You must retroactively update the module. Edit the module, re-upload the image, and check the box. Amazon is treating retroactive compliance as mandatory, not optional. Seller forum guidance from July 22, 2026 says the rule applies to all content created before the June 9 effective date.
For A+ videos: The checkbox workflow applies to video uploads as well. If the video contains photorealistic AI-generated people, check the box when uploading. Amazon does not yet support video-file metadata for synthetic-performer disclosure.

The in-product checkbox is only available for A+ Content, Brand Story, and A+ video uploads. It is not available for main images, secondary images, Store images, or advertising videos. Those asset types require file-level metadata embedding.

If you manage A+ Content at scale (10+ ASINs, multiple modules per listing), the retroactive update is the operational bottleneck. You cannot bulk-edit A+ modules to add the disclosure. Each module must be opened, re-uploaded, and re-submitted individually. For brands with 50+ SKUs using AI-generated lifestyle imagery across A+ modules, the compliance backlog can take days.

This is where product research decisions intersect with compliance cost. When evaluating a new product using Kaldon’s 5-phase launch process, consider visual asset requirements as part of the Build phase. If your category demands A+ Content with lifestyle imagery, and you plan to use AI-generated models, budget time for metadata tagging and potential retroactive updates. If real photography is feasible, the compliance overhead is lower.

Which Content Types Require Labeling (and Which Are Exempt)

Amazon’s rule applies to more than listing images. Here is the full scope:

Requires labeling:

Main images (hero shot)
Secondary images (lifestyle shots, infographics with people)
A+ Content images and videos
Brand Story images
Store images (homepage, category pages, product tiles)
Product videos
A+ videos
Advertising videos (Sponsored Brands video, Sponsored Display video)

Does not require labeling:

Text-based A+ modules (no image upload)
Product-only shots with no people
Images with real people, even if edited with AI
Fictional characters from licensed IP
Stylized or cartoon-like art

The broadest category is advertising videos. If you run Sponsored Brands video campaigns with AI-generated lifestyle footage, each video needs the synthetic-performer tag. Amazon has not yet clarified whether the tag affects ad approval speed or placement, but seller guidance from late July 2026 recommends pre-tagging all video assets before submitting to the ad console.

For Store images, the checkbox workflow applies. You upload Store assets through the Store builder, and the synthetic-performer checkbox appears in the image upload modal. The workflow mirrors A+ Content, not listing images.

If you are launching a new brand and building a Store from scratch, decide upfront whether to use real photography or labeled AI imagery. Mixing the two (some pages with real models, some with synthetic) is compliant, but may confuse shoppers. Consumer trust data from July 2026 shows 68% of shoppers trust brands more when AI images are labeled, but 31% expressed skepticism about AI-generated product photos overall. The safest default is product-only shots for core listings and real UGC for social proof modules.

How to Audit Existing Catalogs for Retroactive Compliance

Amazon treats the June 9, 2026 effective date as mandatory for all content, including assets created before the law. You must retroactively tag existing images if they contain photorealistic AI-generated people.

Here is the audit workflow:

Export a list of all active ASINs. Use Seller Central’s inventory report or a third-party catalog tool.
For each ASIN, review the main image and all secondary images. Identify any image containing a photorealistic AI-generated person. If you used AI generation tools (Midjourney, Stable Diffusion, Amazon’s generative tools), flag those assets.
Review all A+ Content modules and Brand Story panels. Open each module and check for lifestyle imagery with synthetic humans. If the module was created using Amazon’s AI tools, it may already be tagged. If you uploaded third-party AI-generated images, the module needs an update.
Review Store pages. Open your Store builder and audit homepage banners, category headers, and product tiles for AI-generated people.
Review product videos and A+ videos. Check each video for synthetic human elements. If the video includes AI-generated models or lifestyle scenes, it needs the tag.
For assets that need updating: Re-upload the image with the contains-synthetic-performer metadata (for listing images) or edit the A+ module and check the disclosure box (for A+ Content and Store assets).

The compliance backlog depends on catalog size and how heavily you used AI-generated lifestyle imagery. A brand with 10 ASINs and minimal A+ Content can complete the audit in a few hours. A brand with 100+ ASINs, multi-module A+ Content, and AI-heavy Store pages may need several days.

If you used Amazon’s Creative Studio to generate images after June 2026, those assets are auto-tagged. You do not need to update them. The risk is assets created before June or uploaded from third-party tools.

For product research, this audit cost should inform your decision to enter categories with high visual expectations. Apparel, beauty, and home goods categories rely heavily on lifestyle imagery. If you plan to use AI-generated models, the compliance overhead is higher than categories that can succeed with product-only shots (e.g., electronics, tools, office supplies).

What Happens If You Skip Labeling (and How to Fix Rejected Listings)

Missing the synthetic-performer tag can trigger listing rejection, image removal, or account-health penalties. Amazon’s automated review appears to flag untagged synthetic images within hours of upload, based on seller reports from late July 2026.

Here is what happens:

Listing rejection: If you upload a new listing with an AI-generated person in the main image and no metadata tag, Amazon rejects the listing. The rejection notice references the synthetic-performer disclosure requirement. You must re-upload the image with the tag before the listing goes live.
Image removal: If an existing listing contains an untagged AI-generated person, Amazon may remove the image and flag the ASIN. The listing stays live, but the hero image or secondary image is suppressed. You must re-upload the image with the tag to restore it.
Account-health penalty: Repeated violations can trigger an account-health flag. Amazon has not yet specified the penalty threshold (how many untagged images trigger a flag), but seller guidance recommends treating the rule as zero-tolerance.

To fix a rejected listing:

Download the rejected image.
Embed the contains-synthetic-performer keyword in IPTC or XMP metadata.
Re-upload the image to the ASIN.
Resubmit the listing.

If the listing is already live and the image was removed, the fix is the same. Re-upload the tagged image through the listing editor.

For A+ Content, the fix is module-level. Edit the module, re-upload the image, check the synthetic-performer box, and resubmit the module. Amazon reviews the update and restores the content if the tag is present.

The fastest fix is to replace AI-generated people with product-only shots or real photography. If you remove the synthetic human element entirely, the image no longer needs the tag. For brands with tight launch timelines, swapping to product-only shots may be faster than re-tagging and re-uploading dozens of images.

If you are in the Discover or Build phase of product validation, factor compliance risk into your visual strategy. A product that depends on lifestyle imagery with human models now carries compliance overhead if you use AI generation. A product that can succeed with product-on-white shots has lower risk and faster time-to-launch.

How This Changes Product Research and Visual Asset Planning

The labeling requirement affects product research in two ways: visual asset cost and consumer trust.

Visual asset cost: Categories that rely on lifestyle imagery now have a cost fork. Real photography (model hire, studio time, editing) is more expensive upfront but requires no labeling and carries no trust risk. AI-generated lifestyle imagery is cheaper upfront but requires metadata tagging, may reduce conversion if shoppers distrust the label, and adds compliance overhead for catalog updates. When you evaluate a new product, estimate the visual asset requirement as part of total launch cost. If your category demands lifestyle shots with human models, compare the cost of real photography vs. labeled synthetic images. The gap is narrowing, but real photography remains the safer default for high-trust categories (beauty, apparel, baby products).

Consumer trust: Data from July 2026 shows 68% of shoppers trust brands more when AI images are labeled, but 31% expressed skepticism about AI-generated product photos overall. The label itself is neutral. It discloses the asset is AI-generated, but does not vouch for quality or accuracy. Some categories tolerate the label better than others. Electronics and home goods shoppers may not care whether a lifestyle shot is real or synthetic. Beauty and baby product shoppers may prefer real models. If you are launching into a high-trust category, test real photography first. If conversion holds with AI-labeled images, you can scale synthetic assets later. If conversion drops, the label may be costing you sales.

When you use Kaldon’s demand discovery tools to identify unmet demand, the visual asset requirement should be part of the Build phase checklist. If the market is paying for a product but nobody is shipping it yet, ask why. One reason may be high visual barriers to entry. If the category demands expensive lifestyle photography and most sellers cannot afford it, AI-generated imagery with labeling may be the unlock. If the category tolerates product-only shots, the visual barrier is lower and you can launch faster.

The compliance checklist should be part of your pre-inventory decision tree. Before you order stock, confirm you can produce compliant visual assets within budget. If the only viable option is AI-generated lifestyle imagery, budget time for metadata tagging and potential label-driven conversion loss. If real photography is feasible, the compliance path is clearer.

Operational Checklist: What to Do Before Your Next Upload

Here is the step-by-step compliance checklist for sellers uploading new images or updating existing catalogs:

Before uploading a new listing image:

Identify whether the image contains a photorealistic AI-generated person.
If yes, embed the contains-synthetic-performer keyword in IPTC or XMP metadata.
Verify the metadata persists through export.
Upload the image to Seller Central.
Confirm the listing reflects the shopper-facing indicator.

Before uploading new A+ Content or Brand Story assets:

Identify whether the image or video contains a photorealistic AI-generated person.
If yes, check the “contains synthetic performer” box in the upload modal.
Submit the module and verify the disclosure appears on the live listing.

For existing catalogs:

Export a list of all active ASINs.
Audit main images, secondary images, A+ Content, Brand Story, Store pages, and videos for AI-generated people.
For listing images, re-upload with metadata tags.
For A+ Content and Store assets, edit the module and check the disclosure box.
Resubmit all updated content.

For new product launches:

During the Build phase, decide whether to use real photography, AI-generated imagery with labeling, or product-only shots.
If using AI-generated imagery, add metadata tagging to your pre-upload checklist.
Budget time for compliance updates if you plan to scale the product across multiple ASINs or expand into A+ Content later.

The operational cost is highest for brands with large catalogs and heavy A+ Content usage. The compliance burden is lowest for brands using product-only shots or real photography. If you are just starting on Amazon, the safest default is product-on-white for hero images and real UGC for social proof. You can add AI-generated lifestyle imagery later if category data supports it.

For sellers using Kaldon’s 5-phase launch workflow, this checklist fits into the Create phase. Before you finalize listing images, confirm all synthetic-human assets are tagged. The alternative is to generate compliant product-only imagery through Kaldon’s visual tools and skip the labeling workflow entirely.

FAQ: Amazon AI-Generated Image Labeling

Do I need to label AI-edited images if the person is real?

No. Amazon’s rule applies only to photorealistic AI-generated people. If you photograph a real model and use AI to edit the background, adjust lighting, or remove objects, the image does not need labeling. The person must be fully synthetic (created by AI, not photographed) to trigger the requirement.

What happens if I upload an untagged image containing a synthetic person?

Amazon’s automated review will reject the listing or remove the image. You will receive a notice referencing the synthetic-performer disclosure requirement. To fix it, embed the contains-synthetic-performer keyword in IPTC or XMP metadata and re-upload the image. Repeated violations can trigger account-health penalties.

Do I need to update old A+ Content created before June 2026?

Yes. Amazon treats the rule as retroactive. If your existing A+ Content contains photorealistic AI-generated people, you must edit the module, re-upload the image, check the synthetic-performer disclosure box, and resubmit. The compliance deadline is not explicit, but seller guidance recommends completing the audit by end of Q3 2026.

Can I use AI-generated product-only shots without labeling?

Yes, as long as the image contains no people. The rule applies only to photorealistic synthetic humans. If you generate a product-on-white image using AI and there are no human elements, the image does not need the tag. This includes AI-enhanced backgrounds, shadows, and lighting.

Will the synthetic-performer label hurt my conversion rate?

It depends on category and shopper expectations. Consumer trust data from July 2026 shows 68% of shoppers trust brands more when AI images are labeled, but 31% expressed skepticism about AI-generated product photos overall. High-trust categories (beauty, baby, apparel) may see conversion drops if shoppers prefer real models. Electronics and home goods categories appear to tolerate the label better. Test both labeled AI imagery and real photography to measure category-specific impact.

Frequently asked questions

Do I need to label AI-edited images if the person is real?

No. Amazon’s rule applies only to photorealistic AI-generated people. If you photograph a real model and use AI to edit the background or lighting, the image does not need labeling. The person must be fully synthetic to trigger the requirement.

What happens if I upload an untagged image containing a synthetic person?

Amazon will reject the listing or remove the image. You will receive a notice referencing the synthetic-performer disclosure requirement. To fix it, embed the contains-synthetic-performer keyword in metadata and re-upload. Repeated violations can trigger account-health penalties.

Do I need to update old A+ Content created before June 2026?

Yes. Amazon treats the rule as retroactive. If existing A+ Content contains photorealistic AI-generated people, edit the module, re-upload the image, check the disclosure box, and resubmit. Seller guidance recommends completing the audit by end of Q3 2026.

Can I use AI-generated product-only shots without labeling?

Yes, as long as the image contains no people. The rule applies only to photorealistic synthetic humans. AI-enhanced backgrounds, shadows, and lighting do not require the tag if there are no human elements.

Will the synthetic-performer label hurt my conversion rate?

It depends on category. Consumer data shows 68% of shoppers trust brands more when AI images are labeled, but 31% expressed skepticism. High-trust categories (beauty, baby, apparel) may see conversion drops. Test labeled AI imagery vs. real photography to measure impact.

Sources & citations

AmazonAI-generated imagescomplianceeCommerce operationslisting optimization

Last updated Aug 7, 2026

Try Kaldon

See the pipeline for yourself.

Start with 2 free analyses. No credit card required.

Start Free