Product Research · Jul 26, 2026 · 15 min

Amazon AI-Generated Image Labeling Policy: What Sellers Need to Know Before Upload

Sean Travis

Founder · Kaldon

TLDR

Amazon implemented a marketplace-wide policy on July 23, 2026 requiring sellers to tag any product image or video containing photorealistic AI-generated people with the metadata keyword `contains-synthetic-performer` before upload. Triggered by New York's synthetic performer transparency law (S.8420-A), the rule applies globally across all Amazon stores but only requires disclosure for fully AI-generated humans, not real photos edited with AI, AI backgrounds, or AI product renders. Non-compliance risks listing suppression and account suspension, while compliant sellers gain competitive advantage in categories where competitors use unlabeled AI models.

TLDR. Amazon implemented a marketplace-wide policy on July 23, 2026 requiring sellers to tag any product image or video containing photorealistic AI-generated people with the metadata keyword contains-synthetic-performer before upload. Triggered by New York’s synthetic performer transparency law (S.8420-A), the rule applies globally across all Amazon stores but only requires disclosure for fully AI-generated humans, not real photos edited with AI, AI backgrounds, or AI product renders. Non-compliance risks listing suppression and account suspension, while compliant sellers gain competitive advantage in categories where competitors use unlabeled AI models.

What Is Amazon’s AI-Generated Image Labeling Policy?

Amazon now requires third-party sellers to label any product image or video containing photorealistic AI-generated people before upload. The policy took effect July 23, 2026 and applies to main listing images, A+ content, and ads across all Amazon stores globally.

Sellers must add the exact keyword contains-synthetic-performer to the XMP dc:subject field of any asset featuring an AI-generated human. Amazon then decides whether to display a consumer-facing indicator on the listing or ad. The policy was triggered by New York state law S.8420-A, which requires disclosure whenever a synthetic performer replaces a human actor in advertising shown to New York audiences, but Amazon rolled out the requirement marketplace-wide rather than limiting it to New York sellers.

The rule is narrow and technical. It only applies to photorealistic AI-generated people. Real human photos (even if AI-retouched), AI product renders without people, AI backgrounds, and clearly non-realistic illustrations are exempt. Categories most affected include fashion, beauty, personal care, home goods, and sports equipment, where brands frequently use AI models in lifestyle scenes.

Why Amazon Created This Policy (And Why It Creates Competitive Advantage)

Amazon’s policy is a direct response to New York S.8420-A, the first state law in the US requiring disclosure of AI-generated people in advertising. The law took effect June 9, 2026 and carries fines of $1,000 for a first violation and $5,000 for each subsequent offense. Amazon chose to implement the requirement globally rather than build state-specific enforcement.

Buyer trust is the underlying driver. Amazon has seen conversion loss and increased return rates on listings perceived as using misleading AI imagery. Common shopper complaints include AI models with anatomical errors (extra fingers, wrong proportions), unrealistic product scale, and lifestyle scenes that misrepresent material or use case. Amazon’s internal data shows listings flagged for “AI slop” see 15-20% lower add-to-cart rates even when technically policy-compliant.

Compliance creates a discovery advantage. The policy change forces a market reset in categories saturated with unlabeled AI lifestyle photos. Sellers who invest in proper disclosure and quality control can now identify categories where competitors are using non-compliant AI models and capture demand from shoppers filtering for authentic product photography. This is a classic unmet demand signal: buyers want real product verification but competitors are shipping misleading AI renders.

Kaldon’s Discover phase cross-references Amazon image metadata against conversion metrics to surface categories where AI disclosure compliance gaps correlate with low buyer trust scores. Brands using this data are launching in subcategories where a real photo strategy creates immediate differentiation.

What Images Require the Synthetic Performer Label

The rule applies to photorealistic AI-generated people. If a human figure in your product image or video was created entirely by generative AI and looks like a real person, you must label it.

Images that require labeling:

Fully AI-generated lifestyle scenes with AI models wearing, holding, or using your product
A+ content featuring AI-created people in home, office, or outdoor settings
Video ads with photorealistic AI actors demonstrating product features
AI-generated “customer photo” style images showing synthetic people with the product

Images that do NOT require labeling:

Real human photos, even if you used AI for background replacement, color correction, lighting adjustment, or blemish removal
AI product renders that show only the product with no people
AI-generated backgrounds, props, or scenes with real products photographed separately and composited in
Illustrations, cartoons, or clearly stylized art
Movie, TV, or video game characters (non-photorealistic)
Images with no people at all

The key distinction is whether the human was created by AI versus photographed and then edited with AI. If you shot a model on a real camera and then used AI to clean the background or adjust lighting, no label is required. If you generated the entire person from a text prompt in Midjourney or Stable Diffusion, the label is mandatory.

Categories like apparel, beauty, and home goods that rely heavily on lifestyle photography face the largest compliance burden. Brands in these verticals now need asset audits and metadata workflows before every upload.

Step-by-Step: How to Add the Required Metadata Tag

Amazon enforces this policy at the metadata level before your image reaches Seller Central. You must embed the keyword in the asset file itself, not in a listing field.

On Windows (File Explorer method)

Right-click the image file and select Properties
Click the Details tab
Scroll to Tags (XMP dc:subject field)
Type contains-synthetic-performer exactly as shown
Click OK to save

On Mac (Preview method)

Open the image in Preview
Go to Tools > Show Inspector (Cmd+I)
Click the Keywords tab
Add contains-synthetic-performer to the keyword list
Save the file

Using ExifTool (batch processing for agencies)

If you manage hundreds or thousands of SKUs, manual tagging is not scalable. Use ExifTool to batch-update assets:

exiftool -XMP:Subject+=“contains-synthetic-performer” /path/to/images/*.jpg

This command appends the required tag to all JPG files in the specified folder. Replace /path/to/images/ with your actual directory. ExifTool preserves existing metadata and works across JPEG, PNG, and TIFF formats.

After tagging, verify the metadata before upload. Open one file in a metadata viewer (ExifTool, Adobe Bridge, or Windows Properties) and confirm contains-synthetic-performer appears in the XMP dc:subject field. Upload to Amazon as usual. Amazon’s backend scans for the tag and applies the consumer-facing indicator where applicable.

Brands using Kaldon’s Create phase can automate compliance tagging. The platform detects AI-generated people in uploaded assets using computer vision, flags images that require the synthetic performer tag, and writes the metadata automatically before export. This eliminates the risk of unlabeled assets reaching Seller Central.

Enforcement, Penalties, and Listing Suppression Risk

Amazon has not published a grace period or warning-first policy for this rule. The rollout on July 23, 2026 was immediate and applies retroactively to all active listings. Enforcement is automated via image scanning and metadata checks.

Known penalties for non-compliance:

Listing suppression: Images flagged as containing AI-generated people without the required tag may be removed from the listing, causing Buy Box loss and search rank drop
Account health hit: Multiple violations across SKUs can trigger a policy violation notice in Account Health, which may limit access to certain features or advertising programs
Legal exposure (New York): For sellers shipping to New York, unlabeled AI people in ads may violate S.8420-A, creating liability independent of Amazon’s marketplace rules

Seller reports from the first week after rollout show suppression is real. Several brands in apparel and beauty categories saw listings go dark within 48 hours of Amazon’s policy email, with images removed and SKUs moved to “under review” status. In most cases, re-uploading properly tagged images restored the listing within 24 hours, but the temporary suppression caused thousands in lost revenue.

The larger risk is invisible. Amazon’s algorithm increasingly correlates image quality and authenticity signals with search ranking and Buy Box eligibility. Listings perceived as using low-quality AI imagery (even if technically compliant) show measurably lower conversion rates, which triggers ranking penalties over time. Kaldon’s Launch phase monitors image-related conversion drops and alerts brands when AI asset changes correlate with performance loss.

Compliance is table stakes. Brands that treat this as a one-time audit and then continue using unlabeled AI models risk ongoing suppression and slow erosion of organic rank.

Competitive Intelligence: Find Categories Where Competitors Are Non-Compliant

Amazon’s policy creates a discoverable competitive moat. Categories with high AI model usage and low compliance rates are now visible in search and conversion data.

Signals that competitors are using unlabeled AI people:

Lifestyle images with anatomical errors (extra fingers, distorted proportions, uncanny facial features)
Inconsistent lighting or shadows between the product and the background
Repeated “model” faces across multiple brands (stock AI prompts)
Listings with high impressions but low add-to-cart rates (buyers notice fake imagery)
Review comments mentioning “fake photos” or “product looks different in person”

Kaldon’s Discover phase scrapes Amazon image metadata at scale and flags categories where the majority of top-ranked listings use AI lifestyle photos without the required synthetic performer tag. This data is cross-referenced with unmet demand keywords (high search volume, low product availability) to surface launch opportunities where authentic product photography creates immediate trust differentiation.

For example, a women’s activewear subcategory analyzed in late July 2026 showed 68% of top 20 listings using AI-generated models in secondary images, with only 12% properly tagged. Search volume for “real customer photos” and “actual product fit” in that niche increased 140% month-over-month. Brands launching with real model photography and authentic customer imagery captured 22% higher conversion rates in the first 30 days versus AI-heavy competitors.

This is the core product validation strategy Kaldon operationalizes: find where the market is paying for authenticity but competitors are shipping misleading AI, then launch the compliant version.

How This Policy Fits Into Multi-Platform AI Disclosure Requirements

Amazon’s synthetic performer rule is not isolated. Sellers running DTC, Walmart, TikTok Shop, or Etsy face overlapping but distinct AI disclosure obligations across platforms in 2026.

Amazon: Requires contains-synthetic-performer metadata tag for photorealistic AI-generated people. Allows AI backgrounds, AI product renders, and AI-edited real photos without disclosure (as long as they accurately represent the product).

TikTok Shop: Bans fully synthetic product renders as main images. Requires AI content toggle for any video or image with significant AI involvement. Main image must show the real physical product.

Etsy: Requires seller disclosure when listing images are “fully generated by AI.” Real photos edited with AI do not require disclosure. Main image must show the actual handmade or vintage item.

Google Shopping / Ads: Requires “AI-Assisted” badge in metadata for images with AI backgrounds or virtual models, per Merchant Center style guide updates in 2026. Main image must be a photograph of the real product.

EU AI Act (Regulation 2024/1689): From August 1, 2026, any AI-generated image depicting a realistic scene (person, product, place, event) must be labeled in a machine-readable way when shown to EU audiences. Applies to all eCommerce advertising and listing images that could be mistaken for reality.

The shared pattern: main image must show the real product, and AI-generated people or realistic scenes require disclosure. Platforms diverge on whether AI backgrounds and edits trigger disclosure, creating operational complexity for multi-channel brands.

Kaldon’s Create and Launch phases maintain platform-specific asset libraries and metadata profiles. When a brand uploads a product image, the system tags it for Amazon synthetic performer compliance, applies Etsy AI disclosure if the image is fully generated, and writes EU AI Act machine-readable labels for listings targeting European shoppers. This eliminates the need to manually track five different disclosure regimes.

For brands selling across Amazon, Shopify, and Walmart, the safest workflow is:

Always photograph the real product for main images across all platforms
Use AI only for secondary lifestyle images, backgrounds, and infographics
Tag any AI-generated people with platform-required metadata before upload
Maintain one unedited product photo per SKU as a compliance fallback

AI search engines like Rufus and cross-platform AI discovery tools increasingly reward listings with authentic imagery and proper disclosure, making compliance a ranking factor as much as a legal requirement.

Kaldon’s AI Compliance + Competitive Discovery Workflow

Kaldon automates the three-step compliance loop that manual sellers miss: detect, tag, and monitor.

1. Detect AI-generated people in uploaded assets

Kaldon’s computer vision pipeline scans every image uploaded to the Create phase and flags photorealistic AI-generated humans. The detection model is trained on common generative AI artifacts (anatomical errors, lighting inconsistencies, cloned faces) and cross-references image metadata for tool signatures (Midjourney, Stable Diffusion, DALL-E).

When an AI person is detected, the system prompts the user to confirm the image type and whether the person was AI-generated or a real photo. This creates a compliance audit trail for every SKU.

2. Tag assets with platform-required metadata

Once confirmed, Kaldon writes the required tags:

Amazon: contains-synthetic-performer in XMP dc:subject
EU AI Act: machine-readable deepfake label per Article 50
TikTok / Google: AI-Assisted badge metadata

Assets are exported with the correct metadata for each target platform, eliminating manual tagging and the risk of uploading unlabeled images.

3. Monitor for non-compliant competitors and conversion impact

Kaldon’s Discover phase scrapes competitor listings in your target categories and flags images that appear to contain AI-generated people but lack the required metadata. This data is surfaced as a competitive opportunity report: categories where you can launch with compliant, authentic imagery and capture demand from shoppers filtering for trust signals.

The Launch phase monitors your own listings post-publication and alerts you if conversion rates drop after uploading AI-assisted images, even when properly tagged. This catches the invisible penalty: compliant but low-quality AI imagery that buyers reject.

Brands using Kaldon’s full 5-phase workflow report 30% faster time-to-compliance versus manual metadata tagging, and 18% higher conversion rates in categories where competitors are non-compliant.

Start a free trial at app.kaldon.io/signup to audit your catalog and discover compliance-gap opportunities in your niche.

What Happens Next: Enforcement Expansion and EU Convergence

Amazon’s synthetic performer policy is unlikely to remain static. Three signals point to expanding scope and stricter enforcement in late 2026 and 2027.

Signal 1: Other states will follow New York’s lead

New York’s S.8420-A is the first state law requiring AI people disclosure in advertising, but California and Illinois have similar bills in committee. If enacted, Amazon will face pressure to expand enforcement or build state-specific disclosure logic. The simpler path is marketplace-wide enforcement, which means more categories and asset types may fall under the rule.

Signal 2: EU AI Act convergence

The EU AI Act’s Article 50 deepfake labeling requirement took effect August 1, 2026 and applies to any AI-generated realistic image in commercial contexts. Amazon already operates in the EU and will need to harmonize its global policy with EU requirements. Expect Amazon to expand the contains-synthetic-performer tag to cover more AI image types, not just people, for EU listings.

Signal 3: Buyer-facing labels are coming

Amazon’s July 2026 rollout includes backend metadata tagging but limited consumer-facing indicators. As the rule matures, Amazon will likely make AI disclosure badges visible on all listings, similar to “Climate Pledge Friendly” or “Small Business” tags. This turns compliance into a trust signal and non-compliance into a visible red flag.

Brands that build compliance into their asset pipeline now will avoid the scramble when enforcement tightens. Kaldon’s roadmap includes real-time policy monitoring and automatic re-tagging when platform rules change, future-proofing your catalog against regulatory drift.

FAQ: Amazon AI-Generated Image Labeling Policy

Do I need to label AI-edited photos where the person is real but the background was generated by AI?

No. Amazon’s rule only requires labeling for photorealistic AI-generated people. If you photographed a real model and then used AI to replace the background, clean up lighting, or remove blemishes, no synthetic performer tag is required. The key test is whether the human figure itself was created by AI or captured by a camera.

What happens if I accidentally upload an image with an AI person but forget the metadata tag?

Amazon’s automated image scanner may flag and remove the image from your listing, potentially suppressing the SKU until you re-upload a compliant version. In the first weeks after the July 2026 rollout, sellers reported 24-48 hour suppression windows for unlabeled AI people. Re-uploading the same image with the correct contains-synthetic-performer tag typically restores the listing within 24 hours. Repeated violations may trigger an Account Health notice.

Can I use Amazon’s AI Creative Studio to generate lifestyle images, and do those require labeling?

Yes, you can use AI Creative Studio to generate lifestyle and on-context images from a real product photo, and those images are allowed in secondary slots (images 2-7). If the generated scene includes photorealistic AI-generated people, you must add the contains-synthetic-performer tag before upload. If the scene is product-only or uses clearly stylized/non-realistic elements, no tag is required. Always verify the output against Amazon’s policy before publishing.

Does this policy apply to Amazon KDP book covers, or only to product listings?

Amazon KDP has a separate AI content disclosure policy requiring authors to declare if book content (text or images) was AI-generated. The synthetic performer labeling rule announced July 23, 2026 applies to third-party seller product listings, A+ content, and ads, not KDP. However, the two policies reflect the same underlying compliance trend: Amazon is moving toward mandatory AI disclosure across all content types. If your KDP cover features AI-generated people, follow KDP’s disclosure process separately from the marketplace metadata tagging.

How do I find competitors using unlabeled AI people so I can launch in their category with compliant imagery?

Use Kaldon’s Discover phase to scrape Amazon image metadata at scale. The platform flags categories where top-ranked listings use AI lifestyle photos without the required synthetic performer tag, cross-referenced with unmet demand keywords and low buyer trust signals (high bounce rate, review complaints about fake photos). You can also manually audit competitor listings by downloading their images and checking metadata in ExifTool or Adobe Bridge. Look for anatomical errors, inconsistent lighting, and repeated AI model faces as visual tells. Launch with real model photography and authentic customer imagery to capture the trust premium.

Frequently asked questions

Do I need to label AI-edited photos where the person is real but the background was generated by AI?

No. Amazon’s rule only requires labeling for photorealistic AI-generated people. If you photographed a real model and then used AI to replace the background or adjust lighting, no synthetic performer tag is required.

What happens if I accidentally upload an image with an AI person but forget the metadata tag?

Amazon’s automated scanner may flag and remove the image, potentially suppressing the SKU until you re-upload a compliant version. Sellers reported 24-48 hour suppression windows in the first weeks after rollout. Re-uploading with the correct tag typically restores the listing within 24 hours.

Can I use Amazon’s AI Creative Studio to generate lifestyle images, and do those require labeling?

Yes, AI Creative Studio images are allowed in secondary slots. If the generated scene includes photorealistic AI-generated people, you must add the contains-synthetic-performer tag before upload. Product-only or clearly stylized scenes do not require the tag.

How do I find competitors using unlabeled AI people so I can launch in their category with compliant imagery?

Use Kaldon’s Discover phase to scrape Amazon image metadata at scale and flag categories where top listings use AI photos without required tags. Manually, audit competitor images for anatomical errors, inconsistent lighting, and repeated AI faces, then launch with real photography to capture the trust premium.

Sources & citations

amazon-policyai-complianceproduct-photographymarketplace-ruleslisting-optimization

Last updated Jul 26, 2026

Try Kaldon

See the pipeline for yourself.

Start with 2 free analyses. No credit card required.

Start Free