Amazon and AI-generated images: How to use the contains-synthetic-performer tag

Amazon now requires images and videos containing a photorealistic AI-generated person to include a specific metadata keyword before they are uploaded to a product detail page or A+ Content: contains-synthetic-performer.

The keyword must be stored in the file’s XMP dc:subject field. This requirement does not apply to every asset created or edited with artificial intelligence. It targets a narrower case: media depicting a synthetic person designed to look real.

The metadata enables Amazon to identify the relevant media and provide a disclosure where required. It does not replace Amazon’s other image, video and A+ Content policies, and it does not make a misleading asset compliant.

Key takeaways

• The tag applies to photorealistic AI-generated people, not every AI-assisted visual.

• The exact keyword is contains-synthetic-performer.

• It must be stored in the XMP dc:subject field of the final file.

• Media showing only real people is not covered by this specific tag, even if AI tools were used for editing.

• Amazon’s guidance covers product images as well as images and videos used in A+ Content.

• Amazon’s public guidance does not describe any positive or negative ranking effect linked to the tag.

What exactly does Amazon require?

Amazon instructs content owners to add the following value:

contains-synthetic-performer

The value must be added to the XMP dc:subject field. More specifically, Amazon states that the metadata should contain the keyword as a value inside the XMP dc:subject rdf:Bag.

This is not a visible watermark. It is information embedded in the media file that Amazon can read when the asset is uploaded. An IPTC/XMP-compatible metadata editor can add it without changing the visible image.

Amazon explains the requirement in its official guidance on tagging media that contains an AI-generated person.

In Seller Central or Vendor Central, search the Help section for:

How to tag media that contains an AI-generated person

Which media require the contains-synthetic-performer tag?

The tag is required when an image or video contains at least one entirely AI-generated person presented in a photorealistic way.

These cases and exceptions are described in Amazon’s official product image guide.

An exception from this specific metadata requirement does not remove other obligations. Copyright, trademarks, image rights, consumer-protection rules and the accuracy of the product representation still apply.

What about digital replicas of real people?

A digital replica of an identifiable real person should be reviewed separately.

It does not clearly match the narrow case of a fully synthetic person who is not based on an identifiable natural person. However, this does not mean that the asset is automatically permitted or that no other disclosure is required.

A technical metadata tag cannot resolve questions involving consent, publicity rights, impersonation, endorsements or misleading advertising.

When an identifiable person’s likeness is involved, a legal review may therefore be necessary in addition to the technical check.

Where does the Amazon rule apply?

The official documents reviewed cover several types of listing media:

• product gallery images;

• images used in A+ Content;

• videos used in A+ Content;

• other product videos covered by the relevant Amazon media guidance.

Amazon’s A+ Content image guide and A+ Content video guide refer to the same keyword and XMP field for affected media.

The guidance is visible across several regional Seller Central help centres. However, the public pages reviewed do not provide one worldwide launch date or a detailed enforcement schedule for every marketplace.

Teams should therefore check the Help documentation available in the relevant account and country before a major upload.

How do you add the metadata on macOS?

Amazon’s documented method uses Preview to add the keyword to an image:

1. Open the final image in Preview.

2. Select Tools, then Show Inspector, or use Cmd + I.

3. Open the Keywords tab.

4. Select the + button.

5. Enter contains-synthetic-performer exactly.

6. Press Return.

7. Save the file.

Apple confirms that Preview can assign keywords to a PDF or image.

For a controlled production workflow, verify the XMP field with a second tool before uploading the asset to Amazon.

How do you add the metadata on Windows?

Amazon’s Windows instructions use File Explorer:

1. Right-click the final file.

2. Select Properties.

3. Open the Details tab.

4. Find the Tags field.

5. Add contains-synthetic-performer.

6. Apply the changes.

Support varies according to the file format and Windows version.

After saving, confirm that the value was actually written to XMP dc:subject, rather than only to a proprietary field that Amazon may not read.

How do you automate the process with ExifTool?

Teams managing large media libraries can automate the change with ExifTool, a command-line tool that reads and writes metadata across many file formats.

To add the value without deleting existing keywords:

exiftool -XMP-dc:Subject+=contains-synthetic-performer image.png

To verify the field afterwards:

exiftool -XMP-dc:Subject image.png

The result should include:

contains-synthetic-performer

The += operator matters. It appends the new value instead of replacing keywords already stored in the XMP field.

For video files, metadata support depends on the container format and the tool used. Always inspect the final exported file rather than assuming that the metadata survived the workflow.

What are the most common mistakes?

Tagging every AI-generated image

An AI-generated background, an isolated product or an abstract illustration does not automatically trigger this requirement.

Amazon’s rule focuses on media containing a photorealistic synthetic person.

Adding the tag before the final export

Compression, format conversion, resizing or another export can strip metadata.

Add or verify the tag on the exact file that will be sent to Amazon.

Confusing XMP, EXIF and a simple file comment

Amazon specifies the XMP dc:subject field.

ExifTool is the name of a tool. It does not mean that placing the keyword in any EXIF field is sufficient.

Replacing existing keywords

An automated workflow should append the value instead of overwriting the existing list.

Keep the source asset, the submitted file and a log of the change.

Assuming the tag makes the image compliant

Transparency does not correct a misleading image.

The asset must still represent the product accurately. It must not invent a feature, dimension, accessory, result or use that the real product does not provide.

Assuming the tag affects Amazon ranking

The official guidance reviewed describes a transparency and compliance mechanism.

It does not announce a ranking bonus or penalty. Claims about an effect on organic visibility, advertising performance or Amazon recommendation systems would therefore be speculative.

Why is Amazon introducing this requirement?

Photorealistic AI-generated people are increasingly used in lifestyle photography, product demonstrations, videos and user-generated-content-style marketing.

As this type of media becomes easier to produce, marketplaces and advertisers need a reliable way to identify whether the people shown in commercial content are real or synthetic.

Amazon states that the metadata helps it provide disclosures in jurisdictions where transparency is required.

One example of this wider regulatory shift is New York’s synthetic-performer disclosure law, which took effect on June 9, 2026.

The law requires certain advertisements to disclose the presence of a synthetic performer. The official New York State announcement and the published bill text provide the legal wording and scope.

Amazon’s technical rule and New York law should not be treated as identical. The law does not prove the origin of Amazon’s policy, and adding Amazon’s metadata is not a substitute for assessing every legal obligation that may apply to an advertisement.

What does this change for brands?

Marketplace content quality no longer depends only on what shoppers can see.

Teams also need to control:

• how the asset was created;

• whether the people shown are real, synthetic or a mixture of both;

• whether the rendering is photorealistic;

• what rights and approvals support its commercial use;

• which metadata is stored in the final file;

• which transformations were applied before publication;

• which countries and marketplaces use the asset.

For every image or video, a brand should be able to answer:

• Who created the asset, and with which tool?

• Does the depicted person exist?

• Is the result designed to look photorealistic?

• Does the asset mix real and synthetic people?

• Are the relevant rights and approvals documented?

• Is the keyword present in the final file?

• Could an agency, compression tool or digital asset management export have removed it?

• Is the same asset being used in several countries?

What should teams do now?

1. Inventory active media. List product images, A+ Content images and videos that are live or awaiting publication.

2. Classify every asset. Separate real people, AI-edited real people, fully AI-generated people, non-photorealistic characters and media containing no person.

3. Identify affected files. Select every image or video containing at least one photorealistic AI-generated person.

4. Review the rights. Confirm that the creation and commercial use of the asset are permitted and that it does not improperly resemble a real person.

5. Work on the final file. Add contains-synthetic-performer after the last resize, conversion, compression or export.

6. Verify the XMP field. Check dc:subject explicitly before upload.

7. Keep evidence. Archive the source, submitted file, date, marketplace, approval and verification result.

8. Check the published page. Confirm the visual and make sure no unexpected change occurred.

9. Monitor the relevant Amazon Help page. Content policies can change by country, programme and media type.

Conclusion

Amazon is not asking brands to disclose every use of artificial intelligence. The contains-synthetic-performer requirement targets media containing a photorealistic AI-generated person.

The technical edit is relatively simple. The harder part is operational: correctly classifying assets, preserving the metadata through the final export, documenting the rights and ensuring that an agency or automated tool does not break the traceability chain.

Product-page compliance now partly depends on information that shoppers cannot see but platforms can read inside the media file.

Frequently asked questions

Does Amazon require the tag for every AI-generated image?

No. The contains-synthetic-performer tag applies to media containing a photorealistic AI-generated person.

An asset with no person, or only a non-photorealistic character, does not fall within this specific scope.

Should a real photo edited with AI be tagged?

Not under this specific Amazon rule when the media contains only real people, even if AI was used for retouching.

Other requirements concerning image rights, accuracy and consumer protection still apply.

Does the rule apply to A+ Content?

Yes. Amazon’s A+ Content guidance refers to the contains-synthetic-performer keyword and the XMP dc:subject field for affected images and videos.

Must the tag be visible on the image?

No. Amazon requires metadata embedded in the file.

Amazon may then display its own disclosure where appropriate. Adding only visible text to the image does not replace the requested metadata.

Does the tag affect a product’s ranking?

The official pages reviewed do not describe any positive or negative ranking effect. Its documented purpose is transparency and compliance.

What is the difference between XMP and EXIF here?

XMP and EXIF are different metadata frameworks.

Amazon requires a value in XMP dc:subject. ExifTool can write this XMP field despite its name, but placing the keyword in an unrelated EXIF field does not follow Amazon’s published instruction.