Amazon Requires Sellers to Tag AI-Generated People in Listing Images: What You Need to Know
A New York advertising law that took effect on June 9, 2026 requires businesses that knowingly create ads with synthetic performers to disclose that fact. Amazon's public listing-and-A+ notice followed in July 2026. Against that regulatory backdrop, Amazon has built a worldwide tagging workflow so its systems know when to show a customer indicator. Here's what sellers actually need to do, and what the enforcement risk looks like.
If your Amazon listing uses a photorealistic AI-generated model, there is something else that needs to be added before you upload the image.
It is not text on the image. It is not an Amazon checkbox in most listing workflows. It is a metadata tag hidden inside the file:
contains-synthetic-performer
As of Amazon's July 2026 Seller Forums announcement, sellers must identify product media containing photorealistic AI-generated people before uploading it. Amazon then uses that information to display an AI disclosure to customers where applicable. The requirement applies across Amazon's worldwide stores. Sellers on that thread date the notice to July 23, 2026. Amazon's post itself is not date-stamped. It just says “you must now.”
And strangely enough, the story starts not with that July notice, but with a New York law that took effect on June 9, 2026.
Amazon has not banned AI-generated listing images. It has created a disclosure requirement for one specific category of AI media: photorealistic synthetic people. Real models, AI retouching of real people, empty product scenes, and cartoons are outside this tag. Immediate Amazon risk looks relatively low. We would keep using photorealistic AI people where they convert, tag new uploads, and start an orderly cleanup of what is already live. Do not freeze the catalog. Do not leave hundreds of untagged AI-human files sitting for the long term.
The short answer: what Amazon changed
Amazon's July 2026 listing announcement requires qualifying images and videos to carry contains-synthetic-performer in the dc:subject XMP field before they are added to listings and A+ Content “in any of our worldwide stores.” Amazon uses that information to display a customer-facing disclosure where applicable. Source: News_Amazon.
| What | Date |
|---|---|
| New York GBL §396-b synthetic-performer amendment takes effect | June 9, 2026 |
| Amazon Ads: conspicuous disclosure when qualifying ads are served in New York | June 9, 2026 |
| Amazon moderator (Yujiro): listing/A+ compliance “mandatory since” | June 6, 2026 |
| Seller Forums listing and A+ announcement (News_Amazon) | July 23, 2026, as dated by sellers on that thread. Amazon's post is not date-stamped. |
| This article last checked the live sources | August 30, 2026 |
You must now tag product images that contain photorealistic AI-generated people with specific metadata before you add them to your listings and A+ content in any of our worldwide stores.
Recent legislation requires disclosure when images or videos in advertisements contain photorealistic AI-generated people.
To ensure compliance, you must use an IPTC-compatible metadata editor to add the keyword "contains-synthetic-performer" to the dc:subject (XMP) field of your image or video files before adding them to your listings and A+ content.
When you include this metadata, we'll add an indicator to inform customers that the image includes AI-generated people, where applicable.
Source: You must now tag AI-generated people in listings and A+ content (News_Amazon, July 2026). Live tagging steps: How to tag media that contains an AI-generated person.
The announcement does not name the statute. The US law that matches that description is New York General Business Law §396-b, as amended by S.8420-A/A.8887-B (Chapter 617 of 2025).
One distinction matters immediately. The XMP tag is invisible to the shopper. It is not itself the “conspicuous disclosure” New York requires. The statute wants a person looking at the advertisement to be able to tell that the performer is synthetic. Amazon wants a machine-readable flag so it can display that customer indicator where applicable. Sellers who tag the file have done the Amazon workflow. They have not, by tagging alone, printed a New York disclosure onto the image.
First, Amazon has not banned AI-generated listing images
This is the distinction most headlines skip, and it is probably the most useful point for sellers.
Neither New York nor Amazon is saying: if AI was used, disclose AI. The trigger is a synthetic human performer. An AI-generated product shot with no person is not caught merely because AI created it. An AI kitchen, living room, landscape, or background is not automatically in. Ordinary generative fill, retouching, cleanup, lighting, and product rendering are outside this rule unless the result actually contains a synthetic performer.
Amazon is also not saying sellers cannot use AI-generated product imagery. The new requirement is specifically focused on photorealistic AI-generated people.
Amazon says you do not need this metadata when the media:
- Only features real people, even if they were altered with AI tools
- Only features characters from movies, TV programs, streaming content, documentaries, video games, or other expressive works (for example, movie poster art or a video-game cover)
- Doesn't feature any people
- Doesn't feature photorealistic people
So the shorthand is not “AI image = disclosure.” It is: photorealistic fictitious human = tag.
What does “synthetic performer” actually mean?
New York does not say “anything made with AI.” The enacted definition is:
“Synthetic performer” means a digitally created asset created, reproduced, or modified by computer, using generative artificial intelligence or a software algorithm, that is intended to create the impression that the asset is engaging in an audiovisual and/or visual performance of a human performer who is not recognizable as any identifiable natural performer.
Source: S.8420-A (same as A.8887-B), amending GBL §396-b(1)(c). Official page controls.
In plainer language: a realistic-looking fake person. An entirely fictional AI lifestyle model wearing a shirt, holding a coffee maker, sitting on a sofa. Not a celebrity deepfake of an identifiable natural person. Not “we used AI somewhere in the file.”
There is a drafting issue worth not over-reading. The definition includes assets “created, reproduced, or modified” by computer, which is literally broader than “generated from nothing.” The performer still has to be someone who is not recognizable as an identifiable natural person. During Assembly debate on the companion bill, sponsor Linda Rosenthal explained that the definition was intended to be narrow and not to encompass AI editing, alterations, or visual effects applied in an advertisement. Amazon makes the operational cut even cleaner: real people, even if AI-altered, do not get this tag. Source: A.8887-B (Rosenthal).
Amazon then translates the advertising definition into listing language. An Amazon moderator (Yujiro) has said partial visibility does not avoid the tag. If the content still depicts a photorealistic person who was entirely AI-generated, not based on a real person, Amazon says the metadata should be applied. Source: Questions About the New Metadata Requirement.
The statute itself does not give a clean answer for every synthetic hand, arm, or half-face. Amazon is taking the safer operational position, and that matters on Amazon because lifestyle imagery is full of hands holding products, wrists showing watches, and legs in apparel. If that body part still belongs to a fully AI-generated photorealistic human, Amazon says tag it.
| Image or video | Tag required? |
|---|---|
| Fully AI-generated photorealistic lifestyle model | Yes |
| AI-generated realistic person holding or using the product | Yes |
| Only part of a photorealistic AI person is visible | Yes, if it still depicts a fully synthetic photorealistic person |
| AI-generated photorealistic hand or body portion belonging to a synthetic person | Yes, under that same partial-visibility reading |
| Real model photographed normally | No |
| Real model retouched using AI | No, under this specific requirement |
| AI-generated background with a real person | No, under this specific requirement |
| AI-generated room or product scene with no people | No |
| Cartoon or obviously illustrated person | No |
Sellers searching “Does Amazon allow AI images?” will find a lot of overly broad answers. The live Amazon rule is narrower than that search. A proper audit also cannot stop at the image gallery. The same moderator said the requirement applies across buyer-facing media, not just the main image.
Where did this rule come from?
On June 9, 2026, an amendment to New York General Business Law §396-b took effect. Governor Kathy Hochul signed S.8420-A/A.8887-B on December 11, 2025 as Chapter 617, and described it as first-in-the-nation legislation requiring disclosure when advertisements include AI-generated synthetic performers. The act takes effect on the 180th day after it became law.
The operative rule is short. A person engaged in dealing in property or services who, for a commercial purpose, produces or creates an advertisement concerning that property or service must conspicuously disclose that a synthetic performer appears in the advertisement, when that person has actual knowledge of it.
| Violation | Statutory penalty |
|---|---|
| First violation | $1,000 |
| Any subsequent violation | $5,000 |
The statute does not say what one “violation” is. It does not say whether the unit is one image, one advertisement, one campaign, one publication, one impression, or one day. Statements such as “$5,000 per image” or “$5,000 every time a listing is viewed” are speculation. The statute also does not create an express private right of action.
What “conspicuously disclose” does not tell you
This is one of the law's largest gaps. The statute requires a conspicuous disclosure in the advertisement. It does not prescribe the exact language, font size, placement, whether an icon is enough, how long it must display in a video, contrast, or technical format. The gap is obvious from the text itself: the obligation exists and the implementation standard almost does not.
The legislative justification is clearer than the drafting. The goal is for an average viewer to be able to tell, at the point of viewing, that the person they are seeing is synthetic rather than real. That is the customer-facing duty. Amazon's XMP keyword is the upstream signal, not a substitute for that duty.
What the statute leaves out
Advertisements for expressive works such as films, television, streaming content, documentaries, and video games are exempt where the synthetic performer used in the advertising is consistent with its use in the underlying work. Audio-only advertisements are excluded. AI used solely to translate the language of a real human performer is excluded. The statute also preserves other rights under New York privacy and publicity law rather than replacing them. Treat this as a specific advertising-transparency rule, not a complete law of every synthetic human.
The June 2026 date can make this look like a sudden AI regulation. The sponsor memo traces the concept back through A.6758-A in 2021-2022; S.6859 / A.216 in 2023; S.6859-A / A.216-C in 2024; and S.1228-C / A.606-B earlier in 2025. Those did not become this law. S.8420 was introduced on June 9, 2025 and amended the next day into S.8420-A. The Senate passed it 59-0 on June 13, 2025. The Assembly passed the companion on June 17, 2025. It was delivered to the Governor on December 1, 2025 and signed on December 11, 2025.
The policy rationale throughout was consumer transparency: increasingly realistic synthetic performers make it hard to tell a real person from a generated one in digital advertising.
Why does a New York law affect a seller in Europe, Canada, or Japan?
Because Amazon did not implement this as a New-York-only listing rule. Do not conflate the two scopes.
New York law is a state advertising statute. Amazon Ads treats it operationally as applying where advertisements containing synthetic performers are served to shoppers or viewers in New York. Exactly how New York's geographic reach applies to a particular online advertisement is a separate legal question. Amazon's platform rule is much simpler: Amazon applies it worldwide.
Amazon's Seller Central rule is global. The announcement says the metadata is required in any of Amazon's worldwide stores. An Amazon moderator later said the requirement applies to buyer-facing media including product listing images, A+ Content, Brand Story, Store images, product videos, A+ videos, Store videos, and advertising videos. Each tool may have its own implementation. Source: Yujiro_Amazon.
A state advertising law has effectively produced a global Amazon operational requirement. Maintaining a separate image catalog for one US state would be cumbersome. Amazon's own moderator language is “some jurisdictions may require disclosure,” not “New York only.”
Who the statute actually binds
The enacted New York duty sits on the person who produces or creates the advertisement, where that person has actual knowledge. For a typical Amazon listing, that potentially puts the seller or brand in scope: they commissioned or generated a fictional AI model, put it in a commercial product image, and used that image to sell. That is a much more obvious statutory target than Amazon merely hosting the listing.
Subdivision 8 of the enacted S.8420-A text, and the same idea in the codified GBL §396-b, says the section does not apply to the advertising medium that publishes or disseminates a non-compliant advertisement: newspapers, magazines, television networks and stations, streaming services, cable systems, billboards, transit ads, and similar. Subdivision 6 says the statute neither limits nor enlarges Section 230 of the Communications Decency Act.
An earlier print of this bill said an advertising medium would generally be exempt until it received written notice that an ad contained an undisclosed synthetic performer, then would have up to five days, or as soon as technically or practically feasible, to stop distributing it or insert the disclosure. That print also barred the publisher from stripping an existing disclosure. That language did not survive. The final S.8420-A / A.8887-B text we checked against the New York Assembly bill text is an unconditional media exemption. Do not build an Amazon theory on a clause that did not become law.
Whether an online marketplace like Amazon fits that exemption in every factual scenario has not been tested in court. We would not write “Amazon can never be liable.” When Amazon is merely hosting seller-created creative, the statutory language strongly points toward the medium carveout. When Amazon itself generates the advertisement, for example through its own AI advertising tools, that is a different fact pattern. That helps explain why Amazon Ads auto-handles disclosure in some of its own generative workflows.
Separate three actors:
- The seller or brand that deliberately creates or commissions an AI lifestyle image is the clearest candidate for the statutory disclosure obligation.
- The marketplace, when it is only disseminating someone else's ad, has a strong argument that it is the “medium used for advertising.”
- The marketplace, when it creates the advertisement itself, is not in that same “mere publisher” position.
New York did not make Amazon liable for every seller image, so Amazon is not obviously looking at $5,000 times a catalog. The cleaner thesis is: New York created a disclosure obligation around synthetic performers. Against that regulatory backdrop, Amazon built a metadata flag into listing and A+ uploads worldwide so it can display a customer indicator.
Listings, A+, and ads are not the same workflow
Seller Central product media
For standard listing images and other non-A+ assets, Amazon instructs sellers to use an IPTC-compatible metadata editor and add the exact keyword contains-synthetic-performer to the dc:subject field of the file's XMP metadata before upload. Step-by-step instructions live on How to tag media that contains an AI-generated person.
This metadata is not a visible watermark. A shopper looking at the original file would not see those words printed across the photo. It is machine-readable information that tells Amazon how the asset should be handled.
A+ Content
A+ already has an easier path. According to the same Yujiro reply, sellers uploading qualifying A+ assets have two options:
- Embed
contains-synthetic-performerin the XMPdc:subjectfield themselves, or - Select the “AI-generated people” option in A+ Content Manager Creative Assets. Amazon says that when the checkbox is used, it applies the appropriate metadata.
Images generated through Amazon's own A+ generative AI features, such as Creative Studio, are tagged automatically. For AI-generated images that do not contain photorealistic people, Amazon says the existing A+ disclosure option remains sufficient. See also Amazon's A+ Content guidelines.
Amazon Ads, and the clearest New York connection
The News_Amazon announcement itself is written around listings and A+ Content. A later Amazon moderator said advertising videos are in scope, and that each tool may have its own implementation.
Amazon Ads has a separate help page specifically framed around the New York State disclosure requirement (GP6Y2AE2BUM5XWFX; sign-in; re-check live before you launch). As of this writing, that page is the strongest public evidence that the New York statute actually triggered Amazon's architecture, not just a generic “AI images” cleanup.
Effective June 9, 2026, Amazon Ads says advertisements containing synthetic performers, including photorealistic AI-generated depictions of fictitious people, require conspicuous disclosure when served to shoppers or viewers in New York. How the duty is carried depends on how the asset was created:
- If Amazon Ads' own generative AI tools create the creative and the asset stays in Amazon's ad workflow, Amazon says it identifies qualifying synthetic performers and displays the disclosure automatically.
- If the creative was generated outside Amazon, or generated by Amazon's tools and then downloaded, edited, and re-uploaded, the advertiser is expected to declare “Contains synthetic performers” during asset upload.
- For certain third-party-served formats where Amazon cannot control the presentation, the advertiser is responsible for putting the disclosure directly into the creative.
That last point matches the statute better than listing XMP does. New York wants a customer-facing disclosure. Amazon can insert one when it controls the ad unit. When it cannot, the advertiser has to put the disclosure in the file. Confirm the live Ads help in the console. We are not reprinting unverified click paths.
What about images that were already live?
This section matters because Amazon's public moderator answers do not match.
TaylorR_Amazon (on whether already-uploaded images must be updated): “Compliance is mandatory going forward for new uploads. That said, it's best practice to update older images if you plan to make any edits to those listings anyway.” Source: Adding "contains-synthetic-performer" on AI created images.
Kai_Amazon (on retroactive tagging): the disclosure is mandatory. Then: “If you would like to apply the disclosure retroactively, you can do so following the guidance on the help page linked in the announcement.” Source: Do Previously Uploaded AI-Generated People Images Need Retroactive Tagging?.
Yujiro_Amazon: “On retroactivity, compliance has been mandatory since June 6, 2026. For content published before that date, you should retroactively add the metadata following the guidance on the help pages. We recommend updating existing content to stay compliant, as non-compliance could result in violations and potential ASIN removal.” Source: Questions About the New Metadata Requirement.
Those three answers do not say the same thing. One is new-uploads-plus-best-practice. One sounds optional for legacy files. One says retrofit, with ASIN-removal risk.
There is another oddity. The New York statute took effect on June 9, 2026. Amazon Ads uses that same June 9, 2026 date. Yujiro said Amazon's listing requirement has been mandatory since June 6, 2026. The public listing announcement sellers are working from is the July 2026 Seller Forums post. Sellers on that thread date it to July 23, 2026. Amazon has not published one clean timeline that reconciles those dates. Do not build a strategy around the three-day June gap, or around a single forum reply. “Amazon now” in the announcement is Amazon's wording, not a calendar.
The expert position today is: new uploads are clearly mandatory. Amazon clearly wants existing AI-human assets brought into compliance, but its public guidance is inconsistent on whether every legacy file must be replaced immediately. Amazon's conflicting guidance here concerns whether older Amazon files must be re-uploaded with its metadata. It should not be treated as a legal safe harbor for an advertisement that remains live after the New York law took effect. Inventory high-revenue ASINs and work through them. Do not ignore the catalog because one reply sounded softer, and do not pretend the legacy question is settled.
What happens if you don't tag it?
Separate three things: the law, Amazon's published policy, and what we think Amazon is likely to do operationally.
The law. Under enacted New York §396-b, a covered person who, with actual knowledge, produces or creates a commercial advertisement containing an undisclosed synthetic performer can face a $1,000 civil penalty for the first violation and $5,000 for subsequent violations. The unit of a “violation” is undefined. This article is not legal advice. Whether a given Amazon listing or ad is an “advertisement” under New York law, whether it was produced for a commercial purpose, and whether the seller had actual knowledge, are facts-and-counsel questions.
The theory that Amazon itself faces millions of dollars under this statute, because a regulator samples thousands of seller images, is not well supported by the final text. The enacted statute is structured to put the disclosure obligation on the business that produces or creates the advertisement, and to protect the medium that merely publishes it.
Amazon's published policy. The main Seller Central announcement says you must tag qualifying media. It does not specify an enforcement ladder. There is no published rule that one untagged image equals image removal, five equal listing suppression, or ten equal account suspension. One moderator has said non-compliance could result in violations and potential ASIN removal. That is forum guidance, not a penalty schedule. As of August 30, 2026, we have not found evidence of a broad enforcement wave against sellers for this requirement.
Our assessment. Regulatory pressure can still produce an Amazon crackdown. That is not because the enacted statute already fines Amazon $5,000 per seller image. The publisher carveout is still there. It is because Amazon may decide that thousands of sellers ignoring a legally required disclosure is bad for marketplace compliance, consumer trust, advertiser compliance, or its relationship with states. We have seen Amazon move when an external regulatory issue becomes operationally important.
Our take (assumption and bet, not Amazon policy). Immediate Amazon risk looks relatively low. We do not expect Amazon, tomorrow morning, to scan the whole catalog and drop listings over one untagged lifestyle image. As of August 30, 2026, we have not found evidence of a broad public sweep. The requirement is live. Broad enforcement, at least publicly, has not arrived yet.
What we assume comes next is a sample, not a seller-by-seller lawsuit: a regulator or other enforcement action that surfaces a large volume of photorealistic synthetic people with no disclosure. Amazon hosts countless listings, images, and videos. At that catalog scale, the operational and political exposure can start to look like real money, even if the current New York statute is built to hit the advertiser rather than the host. That “millions” figure is an assumption about platform scale, not a claim that enacted GBL §396-b already bills Amazon $5,000 per image.
Then, still an assumption, Amazon does what it usually does when a regulator is sitting on its neck: a crackdown.
How that would show up is already more bet than assumption. Our bet is image removal or listing suppression until the file is tagged or replaced. Not account suspension, and not a process that needs a complex appeal. Fix the media or the tag, and the listing should be able to come back in the ordinary way, or relatively easily. That is curable. That is not a published Amazon promise.
Would we stop everything now because of this? No. Would we leave hundreds of untagged AI-human images sitting for the long term? Also no. We would continue, start fixing this in a controlled order, and keep original source files ready so that if Amazon suddenly starts enforcing, the catalog can be retagged or replaced quickly instead of reconstructed after listings are already down.
Already seeing image removal, A+ rejection, or an ASIN suppressed over listing media? Do not keep re-uploading the same untagged file. Get my case strategy with the Performance Notification or rejection detail and examples of the affected images.
Should sellers stop using AI-generated people?
No. Would we stop everything now because of this? No. Would we leave hundreds of untagged AI-human images sitting for the long term? Also no. We would keep using photorealistic synthetic people where they convert, tag new uploads, start an orderly cleanup of what is already live, and keep original source files so a later swap is operational, not a scavenger hunt. Make sure any customer-facing disclosure required for the relevant advertising surface is handled correctly. AI-generated lifestyle imagery can still make commercial sense where traditional photography would require models, locations, reshoots, and a much larger budget.
This is a disclosure and asset-management problem, not a reason to freeze creative. Sellers should stop treating AI-generated people as something that can be generated, uploaded, and forgotten. If a catalog contains synthetic people, know which ASINs use them, keep the original source files, preserve tagged masters, and make new media compliant before it reaches Amazon.
What sellers should do now
There is little reason to panic, and little reason to wait for a crackdown. New qualifying creative should be compliant now. Existing AI-human assets should be inventoried and progressively remediated. Keep the original files even if you do not retrofit every listing today.
1. Audit for photorealistic synthetic people
Do not rebuild AI-generated product-only imagery, backgrounds, illustrations, or photographs of actual people merely because AI was used somewhere in production. Focus on realistic people who were generated rather than photographed.
2. Prioritize high-volume and high-revenue ASINs
If you have hundreds of affected listings, remediate them in a controlled order. Replacing everything at once creates catalog disruption you do not need.
3. Tag new listing media before upload
Embed contains-synthetic-performer in dc:subject. Keep the tagged master file. Do not rely only on whatever version Amazon later serves from its image CDN.
4. Use the A+ workflow where it applies
For A+, either embed the metadata or use Amazon's “AI-generated people” option. For ads, follow the live Amazon Ads help, not the listing-image steps by default.
5. Keep an asset record even if you do not retrofit everything today
At minimum: the original source files, a tagged master, proof the tag was present before upload, the date and version uploaded, and the ASIN or placement. If Amazon suddenly starts enforcing, you want to replace or retag from a backup, not hunt for assets after listings are already suppressed.
How to add the tag
XMP is a standard way to store metadata inside an image or video file. dc:subject is the subject/keywords field. Amazon wants the exact string contains-synthetic-performer, including the hyphens. Close variants are not the instruction.
Use an IPTC-compatible editor. Follow Amazon's live tagging help for the current Windows, Mac, and video steps. Photographers already using Lightroom, Bridge, or similar can add the keyword there. One technical example is ExifTool:
Example only. Confirm against Amazon's live tagging help.
exiftool -XMP-dc:Subject+=contains-synthetic-performer your-file.jpg
Then upload the tagged source file. Amazon processes listing images after upload. On the announcement thread, a seller reported that the tag remains present when downloading approved A+ modules, but downloaded live listing images no longer contain the XMP tag, and there is no Seller Central report confirming ingestion. Amazon acknowledged that questions had gone to the relevant teams. The thread still does not provide a clean verification method. Preserving the original compliant upload file is the practical control.
The bigger takeaway
The interesting part of this change is not really the XMP tag. It is the direction.
A state passes an advertising transparency law aimed at synthetic people in ads. The duty sits on the business that produces or creates the advertisement. The final statute specifically exempts the medium that merely publishes it. Amazon describes “recent legislation,” then builds a metadata flag into listing and A+ uploads worldwide so it can display a customer indicator. Amazon Ads, on a New York-titled help page, ties conspicuous disclosure to ads served in New York as of June 9, 2026.
Something that began as a New York advertising regulation has effectively become part of global Amazon catalog operations. For sellers, that means reading why Amazon created the policy, not only the prohibited-content page.
On AI-generated people, six lines are enough:
- New York requires businesses that knowingly create commercial advertisements containing synthetic human performers to disclose that fact conspicuously.
- It is not a disclosure requirement for every AI-generated image.
- The producer or creator carries the primary statutory duty. The final law specifically exempts the advertising medium that merely publishes or disseminates the ad.
- Amazon requires qualifying photorealistic AI-human media to communicate that fact through
contains-synthetic-performermetadata, and it applies that Seller Central requirement worldwide. - The requirement is real. The enforcement framework is still developing. Amazon has mentioned possible violations and ASIN removal. Guidance on old assets is inconsistent. We have not seen a broad sweep as of August 30, 2026.
- Immediate Amazon risk looks relatively low. We do not expect a tomorrow-morning catalog sweep over one untagged image. If a later sample or crackdown comes, our bet is image removal or listing suppression until the file is fixed, not account suspension. We would continue, start an orderly cleanup, and keep original files ready. We would not leave hundreds of untagged AI-human images sitting for the long term.
Listing suppressed, A+ rejected, or ads paused over image or video creative? Send the Performance Notification or rejection detail plus examples of the affected media. We map what Amazon actually asked for against the live listing and A+ workflows, and only recommend next steps if there is a real path.
Get my case strategy with those materials. Related lanes: listing reinstatement · regulatory support · who we help.
FAQ
Did Amazon ban AI-generated listing images?
No. Amazon created a disclosure requirement for photorealistic AI-generated people. AI product scenes with no people, cartoons, and real models (including AI-retouched real models) do not need this specific tag under Amazon's published exceptions.
What is the exact metadata Amazon wants?
The keyword contains-synthetic-performer in the dc:subject XMP field, added with an IPTC-compatible editor before upload. See Amazon's tagging help.
Is the XMP tag the New York disclosure?
No. The tag is invisible. New York requires a conspicuous, customer-facing disclosure in the advertisement. Amazon uses the metadata as a signal and says it will add an indicator where applicable. Tagging the file is the Amazon workflow, not a printed statutory legend on the image.
Does this only apply to New York shoppers?
The New York statute concerns commercial advertisements produced or created with actual knowledge of a synthetic performer. Amazon Ads treats the disclosure as applying when those ads are served in New York. Amazon's Seller Central announcement applies the metadata requirement to listings and A+ Content in its worldwide stores. Treat the Amazon operational rule as global even if the statute is not.
Is it $5,000 per image?
The statute says $1,000 for a first violation and $5,000 for subsequent violations. It does not define the unit of a violation. Do not treat “per image” or “per impression” as law.
Do I have to retag every old lifestyle image?
Amazon's own public answers conflict. One moderator said new uploads are mandatory and older files are best practice if you are already editing. Another said you can apply the disclosure retroactively if you would like to. A third said compliance has been mandatory since June 6, 2026, that sellers should retrofit, and that non-compliance could lead to ASIN removal. New uploads should be tagged. For a large existing catalog, inventory high-revenue ASINs first rather than pretending the legacy question is settled.
Is the A+ checkbox enough?
For qualifying A+ assets, Amazon says yes: either embed the metadata or select “AI-generated people” in A+ Content Manager Creative Assets. For regular product listing images and most non-A+ media, tagging the file before upload is the current path.
How do I verify Amazon kept the tag after upload?
Amazon has not published a clean verification report. A seller on the announcement thread reported that approved A+ modules still showed the tag, while downloaded live listing images did not. Keep the tagged master you uploaded, plus proof the tag was present before upload.
Will Amazon suspend my account over one untagged image?
The announcement does not say that. One moderator mentioned violations and possible ASIN removal. Immediate risk looks relatively low: we do not expect Amazon to drop a catalog tomorrow over one untagged image. If a later crackdown comes, our bet is image removal or listing suppression until the file is tagged or replaced, not an account suspension that needs a complex appeal. That is our bet, not an Amazon promise.
Should I stop using AI-generated models until Amazon starts enforcing?
No. We would not stop everything now, and we would not leave hundreds of untagged AI-human images sitting for the long term. The New York duty sits primarily on the producer or creator of the advertisement, and it is already in effect. We would keep using photorealistic synthetic people where they convert, tag new uploads, start an orderly cleanup, keep original source files, and handle any customer-facing disclosure required for the relevant advertising surface.
Is this legal advice?
No. This is general information, not legal advice. ASA Compliance Group is not a law firm. The entity, the regional listing, the ad placement, and New York's actual-knowledge standard still have to be reviewed for each situation.