Amazon AI-People Tag Missing? Check the contains-synthetic-performer Tag
October 27, 2026 · 10 min read · by Aashirvad Kumar
Your credits refresh every month for as long as ListingRVA AI exists. No renewals, no annual increases, nothing to cancel.
Available on select plans. No renewals, ever.
October 27, 2026 · 10 min read · by Aashirvad Kumar
You can check the contains-synthetic-performer tag yourself: open the image in an XMP-capable metadata reader and look for that exact keyword in the dc:subject field. That check proves what you uploaded. It cannot prove what Amazon stored after processing, because Amazon has not published how it handles the keyword once a file goes live, so the reliable routine is to verify before upload, keep the tagged master files, and log which images contain AI people.
The worry behind this is real. In a 42-reply Amazon Seller Forums thread, sellers describe tagging their files, confirming the keyword is embedded, then finding no confirmation anywhere in Seller Central. One seller reports that A+ modules downloaded after approval still carry the tag, while live listing images no longer contain any XMP data. Another has over 2,000 AI-people images across child ASINs and asks whether every one must be re-tagged. Neither point has a staff answer. In short, Amazon asks sellers to add the keyword to the XMP dc:subject field of images that show photorealistic AI-generated people, and the background sits in the synthetic performer rule explained.
Key takeaways
You confirm it by reading the file's XMP metadata and finding the exact keyword contains-synthetic-performer in the dc:subject list. Most photo editors that support IPTC and XMP show this list as Keywords or Subject, and a command-line metadata reader prints it directly.
exiftool -XMP-dc:Subject yourfile.jpg prints the Subject list.That last point catches more sellers than anything else. Export dialogs, image compressors, and some background removers rewrite the file, and several of them drop metadata by default. The layered master can carry the keyword while the small JPEG that goes to Seller Central carries nothing. Check the final file, right before it is uploaded, and keep a copy of that exact file.
Nobody outside Amazon can say for certain, because Amazon has not published how it processes listing images after upload or where it records the keyword. What sellers can see is the image served on the product page, and one seller in the thread reports that those live images contain no XMP data, while A+ modules downloaded after approval still do.
Unconfirmed: the idea that Amazon's reprocessing strips the tag comes from one seller's report in the forum thread. No Amazon staff reply confirms it, and a missing tag on the image a buyer sees does not tell you whether Amazon read and recorded the keyword when you uploaded the file.
Large sites commonly resize and re-encode uploads into several versions for different screens, and those copies often carry little of the original metadata. That is a general pattern across image pipelines, not a documented Amazon behaviour, but it shows why downloading the live image is the wrong test. The file you submitted is the evidence that matters, which is the same reason provenance systems care about keeping metadata intact, as covered in how Content Credentials work.
Amazon applies the metadata for you in two cases, both inside A+, and everything else is your job. According to an Amazon staff reply in July, images made with Amazon's own A+ generative AI tools are tagged automatically, and ticking the "AI-generated people" checkbox under Creative Assets in A+ Content Manager makes Amazon apply the metadata to that image.
| Upload path | Who adds the tag | What you should do |
|---|---|---|
| A+ image made with Amazon's own A+ generative AI tools | Amazon, automatically (per staff reply) | Note which modules used it in your log |
| A+ image you upload that shows an AI-generated person | Amazon, when you tick the "AI-generated people" checkbox in Creative Assets | Tick the box; a tagged file does no harm |
| Listing gallery images | You, before upload | Tag the file, then verify the final export |
| Brand Story and Store images | You, before upload | Same as listing images |
| Product, A+, Store and advertising videos | You, before upload (staff include video in scope) | Tag the master and keep it |
| Creative Studio or other Amazon ad tools | Not stated in the thread | Check the current panel for a disclosure option |
For everything outside A+, the staff answer is blunt: "tagging the file before upload is the only option." The same reply says the rule "applies to all buyer-facing media across Amazon, not just A+ Content", which covers listing images, Brand Story, Store images, and product, A+, Store and advertising videos. The thread does not say whether Creative Studio or other ad tools tag their own AI output, so treat that output as untagged until Amazon says otherwise.
Amazon staff recommend retroactive tagging for older content, so images published earlier are not safe simply because they predate the rule. Amazon staff state June 6 as the start date, writing that "compliance has been mandatory since June 6, 2026", while the announcement post and several sellers in the thread cite late July. Plan around the staff date, since it is the earlier one.
Because the keyword lives inside the file, backfilling means producing tagged copies and replacing the live images with them. The thread does not say whether Amazon offers any other route for older uploads, so for a backlog like the 2,000-image one in the forum, work by risk rather than by upload date:
On penalties, the staff reply says "non-compliance could result in violations and potential ASIN removal." It does not mention fines or describe a specific enforcement action, so treat any claim beyond that sentence as unconfirmed. Re-exporting older images is also a good moment to catch the extra fingers and warped labels that earlier generators produced, using the AI image artifacts checklist.
For this keyword, an AI-generated person is a photorealistic human created by AI rather than photographed. The announcement exempts real people altered with AI, characters from expressive works, images with no people, and people who are not photorealistic, such as illustrations or cartoon figures.
The exemptions do not address partial figures, such as a photorealistic AI-generated hand holding a mug or a cropped torso wearing a jacket. Amazon does not spell these out in the announcement, so tagging them is the cautious choice and costs nothing. If you would rather avoid the question for some products, there are ways to show scale and fit on-body without a full person, set out in product photos without models.
Visible text is not what this rule asks for. The instruction is a keyword in the XMP dc:subject field, added with an IPTC-compatible editor before the image goes into listings and A+, and an on-image label does not replace it.
An overlay is a separate decision with its own trade-offs. Text on a lifestyle image competes with the product, and main image rules on Amazon are the strictest of any slot, so check the current image requirements before adding anything visible. If you are weighing visible labels against metadata more broadly, the wider picture is set out in Amazon AI image disclosure.
You check a catalogue by working from a folder of tagged master files, reading dc:subject in bulk, and comparing the result against a log of which ASIN images contain AI people. Opening files one at a time stops being practical long before you reach a few hundred images.
B0XXXXXXX_PT03.jpg.exiftool -csv -r -XMP-dc:Subject ./masters > tag-audit.csv writes one row per file, and a spreadsheet filter shows every row missing the keyword.Log first, tag second: a spreadsheet of which images contain AI people is the one record you fully control. If Amazon later adds a confirmation screen, or Seller Support asks about a listing, that log plus your tagged master files answers the question faster than re-checking live pages.
A metadata audit only answers one question: is the keyword in the file. It says nothing about whether the gallery has enough images, whether the bullets match what the AI-people images show, or whether a child ASIN is missing its lifestyle shot entirely. So pair it with a listing-level pass, and once the tags are confirmed, run each ASIN through an Amazon listing quality checker.
Find out where it disappears: in your own export chain, or only on the image Amazon serves. If the file you uploaded contains the contains-synthetic-performer tag, you have done what the rule asks, and a missing tag on the live copy is a question for Seller Support rather than a reason to keep re-uploading.
Keep the case factual. Say which file you uploaded, show that it contains the keyword in XMP dc:subject, and ask whether any further step is needed. Do not assume the tag was stripped, and do not assume it was recorded either; until Amazon publishes how the keyword is read, your verified upload and your log are the record.
Between 25 May and 20 October 2026, 133 of the 861 images generated in ListingRVA AI (15%) used a style that shows a person: model lifestyle, model infographic, model studio, or spokesperson. ListingRVA AI embeds the contains-synthetic-performer tag in XMP dc:subject on its AI-people image exports, so those files leave the tool already tagged. What happens after upload to Amazon is outside what this data can show, so the verification steps above still apply.
Source: ListingRVA AI generation records (totals only, no customer data).
Roughly one generated image in seven, in other words, needed the keyword. For most catalogues that is a manageable backfill when the log exists and a slow one when it does not, so start the log before the backlog grows. The person styles are listed with the rest of the ListingRVA AI image features.
Generate model lifestyle, model studio, and spokesperson images with the keyword already in XMP dc:subject, then verify the export before upload. Start free with 50 credits, no credit card.
Start free →Open the file in an IPTC or XMP-capable editor and look at the Keywords or Subject list, or use a free command-line reader such as exiftool with the XMP-dc:Subject field. The keyword must appear exactly as contains-synthetic-performer in XMP dc:subject, in the final file you upload.
Sellers in a 42-reply Seller Forums thread report seeing no confirmation in Seller Central, and Amazon has not published where or whether it shows one. Verify the file before upload and keep the tagged master as your record.
Amazon has not said how it processes listing images after upload. One seller reports live images lack the tag while downloaded A+ modules keep it, but no staff reply confirms that reprocessing strips it, so a missing tag on the live copy is not proof it was lost.
Per an Amazon staff reply, images made with Amazon's own A+ generative AI tools are tagged automatically, and ticking the AI-generated people checkbox in A+ Content Manager Creative Assets makes Amazon apply the metadata. For non-A+ content, staff say tagging the file before upload is the only option.
Amazon staff recommend retroactive tagging for older content and state compliance has been mandatory since June 6, 2026, though sellers cite later dates. Backfill by risk, starting with images on active, selling ASINs, and group child ASINs that share one file.
No, according to the announcement. Real people altered with AI are exempt, as are characters from expressive works, images with no people, and non-photorealistic people. The tag is for photorealistic people created by AI.
ListingRVA AI now works inside Claude. Connect your workspace and ask Claude to generate listing copy and publish product images to your store, without leaving your editor. See the Claude connector →
Comments
No comments yet, be the first.
Leave a comment