Product Photography / Amazon Compliance
Amazon's AI Image Rules for Sellers: What Actually Needs Tagging in 2026
Amazon changed its AI image policy on July 27, 2026, and the coverage since has been noisier than the rule itself. If your listing images show a photorealistic person who was generated entirely by AI, that media now has to carry specific metadata. That is the whole rule. It is narrower than the headlines suggest, and for most product sellers it never applies.
The part worth your attention is different, and it has been sitting in the same Seller Central document for years.
The AI tagging rule applies only to photorealistic AI-generated people. Product-only images, real people, and real people edited with AI tools are all exempt.
The rule that actually pulls listings out of search is the ordinary main-image specification, and it applies to every listing you own.
I photograph products for brands and marketplace sellers, so I read these changes from the production side rather than the policy side. If you want an accurate source library that never raises the question in the first place, that is what commercial product photography is for. Here is what I would do with a catalog this week. If you want the version of this that covers state and EU advertising law rather than Amazon's own rules, I wrote that separately.
What Amazon's AI image rule actually requires
Amazon's seller guidance covers buyer-facing media that contains a photorealistic person generated entirely with AI. That includes listing images, A+ Content, Brand Story assets, Store images and video.
The contains-synthetic-performer keyword
The mechanism is metadata, not a checkbox. Sellers add the keyword contains-synthetic-performer to the dc:subject field in the file's XMP metadata using an IPTC-compatible editor, before upload. Amazon can then surface an indicator to shoppers where it applies.
Two practical notes from doing this on real files. First, metadata does not always survive an export. If your retoucher exports a flattened JPEG out of a different application than the one where the keyword was written, check the final file rather than the working file. Second, if you build A+ Content inside Amazon's own generative tools, content made there can be tagged automatically. Verify it anyway.
What does not need the tag
This is the section most sellers actually need. Amazon's published guidance exempts media that:
- contains only real people, including real people altered with AI-powered editing tools
- contains no people at all, which covers almost every product-only image
- contains people who are not photorealistic
- uses characters from expressive works such as film, television or games, in the circumstances Amazon describes
A studio photograph of a real product on white does not become AI content because Lightroom or Photoshop used a machine-learning feature during retouching. A photograph of a real model does not become a synthetic performer because an AI-assisted tool cleaned up the skin. The question is narrow: was the depicted person generated as a photorealistic synthetic human.
Will Amazon suppress my listing for using AI images?
Not for using AI by itself. Suppression comes from the image specification, and Amazon has been explicit about it for years. If a listing has no main image meeting the requirements, Amazon can suppress the product from search until a compliant one is supplied.
That is the rule with teeth. It has nothing to do with AI, it applies to every listing in your catalog, and it is the one I would audit first. I keep a detailed breakdown of the specification itself in my Amazon product image requirements guide.
The 75-character title limit landed the same week
Amazon also moved product titles to a 75-character limit in every category except media, effective July 27, 2026. It is a separate change, but it arrived alongside the image update, and both hit the same asset: your listing.
If you are already in Seller Central auditing images, audit titles in the same pass. Doing it twice is wasted work.
Where AI-generated product images actually go wrong
Compliance is the low bar. Accuracy is the one that costs money.
Generative models are very good at producing something plausible. Product photography has a stricter requirement: the item on screen has to remain the item in the box. Small drift is where returns come from.
- a button shifts position or disappears
- a seam changes direction
- packaging text becomes almost right
- the product reads larger or smaller than it is
- a material changes from matte to gloss
- a port, control or connector vanishes
- a hand holds the product in a way a hand cannot
None of that is a creative difference when the image is selling one specific SKU. The closer an image sits to the product detail page, the packaging or a paid ad, the more product fidelity matters. That is why I still treat real photography of the physical item as the foundation of an ecommerce library, and generative work as something that happens downstream of it.
A four-step listing audit
- Find the synthetic people. Identify any buyer-facing media containing a fully AI-generated photorealistic person. Those files, and only those files, need the metadata workflow.
- Verify metadata on the exported file. Not the working file. Check what you are actually uploading.
- Audit the main image against the specification. This is the suppression risk, and it applies even if you have never touched an AI tool.
- Audit the gallery as a sales tool. A strong set answers appearance, scale, construction, features, use and what arrives in the box without making the shopper read a single bullet.
Gear I use for accurate, compliant product capture
Colour accuracy is the difference between a compliant image and a returns problem. If a buyer receives a product in a visibly different shade than the listing showed, that is an accuracy issue no policy document will save you from. I shoot a reference target at the start of a set and build the profile from it.
For a pure white background you want a real white sweep and correct exposure, not a background removed in post. Paper gives you a clean gradient falloff and no cut-out halo around the edges of the product.
A single controllable continuous source with a softbox will take a small seller further than any number of speedlights. You see the light as you shape it, which shortens the learning curve considerably.
If your category benefits from 360 spin imagery, a motorised turntable gets you consistent angular steps. Useful for hardware, footwear and anything a buyer wants to rotate before committing.
The heavier alternative when the product has real mass. A light turntable under a heavy item produces wobble, and wobble shows up as a jump between frames.
When photographing the real product is simply the shorter path
Look at the chain attached to synthetic commercial imagery right now: platform metadata rules, state advertising disclosure law, EU transparency obligations, product accuracy, colour fidelity and consistency across a full catalog.
Then look at the alternative. Photograph the item correctly once. You get a master library that runs on Amazon, your own store, retailer portals, social, email and paid media, with no provenance question attached to any of it.
I shoot commercial product photography from my Bridgeville studio and take shipped product work from anywhere in the country. For parts, accessories and anything that bolts to a vehicle, I run that through automotive product photography because the lighting problems are different. If you are getting ready to ship something, my guide on how to ship your product to a photographer covers what to include and what to tell me before the box arrives.
Pittsburgh Studio / Nationwide Shipped Product Work
Listing images that pass, and then sell
Compliance gets the image onto the page. Photography decides if the shopper buys. Send me the product, the categories you sell in and where the images will run.
Start a ProjectFAQ
Does Amazon allow AI-generated product images?
Yes. Amazon has no blanket ban on AI-generated or AI-edited product images. The requirement is accuracy. Every image, however it was made, has to represent the physical product the customer receives and meet Amazon's standard image specifications.
What is the contains-synthetic-performer tag?
It is an IPTC metadata keyword Amazon requires on buyer-facing media that shows a photorealistic person generated entirely by AI. Sellers add the keyword contains-synthetic-performer to the dc:subject field in the file's XMP metadata using an IPTC-compatible editor before uploading.
Do I need to tag a photo of a real model that was edited with AI?
No. Amazon's guidance exempts media containing only real people, including real people whose images were altered with AI-powered editing tools. It also exempts images with no people, non-photorealistic figures and characters from expressive works.
Will Amazon suppress my listing for using AI images?
Not for using AI by itself. Suppression comes from the ordinary image requirements. A main image that does not meet Amazon's specification can be pulled from search until a compliant one is supplied, and that rule applies to every listing regardless of how the image was created.
What is the new Amazon product title limit?
Amazon moved product titles to a 75-character limit in all categories except media on July 27, 2026. It landed the same week as the AI metadata change, so a listing audit in 2026 should cover titles and images together.
What should I send a product photographer for an Amazon listing?
Send the physical product, the exact variations you sell, any accessory that ships in the box, your category style guide and a short note on where the images will run. That last detail decides the shot list and the usage license more than anything else.
Related reading: my Amazon product image requirements guide, how to ship your product to a photographer, and what UGC content is and why brands pay creators to make it.
I am Ian Jones, a commercial photographer and videographer working out of a Bridgeville studio outside Pittsburgh. See every service I offer, read client reviews, or start a project.
Sources: Amazon Seller Central product image guide and the seller announcement on photorealistic AI-generated people, accessed August 2026. Amazon Seller Central announcement on product title length, July 27, 2026. Policies change, so verify before you rely on any of it.