Are AI product photos allowed? Ten platform rules (2026)

AI product photos are permitted in some situations, restricted in others, and subject to the rules for the exact platform and image placement. Accurate product representation, required disclosures and image-specific requirements still apply even when a platform supports AI tools.

This review was checked on October 21, 2026. eBay, Walmart and TikTok Shop entries below use US guidance. For dimensions, formats and source qualifications alongside these AI rules, use the marketplace image and AI rules reference.

What does “allowed” actually mean for a product photo?

Start with the job the image performs. A main listing photo identifies the item being sold. A supporting photo can explain a feature or show a setting. A paid advertisement and an organic social post have different publishing controls. The same file may therefore need a different decision in each placement.

Then separate three questions: does the image accurately describe the item, is this type of image accepted in that placement, and has the required disclosure been applied? Passing one does not answer the others.

For example, an AI scene can preserve a mug’s design while implying that an extra matching plate is included. A realistic model can make a garment look longer than it is. A correct metadata tag records something about the image’s origin; it does not establish that the pictured product matches the order.

What do the ten platforms say in 2026?

The table summarizes the reviewed sources rather than declaring every AI workflow approved. “Not verified” means a general permission or requirement was not established by the source reviewed.

When do Amazon and Google require different metadata?

Amazon’s announcement focuses on photorealistic AI-generated people in listing and A+ media. It specifies the keyword contains-synthetic-performer in XMP dc:subject. It lists exceptions including images with no people and images featuring only real people, even when those people have been edited with AI.

Google’s generative-AI image guidance has a broader trigger: AI generation rather than the presence of a person. It specifies IPTC DigitalSourceType source metadata, including TrainedAlgorithmicMedia, and tells merchants to preserve embedded source tags.

Consider a generated kitchen scene containing a real product cutout and no person. Amazon’s people-disclosure rule is not triggered merely because the kitchen was generated. That does not settle whether the image meets Amazon’s other listing rules. Google’s AI-source metadata guidance still needs consideration.

For older Amazon media, avoid the simple claim “not retroactive.” One official moderator clarification describes mandatory tagging for new uploads and recommends updating older images when editing listings. Another moderator reply gives broader historical compliance advice. The replies do not support a confident blanket exemption for old images.

Can an Etsy seller replace actual-item photos with AI scenes?

Etsy’s seller policy requires your own photographs or video rather than stock photos or artistic renderings, subject to limited exceptions. It separately requires disclosure when AI was used to create an item. Those are different rules.

Suppose you sell an AI-designed digital print. Describing how the artwork was created addresses the item’s creation process. It does not automatically establish that a fabricated room image, framed mockup or altered physical product photo meets the relevant listing-image exception.

The detailed exception conditions were not fully accessible during this review. Read the policy for your item type rather than extending an exception from a different seller’s example. For a physical item, keep genuine photos that establish what the buyer receives. The Etsy product-photo workflow can help organize that work, but a tool cannot grant a policy exception.

Why does TikTok Shop US need extra care?

TikTok Shop’s AI-image guidance says AIGC can be used when it accurately represents the real product. It warns against changes to product attributes, false scale, invented bundle contents and unsupported effects.

Its US Product Listing Policy separately requires the main image to show the product’s front physical view on a pure white background. It also prohibits placeholders and digital renderings of the product.

Read those requirements together. The conditional AI guidance is not a reason to assume a fully rendered main product image is acceptable. A model, scene or effect can look accurate to its creator and still be unsuitable for the required main-image placement.

The distinction also matters when planning exports: a lifestyle image made for an advertisement should not automatically become the main listing photo. Keep separate approved files for those uses.

Do AI labels on ads settle catalog and organic-content rules?

No. Meta’s reviewed help page concerns ad-image AI info labels and explicitly distinguishes organic content. It describes label behavior; it does not establish a general catalog permission. We could not verify a universal Meta catalog AI-image rule from that source.

Pinterest also distinguishes how AI information appears on non-promoted Pins and ads. Its systems use metadata and classifiers. You should not treat the lack of a visible label in one view as evidence that no AI has been detected or that disclosure can be ignored.

Zalando presents another reason to check the destination. Its guide rejects visible on-image labeling of the kind it describes and advises against submitting content that requires visible legal labeling. Its invisible-marking heading says required by December 2026, while the body uses both recommendation and mandatory language. The guide does not specify one universal metadata field or an exact December day.

If a proposed image needs a legal disclosure the destination cannot accommodate, change the content or seek platform-specific and legal guidance. Hiding the disclosure is not a sound publishing workaround.

How should you review a generated photo before publishing?

Use a short decision record for each planned image, especially when one product appears across several channels.

1. Identify the placement. Record main listing image, supporting gallery photo, catalog feed, ad or organic post.
2. Compare the product. Inspect shape, colour, material, quantity, logos, closures, accessories and anything the buyer could mistake for an included item.
3. Check the actual-item evidence. Keep original views that show the product and any wear or defects. eBay’s US content policy also requires truthful, accurate, rights-compliant content.
4. Apply the destination’s disclosure. Record what you added and which policy prompted it. Do not replace a required visible label with hidden metadata.
5. Inspect the final export. Check dimensions, file format and metadata after the last conversion, not only in the original working file.

The marketplace photo checker can assist with file preparation. It cannot approve product claims or every category exception. When reducing a file, review how the image optimizer handles AI disclosure tags and inspect the output you will actually upload.

CatalogCut Stage checks generated images against your original photo before delivery, and failed checks return their credit. The Stage workflow explains that process. Product checking is useful, but your final review still needs to match the destination, product and disclosure context.

What to do next

Choose one destination and one image placement before making a whole listing set. Keep the original product evidence, resolve uncertain policy scope, then use the AI image metadata tagger to inspect or add the Amazon and Google tags appropriate to that file.