Marketplace Rules: Where Generated Product Shots Are Banned

"AI product photography" reads as a style choice everywhere except on the marketplaces sellers actually publish to, where it's frequently a policy question with a real answer. The line these platforms draw isn't "does this look realistic." It's narrower and stricter than that: is this a photo of the specific unit the buyer is going to receive. A generated hero shot can be indistinguishable from a photograph and still fail that test, because the test isn't about visual quality at all.

The rule that trips people up: "the actual item," not "an item like it"

Etsy draws this line explicitly. Per Etsy's listing image requirements, sellers must use original photos of the actual finished product a buyer will receive — items they made, designed, handpicked, or sourced themselves — not stock photography and not renderings. The narrow exceptions prove how tight the rule is: stock imagery is permitted only for personalized or made-to-order items and goods produced with a production partner, and even then the first listing image still has to show a finished, customized example. Blank-product mockups with placeholder text ("Your Text Here") are explicitly prohibited as a main image.

That means a beautifully generated hero shot of a candle that closely resembles the seller's actual candle — same color, same label style, same shape — is a policy violation on Etsy even if a buyer couldn't tell the difference by eye. The rule isn't about realism. It's about whether the pixels in the listing correspond to a real photograph of the real unit.

Amazon and Google Merchant Center: same instinct, different enforcement

Amazon's main image spec doesn't frame itself as an anti-AI rule, but it interacts badly with generated imagery in practice. Per Amazon's seller reference documentation, the main product image must sit on a pure white background — RGB 255, 255, 255, not off-white or light gray — with the product filling at least 85% of the frame. That's a compositional and technical spec enforced largely by automated classifiers that check the literal pixel values of the background. Generated product shots routinely drift off pure white, because generation models are trained on stylized product photography with soft shadows, gradient backgrounds, and warm or cool color casts baked in — exactly the qualities that read as "professional" in most contexts and get an Amazon main image auto-suppressed. It's less an outright ban and more a spec that a typical generation model's default output doesn't meet without deliberate correction.

Google Merchant Center takes a genuinely different posture, and it's worth naming the difference precisely rather than lumping all three platforms together. Per Google's product image guidelines, the framing rule mirrors Amazon's instinct — "frame your product in the image space so that it takes up no less than 75%, but not more than 90%, of the full image" — but Google does not categorically prohibit AI-generated product images. Instead, it requires transparency: images created with generative AI tools must retain their embedded IPTC DigitalSourceType metadata disclosing AI origin, and stripping that metadata to disguise a generated image as a photograph is the actual violation. That's a meaningfully different model — disclosure-gated permission rather than a blanket ban — and it means the compliant move on Google Shopping is keeping your generation tool's metadata intact, not avoiding generation entirely.

Why this trips up AI-assisted sellers specifically

The failure pattern is consistent across all three: a seller generates a polished, on-brand hero image for the listing's first slot — the one Etsy requires to be the actual unit, the one Amazon's classifier scores against a literal background spec, the one Google will accept if the metadata says what it is — and runs into whichever of the three problems applies. On Etsy, that's an outright violation regardless of how the image looks. On Amazon, it's a background and framing mismatch that risks automated suppression. On Google, it's fine only if the AI disclosure metadata survives the export.

The risk that matters more than any single platform's enforcement, though, is buyer trust. A listing that presents generated art as if it photographs the exact item, followed by a buyer receiving something visibly different, is a returns-and-complaints problem regardless of whether any classifier ever flags it. That's the deeper reason these rules exist in the first place — accurate listing images predate AI by decades as a marketplace trust mechanism, and generation just made it easier to drift from the actual unit without meaning to.

Where generated visuals are still the right tool

None of this argues against AI-assisted product content — it argues for being precise about which image is doing which job. The main or first listing image is a factual claim about the specific unit for sale, and that's where the rules above apply directly. Everything downstream of that — lifestyle context shots, social ad creative, video content that drives traffic to the listing rather than standing in as the listing photo itself — sits in a different category, and generation is typically welcome there, sometimes explicitly so.

The distinction is really a usage rights question: what is this specific image being used to claim, and to whom. A generated lifestyle scene used in a Facebook ad that clicks through to a listing with real, compliant product photos isn't making the "this is the exact unit" claim at all — the listing photo is. Keep that separation clean and generation stays a genuine asset rather than a liability. Our broader guide to where AI content policy gets specific by use case covers the adjacent rules worth checking before you publish.

A practical pre-listing checklist

Before a generated or AI-enhanced image goes on any of these three platforms, ask one question: is this the main listing image, or supporting content?

• Main image on Etsy → must be a real photo of the actual unit. No exceptions outside the narrow personalized/production-partner cases, and even those require a genuine finished example as the first image.
• Main image on Amazon → pure white (255,255,255) background, product filling 85%+ of frame. A real photograph corrected to spec is the safer default; a generated image needs deliberate background and framing correction to pass, not just a prompt asking for "white background."
• Main image on Google Shopping → 75–90% framing, and if AI-generated, keep the embedded IPTC DigitalSourceType metadata intact rather than stripping it on export.
• Anything that isn't the main listing image — social creative, ads, lifestyle content, promotional video — generation is broadly fine. Just don't let it misrepresent material facts like size, included accessories, or color accuracy.

A Versely walkthrough: building compliant marketing content around a real listing photo

The workflow that stays on the right side of every rule above starts from a real photograph, not a from-scratch generation. Upload the actual product photo you already have — the one that satisfies Etsy's or Amazon's main-image requirement — and ask the Versely agent to animate it: "Turn this product photo into a 6-second showcase video with a soft rotating light sweep." That routes to generate_video_from_image using your uploaded photo as the starting frame, which means the unit appearing in the resulting video is still your real, photographed product — just brought to motion — rather than a generated stand-in. Build your supporting content this way and the compliance-sensitive listing image never has to be anything but what it already is: an actual photo of what you're selling. For the fuller production path, Versely's AI product video generator is built around exactly this real-photo-as-source pattern.

FAQ

Can I use AI-generated images as my main product photo on Etsy?

No, with narrow exceptions. Etsy requires the main image to be an original photo of the actual finished item a buyer will receive. Stock photos and renderings are only permitted for personalized or production-partner items, and even then the first image must show a genuine finished example — not a generated stand-in.

Does Amazon ban AI-generated product photos?

Not by name, but its main-image spec — a pure white RGB(255,255,255) background with the product filling 85%+ of the frame — is enforced by automated classifiers that most generation models' default output doesn't meet without deliberate correction. In practice, a real photo corrected to spec is the more reliable path for the main image slot.

Is Google Merchant Center stricter or more lenient than Etsy and Amazon on AI images?

More lenient, structurally. Google allows AI-generated product images as long as the embedded IPTC DigitalSourceType metadata disclosing AI origin stays intact — it's a disclosure requirement rather than a ban. Stripping that metadata to disguise a generated image as a photograph is the actual violation.

Where can I safely use AI-generated product visuals if the main listing image has to be real?

Anywhere that isn't standing in as "this is the exact unit for sale" — lifestyle context images, social ad creative, and promotional video built from your real product photo are all typically fine across these platforms, since they aren't making the same factual claim a main listing image makes.

What's the safest AI workflow for marketplace sellers who still want generated content?

Start from a real, compliant product photo and use generation for what comes after it — video, motion, supporting creative — rather than generating the hero listing image from nothing. That keeps the one image every platform scrutinizes most closely as an actual photograph, while still letting generation do real work on everything around it.

Keep the listing photo real and let generation do the rest — build your supporting product video from an actual product shot with Versely's AI product video generator.