Quick Answer

No major online marketplace or ad platform bans AI-generated product images. Every large channel, from eBay to Google to Wayfair, even ships its own AI imagery tools. What gets a listing suppressed or an ad rejected is misrepresentation, an image that shows a product different from what ships, not the fact that AI made the pixels. Disclosure is a separate and much narrower layer: only a handful of channels (Etsy, TikTok Shop, Google Merchant Center) ask the seller to label AI, most apply any label automatically or not at all, and the EU AI Act adds a buyer-facing label for EU sales from August 2, 2026.

This article is general information, not legal advice. For the law behind the disclosure rules, see the AI product photography legal guide, and for decisions that affect your business, talk to a lawyer in your jurisdiction.

The two questions behind every platform's AI-image rule

Every platform's AI-image policy answers two separate questions, and most of the confusion online comes from collapsing them into one. The first is permission: may I use AI images here at all? The second is disclosure: must I label them as AI? They are different questions with different answers, and they get flagged for different reasons.

• Permission layer. Across nearly every channel the answer is yes, when the image is anchored to a real product and represents what ships. Only two genuine permission-layer restrictions exist: TikTok Shop's ban on fully synthetic renders as a main image, and eBay's requirement of a real photo of the actual unit for used goods.
• Disclosure layer. This is where platforms actually diverge, and even here the seller-facing burden is small. Most labels are applied automatically by the platform or are absent entirely.

Every major marketplace already ships its own generative-AI imagery: eBay's background-swap tool in the seller app, Google's Product Studio inside Merchant Center, Wayfair's Decorify and Muse shopper surfaces. A platform does not ship a feature it treats as a policy violation, so the permission question is largely settled. What each platform enforces instead is accuracy: the image must depict the product the buyer receives, a rule about outcomes, not about the tool that made the pixels.

That is the single principle to carry through the rest of this article. Across platforms, the rule that actually bites is accuracy, not provenance. Almost nothing gets flagged for being AI. Things get flagged for misrepresenting the product.

Which platforms allow AI product images (and which require disclosure)?

AI-generated product images are permitted on all thirteen of the major sales and ad channels below. Only TikTok Shop and eBay (used goods) restrict them at the permission layer, and only a minority ask the seller to disclose. The table separates the two questions so each row survives being read on its own.

The permission layer: the only places AI product images are actually restricted

Only two platform rules restrict AI product images at the permission layer. TikTok Shop bans fully synthetic "digital renderings" as a product image, and eBay requires a real photo of the actual unit as the main image for used goods. Everywhere else, a fully AI-made image is allowed as long as it accurately represents the real product.

TikTok Shop bans fully synthetic product images

TikTok Shop is the strictest platform on paper: its Product Listing Policy bans placeholder and rendered images, so a fully synthetic main image is a violation whether or not it is disclosed. The policy bans "placeholders and digital renderings of a product." AI editing of a real product photo stays allowed (quality, lighting, color, translation) when it is truthful, but AI cannot alter size, color, shape, or features so the image no longer matches what ships.

First-time violations get a 72-hour compliance warning before any account-level effects, which gives sellers room to correct a listing rather than losing the account outright. For the fuller breakdown, see the help-desk answer on whether TikTok Shop and Instagram Shop allow AI-generated product images.

eBay requires the actual unit for used-goods main images

eBay allows AI editing and AI-generated backgrounds, and it ships its own background tool, but a used item's main image must be a real photo of the actual unit. The exceptions are Books, Movies, Music, and Video Games, where stock imagery is fine. No in-listing AI label is required as of 2026.

The non-AI picture rules still bind regardless of how the image was made: no watermarks, borders, or promotional text, and a 500 px minimum on the longest side (800 to 1600 px recommended). See the help-desk answer on whether eBay allows AI-generated product photos, and the eBay image spec requirements for the full picture policy.

Amazon and Walmart main images must depict the actual product

Amazon and Walmart both permit AI-edited images but enforce an "actual product image" standard on the main slot. AI edits that start from a real product photo qualify; fully synthetic text-to-image renders that misrepresent the product do not. The distinction is about what the image shows, not whether AI touched it.

Amazon's main image must be accurate on pure white (RGB 255,255,255) with the product filling roughly 85% of the frame, and category rules are unchanged by AI: apparel flat-lay or ghost mannequin in the main slot, footwear a single shoe angled 45 degrees. Walmart requires "actual product images" at 1:1, RGB 255,255,255 white, and a 1500 px minimum longest side for zoom; below 500 px the listing auto-unpublishes. Amazon's separate disclosure question is handled in its own section below. See the help-desk answer on whether Walmart Marketplace allows AI-generated product images, plus the Amazon image requirements and Walmart image requirements for the exact specs.

The disclosure layer: who actually asks you to label AI

Of the thirteen channels, only about four impose a seller-facing disclosure action. Etsy (when AI made the item), Google Merchant Center (machine-readable metadata), TikTok Shop (for substantial AI changes), and the EU AI Act (for EU sales) each ask for something. Pinterest and Meta apply any label themselves, and Walmart, Wayfair, Faire, and Shopify publish no AI-disclosure requirement at all.

Etsy: disclose when AI created or altered the item

From January 14, 2026, Etsy requires disclosure only when generative AI created or altered the item being sold, not when AI merely edits the background or color-corrects a photo of a real handmade product. AI-generated art, AI-assisted prints, and AI-generated designs on a physical product are the trigger. Editing the scene around a real handmade item is not.

When the trigger applies, three steps follow: tick "I used AI-generative technology," select "Designed by a seller" rather than "Made by a seller," and state the AI involvement in the description. Etsy also expects at least one image of the actual physical product. Shopify, by contrast, has no dedicated AI-photo disclosure tag; its binding rule is accurate representation, which mirrors the FTC standard, and the EU AI Act and Meta ad labels still apply. The help-desk answer on whether you must disclose AI-generated images on Etsy or Shopify walks through the listing form.

Google Merchant Center: a machine-readable metadata tag, not a visible badge

Google Shopping's only AI-specific rule is metadata. Since February 2024, AI-generated images submitted to Merchant Center must retain the IPTC DigitalSourceType tag, and Google requires no visible buyer-facing "AI" badge. Google's help page states that "all images created using generative AI must contain meta data indicating that the image was AI-generated by using the IPTC DigitalSourceType metadata tag," and instructs sellers not to remove it (Google Merchant Center).

The exact value string to write into that tag is published on Google's page and should be taken from there at the time you export, because the IPTC vocabulary has been revised over time. The practical trap is that CDNs and re-encoding routinely strip metadata, so a compliant image can arrive tagless. Make sure the tag survives export and hosting. See the help-desk answer on whether Google Shopping accepts AI-generated product images and the Google Shopping image requirements for the surrounding specs.

Pinterest: the label is applied for you

Pinterest allows AI product pins and applies any "AI modified" label itself, so there is no seller action to take and no stated ranking penalty. It has done this since April 30, 2025, using IPTC metadata plus its own classifiers. Pinterest's help page says it "may show a label on the close-up of an image Pin when we detect that it was generated or altered with AI" (Pinterest Gen AI labels).

Pinterest's Merchant Guidelines never mention "AI," "synthetic," or "generated." The bar there is accuracy and quality, not provenance. See the help-desk answer on whether Pinterest allows AI-generated product pins.

Meta and Google ads: disclosure lands on the ad creative

Meta and Google both judge AI ad creative under misrepresentation, not an AI ban. Meta requires an AI Content Label on AI-generated or substantially AI-modified ad creative, live and enforced in 2026, while Google requires an AI label only for election ads. Google's Misrepresentation policy states plainly that "manipulating media to deceive, defraud, or mislead others is not allowed" (Google Ads).

Meta has auto-labeled ads made with its own generative-AI tools since February 2025 and detects third-party AI via C2PA metadata; cropping, brightness, and manual color correction do not trigger the label, and undisclosed AI ad creative is an active reason for rejection. On Google, ordinary product ads need no AI label; only the election "Altered or synthetic content" checkbox is mandatory, and inconsequential edits like resizing, cropping, and background changes are carved out. See the help-desk answer on whether AI-generated product images are allowed in Meta and Google ads.

The marketplaces that publish no AI rule at all: Walmart, Wayfair, Faire, Shopify

Walmart, Wayfair, Faire, and Shopify publish no AI-image disclosure requirement whatsoever. Their binding rule is owned rights plus accurate representation, enforced against the image's content, not its origin. Wayfair's Supplier Code states that "suppliers must own or have the right to use all imagery and videos for their products" (Wayfair).

Enforcement on these channels is accuracy-side. Wayfair runs multimodal AI against supplier images to validate dimensions and attributes, and Faire requires the real product on a white or light background at spec. None of them ask the seller to declare that an image is AI. See the help-desk answers on whether Wayfair accepts AI-generated product images and whether Faire allows AI images for wholesale listings.

The EU AI Act overlay (from August 2, 2026)

For any storefront selling into the EU, the EU AI Act Article 50 adds a buyer-facing transparency label on top of every platform's own rules, applying from August 2, 2026. Article 50(4) provides that "deployers of an AI system that generates or manipulates image, audio or video content constituting a deep fake, shall disclose that the content has been artificially generated or manipulated" (EU AI Act, Article 50). Penalties for transparency breaches reach EUR 15 million or 3% of global annual turnover.

One nuance is still moving: a provisional "AI Omnibus" agreement from May 2026 would give generative-AI systems already on the market before August 2, 2026 until December 2, 2026 to meet the machine-readable marking requirement. Treat that extension as provisional, not settled law. For the mechanics, the AI product photography legal guide covers the law behind this overlay.

Why AI labels and metadata are not a loophole

Stripping metadata to dodge an AI label backfires. Pinterest, Meta, and others run image classifiers that fire even when no metadata is present, and removing provenance data is treated as a separate violation, not a clean escape. The classifiers exist precisely because upload pipelines strip embedded tags during re-encoding.

Content provenance is broad but voluntary. The C2PA coalition has 6,000+ members as of January 2026, including Google, Meta, OpenAI, Sony, Nikon, and Leica. Platforms read Content Credentials at upload but do not require them, so the practical takeaway is simple: keep provenance data intact, do not try to scrub it.

The 2026 disclosure timeline

Four disclosure dates land across 2026, and none of them ban AI. Each adds a narrow labeling step, most of them applied automatically or only in a specific case.

For context, two labels were already live before 2026: Pinterest's auto-label since April 30, 2025, and Google's IPTC metadata requirement since February 2024.

Does Amazon require AI disclosure on product images?

No. Amazon publishes no Seller Central policy requiring you to disclose AI-generated product listing images. The only Amazon AI-disclosure rule that actually exists applies to Kindle Direct Publishing (books), which is a separate program, and that is the source of most of the confusion.

A cluster of blog posts claims otherwise. One widely circulated post describes a "2026 generative-AI content disclosure update," a metadata tag required in A+ Content, a figure that "47% of AI product photos need disclosure labels," and a mid-2026 enforcement window that triggers automatic removals. None of these carry an official Amazon citation, and fresh verification against Amazon's own documentation did not find any of them. They are worth naming once, only to set the record straight.

Here is what Amazon actually publishes. Its product image guide (G1881) sets the accuracy and white-background rules and says nothing about AI disclosure. AI content is permitted as long as it meets the same content standards as human-made content. The one place Amazon does require disclosure is its KDP content guidelines (G200672390), which ask publishers to inform Amazon of AI-generated text, images, or translations, and that is books, not Seller Central listings. See the help-desk answer on Amazon's policy on AI-generated product images.

What actually gets an AI product image flagged?

Almost nothing gets an AI product image flagged for being AI. What triggers a takedown is material misrepresentation: an image that shows a color, feature, accessory, or condition the buyer will not actually receive. The line is the same on every channel, and it is drawn around the product, not the tool.

The cost of getting it wrong climbs a ladder, and every rung is triggered by misrepresentation or a rights violation, never by the fact that an image was AI-generated. TikTok Shop issues a 72-hour compliance warning, then account-level effects. Amazon suppresses the listing, then suspends the account. Removing a watermark or other copyright-management information carries DMCA Section 1202 statutory damages of $2,500 to $25,000 per image. EU AI Act transparency breaches reach EUR 15 million or 3% of global turnover. The enforcement cost of "used AI" is effectively zero; the enforcement cost of "misrepresented the product" is high.

The named failure modes are the ones AI tools produce most often. Wrong color versus the SKU and warped logos or ingredient panels appear on nearly every platform's flag list. See the help-desk answers on how to avoid getting flagged for misleading content, why AI changes product color and how to keep it accurate, and how to stop AI from garbling text and logos.

How do you use AI product photos without getting flagged?

The one property that satisfies every platform's rule at once is accuracy: an image anchored to the real product, preserving its color, text, logos, geometry, and included accessories. The safest AI workflow is therefore the one that behaves like editing a photograph rather than synthesizing a new item.

Consider a US Shopify seller who also lists on Amazon and Walmart, runs Meta ads into the EU, and sells a handmade line on Etsy. For one AI-edited image, they must satisfy FTC accuracy, Amazon and Walmart's "actual product" main-image rules, Meta's AI ad label, Etsy's item-level disclosure if AI shaped the product, and the EU AI Act's buyer-facing label from August 2, 2026. The rules stack rather than replace each other, and the property that clears all of them at once is accuracy. Labels and metadata are procedural. Getting accuracy wrong is what carries the real cost.

This is exactly where generic prompt-to-image tools work against you. ChatGPT, Gemini, Midjourney, and DALL-E reinterpret the product on every image they make, so logos shift, text warps, and materials change from one output to the next. That reinterpretation is what trips the accuracy rule every platform enforces, and it is precisely what TikTok Shop's "no digital renderings" rule and Amazon and Walmart's "actual product image" rule are written to catch.

Nightjar, by contrast, is built specifically for product photography, and it anchors each image it creates (what it calls a Generation) to one to five real product photos the seller uploads (stored as Assets). Because the output starts from the actual product's pixels, it is designed to preserve the real text, logos, and geometry rather than invent a new item. A few properties of that workflow map directly onto the rules above:

• A saved, reusable setup as an audit record. Nightjar saves the whole shoot configuration as a Recipe: the photographic look (its reusable Photography Style, which fixes lighting and camera feel), the Framing and Shadow for product-only shots or the Pose and Camera Distance when a model is in frame, the Background, the chosen Fashion Model, Custom Directions, and output settings. Two images from one Recipe look like one shoot, and the Recipe is itself a structured record of how each image was made, which is the audit trail a marketplace or regulator asks for.
• Source and output kept together. The Library keeps the source product Assets, the generated Assets, and their metadata in one place, so a seller can show what was generated, when, and from what source. That chain of custody is what a takedown appeal or a compliance review turns on.
• Output controls that map to platform specs. A 1:1 aspect ratio meets the square requirements on Walmart, Faire, and Etsy; Upscale brings an Asset to a 2K or 4K long-edge target for zoom-resolution floors while preserving the product; a clean white Background meets pure-white main-image rules.
• A lower-risk posture for the person in the frame. Reusable Fashion Models give the likeness layer a fixed synthetic identity the brand has rights to; a custom Fashion Model can be based on a real person only when the user holds the rights to that likeness.

The through line is product preservation as a design goal rather than a setting. For the accuracy mechanics, see the help-desk answer on how to keep AI from altering the product's shape, and the workflow guides on how to make AI product photos look real and consistent AI product photography.

Is it legal to use AI product photos?

Using AI product photos is legal in the US and EU as long as the image is accurate and you hold the rights to what it depicts. The FTC treats a materially misleading image as deceptive regardless of how it was made, and the EU AI Act adds a transparency label for EU sales from August 2, 2026.

The practical US test under FTC Section 5 is the net impression of the whole image: would the buyer notice the change when the box arrives? On rights, removing a watermark from a supplier's image is usually a DMCA Section 1202 violation, so the safer path is to shoot your own product once and generate from that. This section is deliberately shallow, and it is not legal advice. For copyright, likeness and right of publicity, the EU AI Act mechanics, and an audit-trail checklist, the AI product photography legal guide is the place to go.

Frequently Asked Questions

Are AI-generated product images allowed on Amazon, Etsy, and eBay? Yes on all three. Amazon and eBay allow AI edits of real product photos, with eBay's used-goods main images required to show the actual unit, and Etsy allows AI images while only asking you to disclose when AI created or altered the item itself.

Do I have to disclose that a product image is AI-generated? Usually no. Only Etsy (when AI made the item), Google Merchant Center (machine-readable metadata), and TikTok Shop (substantial changes) ask for a seller action; Pinterest and Meta apply labels themselves; and the EU AI Act adds a buyer-facing label for EU sales from August 2, 2026.

Which marketplaces ban fully AI-generated (synthetic) product images? Only TikTok Shop bans fully synthetic "digital renderings" outright, and eBay requires the actual unit for used-goods main images. Everywhere else, a fully AI image is fine as long as it accurately represents the real product.

What gets an AI product image flagged or removed? Misrepresentation: wrong color versus the SKU, invented features, accessories not in the box, hidden defects on used items, or warped logos and ingredient panels. It is the misrepresentation that triggers the flag, not the fact that AI made the image.

I sell used or regulated goods. Are the rules different? Yes on two fronts. eBay used-item main images must be the actual unit, and cosmetics and supplement imagery must not imply unearned benefits or ingredients under FDA and FTC rules. See the help-desk answers on the FDA and legal rules for AI-rendering cosmetics and supplements and how to generate AI photos for supplements without misleading claims.

My supplier's photo has a watermark. Can I use AI to remove it? Generally no. Removing copyright-management information is a DMCA Section 1202 violation with statutory damages of $2,500 to $25,000 per image. Shoot the physical product once and generate from your own photo instead. See the help-desk answer on whether you can legally remove a watermarked background from a supplier image.

Do the rules change on every channel I sell on? The specifics do, but the two-layer model (permission versus disclosure) and the single accuracy principle apply to any channel, including ones not on this list.

References

Platform policies

• Amazon Seller Central product image guide (G1881)
• Amazon KDP content guidelines (G200672390) - the only verifiable Amazon AI-disclosure policy (books)
• eBay picture policy
• Google Merchant Center AI-image metadata requirement
• Google Ads Misrepresentation / Manipulated media and Google Ads Misrepresentation
• Wayfair Supplier Code of Conduct
• Faire Brand Terms of Service
• Pinterest Gen AI labels
• TikTok Shop AI-Generated Content Restrictions
• Etsy Seller Handbook on AI creations

Legal and standards

• EU AI Act, Article 50
• FTC Deception Policy Statement
• DMCA Section 1202 / 1203 statutory damages

Industry context

• Meta newsroom on GenAI ad transparency (Feb 2025)
• eBay background-swap tool
• C2PA adoption tracker 2026
• rewarx.com "47% of AI product photos need disclosure labels" - cited only to correct the record; carries no official platform sourcing
• Nightjar - AI product photography