Why AI product images get rejected by marketplaces
Every marketplace and retail portal you sell through enforces its own rulebook for product images, and those rules are stricter than most brands expect. A main image on a white background with no added text. The product filling most of the frame. No logos, badges, props, or borders on the primary shot. Get it wrong and the listing is rejected, suppressed in search, or quietly down-ranked - often without a clear explanation. As teams rush to produce these images faster, many now reach for AI generators. The pictures look convincing in isolation, then fail the moment they hit a platform's automated checks.
The problem is not creativity; it is compliance. A marketplace spec is a hard target - an exact background value, a minimum resolution, a product that matches the real SKU down to the label. AI tools are built to produce something plausible, not something that reliably hits a fixed specification. This guide walks through the image rules the major platforms actually enforce, why AI-generated product shots so often trip over them, and how a CGI or hybrid workflow delivers images that pass on the first upload - and double as assets you can reuse everywhere else. For the fundamentals, see our overview of photorealistic 3D product rendering.
Every image in this article is CGI by Transparent House, not photography. They all come from one 3D model of a waterfall bathroom faucet, rendered for each slot a marketplace listing needs: a white-background product shot, detail and feature close-ups, an installation view, and lifestyle scenes.
The image rules marketplaces actually enforce
Most large marketplaces converge on the same core requirements for a primary product image. The background should be pure white - Amazon specifies an RGB value of 255, 255, 255 - so the product separates cleanly and crops consistently across the grid. The item itself should fill most of the frame; Amazon asks for at least 85 percent. The main image must show the actual product, with no text, logos, watermarks, badges, borders, or extra props layered on top. Resolution has to be high enough for zoom, typically 1000 pixels or more on the longest side. Google's Merchant Center rules run in parallel: an accurate image of the product, no promotional overlays or watermarks, and a clean background. Secondary and lifestyle images get more latitude, but the main image is where listings live or die, and it is the least forgiving. A guide to platform-specific workflows is in our Amazon 3D product rendering guide.
Transparent House waterfall faucet detail render Transparent House faucet cartridge cutaway render Transparent House faucet installation render
Secondary slots are where detail, feature, and installation images belong, like the close-up, cartridge cutaway, and under-sink view above. Platforms allow more freedom here, but each image still has to show the real product.
Why AI-generated product images get rejected
AI image generators fail these specs for structural reasons, not because of a bad prompt. First, backgrounds drift: a generator produces a near-white field with subtle gradients, shadows, or off-white casts rather than a locked 255, 255, 255, so automated background checks flag it. Second, the product is not your product - an AI model approximates a plausible object, inventing logos, warping proportions, and misspelling text on packaging, which breaks the rule that the image must represent the real SKU. Third, consistency collapses across a catalog: every generation is a fresh roll of the dice, so the twentieth image will not match the first in scale, angle, or finish. Fourth, resolution and edge quality are often soft, with artifacts around fine details. You can fix any single image by hand, but at catalog scale the rework erases the speed AI promised. We tested these limits directly in Can AI replace a real shoot?.
How CGI is built to the spec from the first render
CGI inverts the problem. Instead of generating something that resembles a product, a 3D artist builds the actual product from your specifications - real dimensions, real materials, the correct logo placement and label artwork. Because the scene is fully controlled, the background can be set to an exact white value, the camera framed so the product fills the required percentage, and the output rendered at any resolution without quality loss. The same model produces a compliant white-background main image, a set of detail and scale shots, and lifestyle scenes - all consistent because they come from one source asset. When a platform changes its rules or you enter a new marketplace, you re-render to the new spec rather than reshooting or re-generating. This is the core of a hybrid AI and CGI workflow: use AI where it is genuinely useful, such as early concepting, and rely on professional 3D product rendering for the final, spec-compliant deliverable.
The production process, step by step
Gather inputs: your product's CAD or dimensioned drawings, material samples or color codes, label and logo artwork, and the exact image specs for each target platform.
Build the 3D model to real-world dimensions, then apply accurate materials, finishes, and packaging artwork so the render matches the physical SKU.
Set up the compliant shot: a pure-white environment, correct framing and product-to-frame ratio, and a camera and lighting rig that reads cleanly on a marketplace grid.
Render the full image set - main white-background shot plus detail, scale, and in-use variants - at resolutions that satisfy zoom requirements.
Run a compliance check against each platform's rulebook, confirming background value, framing, and the absence of text or props on the main image.
Deliver, then archive the source asset so future variants, seasonal updates, or new-platform specs are a re-render, not a new project.
A typical scenario
Consider a bathroom fixtures brand launching a new line of waterfall faucets on Amazon and on home-improvement marketplaces such as Home Depot and Lowe's at the same time. Its existing photography shows the faucets in styled bathrooms, which works for secondary slots but breaks the main-image rules. An attempt to generate clean shots with AI produced near-white backgrounds, a spout that was subtly the wrong shape, and handles that did not match the real SKU, so the listings failed review. Working from the faucet's CAD files, a 3D team builds the model once and renders everything from it: a white-background main image framed to the required fill, a close-up of the brushed finish, a cutaway of the cartridge, an installation view with supply lines, and lifestyle scenes for the brand's own site. When the brand adds a matte black or polished chrome finish, it is a material swap and a re-render rather than a new shoot. This mirrors the economics in why CGI is more profitable for high volumes.
Common mistakes to avoid
Treating the main image like a creative canvas. Save badges, callouts, and lifestyle framing for secondary images, where platforms allow them.
Assuming near-white is white. Automated checks read exact values; a soft gradient or warm cast can trigger rejection.
Letting AI invent your branding. A generated logo or mislabeled package is both a compliance failure and a legal risk.
Ignoring resolution and zoom minimums, which leave listings looking low-quality next to competitors.
Producing images one at a time instead of from a single source asset, so a catalog drifts out of consistency.
Getting started
Start by collecting the image specifications for every platform you sell on and auditing your current listings against them - you will usually find more violations than expected. Prioritize the SKUs that drive the most traffic, and gather their CAD files, material references, and label artwork. A 3D partner can then build those products once and output a compliant, consistent image set for each channel, with the source assets kept for future updates. If you are weighing this against photography or AI generation, our comparison of 3D rendering versus traditional photography and our guide to choosing a 3D visualization partner are good next reads.
Frequently asked questions
Why do marketplaces reject AI-generated product images?
Usually because the image breaks a hard rule: the background is not a true white value, the product is an AI approximation rather than the real SKU with correct logos and text, or the resolution and edges are not clean enough. Automated checks read these precisely, so a picture that looks fine to a person can still fail review.
What are the most common main-image rules across platforms?
A pure white background, the product filling most of the frame, no added text, logos, watermarks, borders, or props on the main image, and a resolution high enough for zoom. Amazon, for example, specifies a 255, 255, 255 background and wants the product to fill at least 85 percent of the frame.
Can I just fix AI images by hand to make them compliant?
You can fix a single image, but at catalog scale the manual rework usually erases the speed AI promised, and consistency still drifts from one image to the next. Building from an accurate 3D model avoids the rework because every image is compliant by construction.
Does CGI work for lifestyle and secondary images too?
Yes. The same model that produces a compliant white-background main image also produces detail, scale, and lifestyle shots. Because they share one source asset, the whole set stays consistent, and lifestyle images can use the creative latitude platforms allow on secondary slots.
How long does it take to produce a compliant image set?
It depends on how many SKUs you have and whether usable models or CAD files already exist. Once a product is modeled, generating the full compliant image set is fast, and future variants or new-platform specs are re-renders rather than new projects.
