Most AI product images fail in the same quiet way. They look great in the generator, then a marketplace suppresses the listing or a customer receives a product that looks different from the photo.
If you plan to use AI product images for an Amazon listing or a Shopify store, the work is not in making the image. It is in checking it. Here is the list we run before any marketplace upload.
Does the product in the image match the product in the box?
This is the check that matters most, and AI is the tool most likely to break it. A generator can quietly change a bottle's shoulder, move a seam, shift a color by a few shades or invent a texture. The image still looks good. It is just no longer your product.
Put the AI image next to the real product, or its dimensioned render, and compare four things:
- Shape and proportions: height, width, cap size, handle position.
- Color: compare against a physical sample or a color-managed photo, not against your screen.
- Material: matte versus gloss, metal versus plastic, foil versus ink.
- Count and parts: the right number of items, lids and accessories.
If any of these differ, the image is a misrepresentation, whatever the intent. It is also a classic cause of "not as described" returns.
Can a customer read the label?
Text is where generated images still break most often. Letters melt, ingredient lists turn into noise, and logos get redrawn slightly wrong. A label that is blurry on a supermarket shelf is a design problem. A label with invented words on a product page is a compliance problem.
Our approach is to never let the generator draw the label. The label is placed in compositing from the real artwork file, with proper perspective and lighting. If the label is not legible at full size, zoom in and fix it before upload.
What does the platform actually require?
Marketplaces publish their own rules, and they differ by platform and by image slot.
Shopify is the simpler case. Its help center says a square image of 2048 x 2048 px usually displays best, accepts files up to 5000 x 5000 px (25 megapixels) and under 20 MB, and lists PNG as the top choice, followed by JPEG (Shopify Help Center). Shopify sets no background rule; your theme and your brand do.
Amazon is stricter on the main image. Amazon's own Seller Central pages sit behind a login, so we could not quote them directly. What seller guides and Seller Central forum threads consistently describe is this: a pure white background (RGB 255, 255, 255), the product filling most of the frame, and no text, logos, watermarks or props on the main image. Pixel minimums vary between sources, from 1,000 to 1,600 px on the longest side. Treat all of these as secondhand, and read the current Product Image Requirements for your category in Seller Central before you produce anything.
Several of those guides also report that Amazon's checks look at technical criteria and not at whether a camera or an AI made the image, and that AI lifestyle scenes belong in the secondary slots. That is a reasonable working rule, but it is a seller-blog claim and not an Amazon statement. Policies change, and apparel and other categories have extra rules.
Is the background really white?
On Amazon, "looks white" is not the test. A background that is 250 instead of 255 can fail an automated check, and AI images often carry a faint gradient or a soft shadow tint. We clip the background to true white in compositing, keep a natural contact shadow under the product, and then measure the corner pixels before export.
Did you disclose where it matters?
Disclosure rules depend on the platform, the country and the ad format. Licensing and tool terms are a separate question, and we cover them in Commercial Use of AI-Generated Images: What to Ask Before They Run in Ads. For a wider look at what buyers ask about AI imagery, see our AI-Based Images FAQ.
Where AI helps and where a real photo wins
In our work, AI imagery earns its place in secondary slots: lifestyle scenes, seasonal variations and context shots built around a real product reference. The main image, the one that decides the click and the return rate, is usually best built from a real photograph or an accurate 3D render, cleaned up and composited. We have worked with AI imagery for about the last year, and Mehmet has 26 years in post-production, so this is the split we recommend, not a rule from the platforms.
The practical takeaway: a seven-point pre-upload check
- Compare shape, color, material and parts against the real product.
- Read every label at 100 percent. If it is not legible, it is not ready.
- Read the current platform rules for your category before producing.
- Export at the size the platform asks for, in a file type it accepts.
- Measure the background on Amazon main images instead of trusting your eyes.
- Keep a real photo or render as the main image where accuracy decides the sale.
- Save the prompt, the tool and the date, in case a listing is questioned.
If you want a second pair of eyes on a set before it goes live, you can see how we work at 35milimetre.com.