AI Product Photo Review Checklist for Ecommerce

Attach the source image that the business already approved for the product. Confirm that it is owned or licensed, that it shows the current packaging, and that it includes every feature the buyer must recognize. Store the source beside the generated result so the reviewer compares two known files rather than a memory. If the product has variants, review each variant separately; a result that is correct for one color or size is not evidence for another. The source is the reference, not the generated image.

Count the products and every visible part. Trace the silhouette, check straight edges, and inspect the join between the product and its shadow. Look for a second bottle, an extra handle, a duplicated logo, a merged cap, or a reflection that implies a different shape. A transparent or reflective product needs a closer check because the model can bend internal geometry while the overall image still looks plausible. Reject the result when the buyer would receive a different object, even if the scene is more attractive than the source.

Zoom to the label and compare every character with the approved artwork. Generated text can look sharp while spelling a different word or changing a number, and small print is the most common place to find corruption. Check logo position, proportions, and color rather than only whether a logo exists. If the label carries a regulated claim, ingredient list, age warning, or certification mark, do not publish a generated version unless a qualified reviewer confirms it. Keep the factual photograph for any view where the exact text is the product evidence.

Compare the result with a color reference under neutral light. A campaign scene may intentionally change the light, but it must not change the product color, finish, or material. Check whether gloss, fabric, glass, metal, or skin-like surfaces still read as the same material. Look for highlights that hide a defect or shadows that add a feature the product does not have. If the destination marketplace requires a plain background for the main image, keep the generated scene as a secondary asset and preserve a compliant factual image.

List every claim visible in the image and verify it against the product record. Be stricter for cosmetics, supplements, medical devices, children products, food, and anything with safety or certification language. A generated prop can accidentally imply an included accessory, a test result, or a before-and-after outcome. Check that the scene does not suggest a performance the product cannot support. When the category requires professional review, route the image to that reviewer instead of treating the checklist as approval.

Record the source, prompt or workflow, settings, credit cost, reviewer, and the approve or reject reason. If the image is rejected, name the defect so the next attempt does not repeat it. If it is approved, export at the destination size, write descriptive alt text, and check the file against the channel limits. Preview the result in the listing grid and on a phone. Keep the factual source and the approved derivative together so a later reviewer can reconstruct why the image was allowed to ship.