Amazon Brand Registry photos: no computer-generated mockups

Can you use an AI product image to enroll a brand in Amazon Brand Registry? Amazon's enrollment page says stock images, digitally altered images and computer-generated mockups are not accepted, and that the product or its packaging must show your brand name permanently affixed. An AI render is a computer-generated picture of a product, so by that wording you should not submit one. This post reads Amazon's Brand Registry requirements page on 2026-10-03.

One caveat up front: the page does not mention AI image generators at all. Its "computer-generated mockups" wording was written before AI image tools were common, and I am applying it by plain meaning, not quoting an Amazon AI rule. If an enrollment decision matters, treat the safe reading as the rule: submit a real, untouched photograph.

What does the Brand Registry page accept and reject?

The page asks for an image of your product or product packaging in which the brand name is permanently affixed, usually applied during production by printing, sewing, laser etching or engraving. It rejects anything temporary or altered.

How do I keep enrollment photos and AI images apart?

Treat the two as separate folders with separate jobs. The enrollment set is photographs of the physical product and its packaging, the brand name readable, with no retouching, background swaps or generated elements. Store the originals before anyone opens an editor, and name the files so nobody reuses one as an input to an image model by accident.

The listing set is where AI belongs: lifestyle scenes, extra angles, a cleaner background. Once the brand is enrolled, build those from the same real photos as references. On Sume, openai/gpt-image-2.5 takes up to 16 image references, so a handful of true product photos can anchor a generated scene. The Brand Registry page has no rule about this later step, but every generated image still has to match the product you ship.

What should a generated listing image keep faithful?

The reason Amazon wants the brand name affixed is that it ties a listing to a real product. Carry that into generated images: ask the model to keep the label, logo and packaging text exactly as in the reference, and reject outputs where the text is garbled or invented. A render with a mangled logo is the picture that causes the most trouble later, because it shows a brand mark that does not exist on the product.

A request for one listing image on Sume looks like the call below. The reference must be a public HTTPS URL, and a custom image_size on this model needs both edges as multiples of 16.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-image-2.5",
    "prompt": "Place this exact product on a kitchen counter, soft morning light. Keep the label text and logo unchanged.",
    "image_size": "1024x1024",
    "input_references": [
      {"type": "image_url", "image_url": {"url": "https://example.com/real-product-photo.jpg"}}
    ]
  }'

What do I check before I submit?

Run through four questions: is each enrollment image a photograph, is the brand name readable and permanently on the product, has any file been through an AI or editing step, and are the originals stored apart from the generated listing set. If an answer is no, reshoot. The cost of a reshoot is a phone and a table; the cost of a rejected application is time. For the disclosure side of AI listing media, see Amazon's synthetic performer tag, and for the marketplace-wide accuracy rule, Walmart's AI content rule.