Amazon Can Draft a Listing From Three Photos. Give It Evidence, Not Advertising.

Amazon can now turn up to three product photos into a draft listing. That is useful, but only if the photos function as evidence. A polished lifestyle image may persuade a shopper while giving the listing model very little reliable information about the product itself.

The safest workflow is to separate listing creation from listing promotion. First, give Amazon clear views that help it identify the object. Then check every generated field against a short product truth sheet. Only after the catalogue data is accurate should you build the customer-facing gallery.

What Amazon actually creates from an image

Amazon’s current Seller Central workflow lets a seller open Add Products, choose the Product Images tab, upload up to three images, and submit them for AI-assisted listing creation. Amazon says the system can propose a title, product description, bullet points, and attributes. The result is a draft that you can review and edit before submission, not a verified product record.

That distinction matters. A model can recognise that an object looks like a gold ring, a ceramic mug, or a leather handbag. It cannot establish from appearance alone that the ring is solid 18-carat gold, the mug is dishwasher-safe, or the bag uses full-grain leather. Those are inventory facts, not visual impressions.

Build a three-image evidence pack

Use the available image slots to reduce ambiguity. Do not simply upload three near-identical hero shots.

Image one should be a clean identification view. Show the complete product, sharply focused, with minimal props and a plain background. This gives the system the best chance of recognising the product type, overall colour, shape, and major components.

Image two should explain structure. Photograph the back, side, underside, closure, connector, or other angle that distinguishes the item from similar products. For a ring, this could show the setting profile and band. For a bag, show the strap attachment and opening. For a kitchen tool, show the working end and handle connection.

Image three should reveal a detail that affects the listing. This might be a hallmark, model number, included accessory, pack contents, control panel, texture, or fastening mechanism. Keep the detail legible, but do not assume the AI will transcribe small text correctly. You will still enter and verify identifiers manually.

Avoid using a lifestyle photo as the main evidence image when it contains several props, a model, or strong styling. The system may confuse an accessory with part of the product, infer the wrong size from perspective, or describe a prop that is not included.

Write a product truth sheet first

Before opening Seller Central, collect the facts that a photograph cannot prove. A simple one-page note is enough. Include the exact brand and product name, GTIN or exemption status, dimensions, weight, materials, finish, colour name, pack quantity, included components, variation names, compatibility, care instructions, country of origin where required, and any safety or compliance information.

Jewellery sellers should be especially strict. Record whether the metal is solid, filled, vermeil, plated, or merely gold-coloured. Note the fineness, stone identity, whether a stone is natural or laboratory-grown, stone weight, ring size, chain length, clasp type, and any treatment or certification that applies. A photo cannot reliably prove any of these.

Create the draft, then audit in risk order

In Seller Central, go to Catalog, choose Add Products, and open Product Images. Upload the evidence pack, submit it, and let Amazon propose the listing fields. Save the result as a draft before making corrections.

Review the draft in this order:

First, check product identity and category. A wrong category can load the wrong required attributes, create irrelevant browse placement, or lead to suppression. Confirm that the listing is for the exact product rather than a visually similar type.

Second, check every number and variant. Dimensions, counts, capacity, weight, model numbers, colour names, and size relationships must match the SKU. For a jewellery variation family, confirm that yellow gold, rose gold, and silver-coloured versions have not been collapsed into one material claim.

Third, remove unsupported claims. Delete words such as waterproof, hypoallergenic, sustainable, medical-grade, genuine, handmade, or child-safe unless your records support them and the wording complies with the relevant rules. Do the same for inferred materials and gemstones.

Fourth, edit for buyers. Replace generic praise with useful details: what is included, how the product fastens, what it fits, and what a shopper should measure before ordering. Keep important search terms natural. Amazon’s AI can suggest copy, but it does not replace your knowledge of customer language or the item’s real limitations.

Know when images are the wrong starting point

Image-led creation is best for visually distinctive, relatively simple products. It is a poor starting point for bundles with many components, nearly identical electronic models, products whose main differences are internal, regulated goods, replacement parts with precise compatibility, or jewellery variants separated only by fineness or stone treatment.

For those items, start with a written description, an authorised product-page URL, or a structured spreadsheet. Amazon supports all three routes. Even then, use the truth sheet and the same audit order.

Keep evidence images separate from sales images

The photos that help AI identify a product are not automatically the best gallery. Once the listing data is correct, build the customer-facing sequence: a compliant main image, alternate angles, scale, detail, use, packaging, and any permitted explanatory graphics. Amazon’s main-image rules require an accurate view of the product on a pure white background and prohibit distracting additions.

This separation also makes creative testing safer. You can change lifestyle scenes or supporting graphics without treating each visual experiment as a source of catalogue truth. If you use AI to create promotional imagery, compare the result with the original product and check geometry, colour, logo, texture, and included components before upload.

The useful role for AI

The real time saving is not publishing a listing from a photograph without reading it. It is replacing a blank form with a structured first draft. Photograph visible facts clearly, supply invisible facts yourself, and review the riskiest fields before polishing the copy. That keeps the speed benefit without turning a plausible visual guess into a product promise.

Sources

Amazon, “How to use Amazon AI to create product listings,” published 3 November 2025: https://sell.amazon.com/blog/amazon-listing-ai

Amazon UK, “How to Create a Listing on Amazon”: https://sell.amazon.co.uk/learn/create-listing

Amazon, “How to create Amazon product listings,” published 29 April 2026: https://sell.amazon.com/blog/amazon-product-listings

Sources reviewed 7 October 2026. Feature availability, interface labels, and marketplace rules can change, so check the current Seller Central guidance for the marketplace and category you use.