Make Amazon Listing Images with ChatGPT or Gemini (2026 Guide)

Yes, you can make Amazon listing images with ChatGPT or Gemini, and many sellers do. They work well for lifestyle scenes, backgrounds and simple infographics. They are weaker at the two things Amazon shoppers care about most: showing your exact product and spelling every word correctly. This guide shows a workflow that works, prompts you can copy, and the checks that catch the usual mistakes.

What ChatGPT and Gemini are good at

Both are general image models. You describe a picture in plain words, add reference photos, and ask for changes in follow-up messages. Reviews from tool vendors such as Designkit describe ChatGPT's image model as good at lifestyle scenes, realistic lighting and short, simple text. A guide to Google's Gemini image model, known as Nano Banana Pro, says it accepts many reference images at once and handles text better than older models (E-commerce AI newsletter). Treat these vendor claims as directional, because none of them reports measured results, and models change quickly.

• Good fit: lifestyle scenes, new backgrounds, seasonal versions, simple infographic layouts and mood.
• Weaker fit: the product itself, printed text on the pack, long sentences, and keeping several images consistent with each other.

Step-by-step: make an Amazon listing image with ChatGPT or Gemini

1. Collect 3 to 6 clear photos of your real product. Include the front, the back with the label, a side view and a close-up of any detail that matters. Plain backgrounds and good light give the model the best reference.
2. Decide the image's one job. Main features, lifestyle, dimensions, what's in the box. One message per image. Our listing images guide lists the nine slots.
3. Upload the photos and write a specific prompt. Describe the layout, the scene and the exact text, and tell the model to keep the product unchanged (examples below).
4. Generate one image at a time. Change one thing per follow-up message, for example only the background, and say what to keep identical.
5. Keep text short or add it yourself. Ask for a headline of a few words at most, or generate the picture without text and add the words in a design tool such as Canva or Photoshop, where you control every letter.
6. Check the product against the real one. Compare colour, shape, label, parts and quantity. Regenerate until it matches.
7. Export at the right size. Use at least 1,000 pixels on the longest side so shoppers can zoom, and 2,000 pixels is a common target. See the image size cheat sheet. Check the file after any resize or crop.
8. Follow Amazon's rules. Read are AI-generated Amazon listing images allowed, including the label for photorealistic AI-generated people.

Prompt templates you can copy

Replace the words in brackets. Keep the instruction to preserve the product in every prompt.

• Lifestyle: 'Use the attached photos as the exact product. Do not change its shape, colour, label or proportions. Show it in [scene, for example a bright modern kitchen] with [who uses it, for example a woman in her 30s pouring a drink]. Natural window light, shallow depth of field, no text, no logos.'
• Features infographic: 'Use the attached photos as the exact product and keep it unchanged. Product large on the right. On the left, a headline "[3 to 5 words]" and three short callouts with simple icons: "[feature 1]", "[feature 2]", "[feature 3]". Clean [colour] background, readable at phone size.'
• Dimensions: 'Use the attached photos as the exact product. Show it straight on with dimension lines labelled "[height]" and "[width]". Minimal white and grey layout.'
• Fix one thing: 'Keep everything exactly the same, including the product. Only change [the background to a light wooden table].'

Where ChatGPT and Gemini fail for Amazon listings

1. The product drifts. Reviews report that shapes, textures and small details can change, and that consistency across images is not guaranteed. In our own work building APlus Listing, a recurring problem was a product that came out in the wrong colour in one image of an otherwise correct set. Check every image.
2. Printed text is unreliable. A test by Rewarx, a vendor of AI image tools, called text and labels the most serious weakness. Pack labels, ingredient lists and fine print are the riskiest. Never trust the model's version of your label.
3. Textures and reflections. Glass, glossy packaging, leather, fabric and metal can look smoothed or have odd reflections, which shoppers notice.
4. A set does not match. Each image is a separate conversation turn, so colours, fonts, icon style and lighting vary across the nine slots unless you manage them carefully.
5. It takes time. Writing prompts, regenerating, fixing text and resizing for nine images adds up, and the time is yours.

ChatGPT or Gemini: which should you use?

Try both on the same product and compare the results with your real photos. Gemini's image model is promoted for working with many reference images at once, but the best choice depends on your product and the current model versions. Check current limits and pricing before you plan around either, since higher resolutions may need a paid plan.

When a dedicated tool is the better choice

A general chat tool is a good way to make one or two images. A dedicated listing tool is better when you need a complete, consistent set, because it adds what a chat window does not:

• Your product photos as a fixed reference for every image, with a check that the result matches the real product.
• One consistent style across all the images and your A+ Content.
• Automatic text proof-reading, so misspelled words are caught before you see them.
• Amazon sizes built in, for the listing images and for the desktop and mobile A+ bands.

APlus Listing works this way: you enter an ASIN, and it builds the A+ Content and listing images from your real product photos and listing details. One product costs 10 credits ($10), and regenerating an image costs 1 credit. See the pricing page, the examples and our guide to making Amazon listing images with AI.

Quality checklist before you upload

• The product matches your real one in colour, shape, label, parts and quantity.
• Every word and number is spelled correctly and every claim is true.
• No invented logos or brand marks in the background.
• Main image: pure white background, product filling about 85% of the frame, no text.
• Photorealistic AI-generated people are tagged as Amazon requires.
• The file is large enough and the text is readable at phone size.