Blog / AI & search /
AI product photos: where they’re fine and where they aren’t
Say your product photo looks a bit flat, so you ask an image generator for a lovely lifestyle scene. It looks gorgeous. It also gave the bag a clasp it doesn’t have, and the colour is slightly off.
Beautiful, yes. Honest, no. The customer will get something different.
AI can help with image work, especially when it improves presentation without changing product facts. The line is accuracy. Don’t let a generated scene invent features, scale, material or use. And if you sell through marketplace feeds, check their disclosure and metadata rules before uploading an AI-made image.
Decide what the image needs to show
A product image helps a shopper understand what is being sold. The main photo should show the actual product clearly. Additional views can explain dimensions, texture, packaging, included components or how the item looks in a real environment.
Before using AI, write down what must stay unchanged: shape, colour, proportions, materials, labels, accessories, condition. For packaged food, the label matters. For furniture, scale and construction matter. A prompt asking the model to keep these details doesn’t mean it will.
A generated lifestyle background can work if it doesn’t suggest the props or accessories are part of the sale. If that could confuse someone, add a caption or show a separate, accurate product image.
Where AI assistance can help
Background removal, cropping, colour correction and image resizing can make a product asset easier to use across a catalogue. Generative tools may also help create a scene or extend a background. Review the output against the source photo and the product’s verified details.
A clean edit leaves the product itself alone. If the tool changes stitching, buttons, labels, ports or controls, reject it or have a human fix it. Plausible isn’t good enough. Compare it with the real item and the supplier’s approved images.
AI can also help a small team explore a composition before a proper shoot. Treat that as a concept, not automatically as final product photography. A mock-up may help decide the direction while a final image is made from the real item.
Where generated images can mislead
Never use an image that implies a feature, size, finish or accessory the product doesn’t have. Don’t alter a model’s body, skin tone or surroundings in a way that suggests a fit or outcome the business can’t support. An image of the product in use should show safe, realistic use.
Be especially cautious with products where visual accuracy affects safety, compatibility or health. A generated image may omit a warning, change a label or show an unsafe use. Use manufacturer-approved information and have a qualified person review claims where needed.
Do not use generative edits to conceal defects, alter before-and-after evidence or create fake customer scenes. If the image is presented as a real customer or real result, the viewer should not be misled about how it was made.
Follow channel requirements
Google Merchant Center’s 2024 product data guidance required AI-generated product images submitted in Shopping feeds to retain the specified IPTC metadata indicating generative AI. The exact requirements can change, so check the current feed documentation before publishing. Do not strip metadata during export or compression if the channel requires it.
A channel’s image rules are separate from what’s acceptable on your own product page. Check the attributes, disclosure requirements and policy for every destination. An image that passes in one feed may not suit another advert or marketplace.
Keep a record of which image is original, which is edited and where the final asset is used. This makes it easier to replace an image if a product changes or a platform’s rules are updated.
Review images before they go live
Use a checklist: compare the image with the real product, verify labels and accessories, check dimensions and crop, inspect for generated artefacts and confirm that the intended channel accepts the asset. Ask a second person to review when the image could affect safety or purchase decisions.
Check the product page on mobile and desktop. The image should load promptly, display clearly and not hide product information. Add descriptive alt text based on what the image shows; do not cram a keyword into the alt text or claim details that are not visible.
If the image changes what a customer thinks is included, replace it. I’d take a plainer photo over a return caused by a misleading one.
Keep a useful original asset
Store the original image and the approved edit separately. Use a predictable naming and version process so the team can tell which file belongs to which product. Avoid overwriting the only authentic photo with a generated variant.
When a product specification changes, review the imagery as well as the copy. A generated background can remain attractive while a small product detail becomes wrong. Give someone responsibility for reviewing the asset library when inventory or packaging changes.
For product image basics, see Image SEO on WordPress: file names, alt text and size and Shopify product titles: how long, what order, what to leave out. For a broader AI workflow, read AI and E-E-A-T: adding the experience AI can’t.
Related reading
- Image SEO on WordPress: file names, alt text and size
- Shopify product titles: how long, what order, what to leave out
- AI and E-E-A-T: adding the experience AI can’t
Put one generated product image next to the real item and the channel’s current requirements. What could a shopper misunderstand if you published it as it is?