9 Steps: Visual Search That Sells
A simple checklist for e-commerce owners: 9 steps to make product images understandable for visual search and AI agents. No coding required.

Key takeaways
- Images must clearly show the product: standard angles, neutral backgrounds, scale, and variants.
- Write alt texts and file names as if for a person; avoid 'IMG_1234'.
- Add structured data for Product and ImageObject, along with an image sitemap.
- Use WebP/AVIF for speed and ensure proper rights and labeling for AI images.
Customers are increasingly searching 'by image' (visual search). AI shopping agents are doing the same. If your photos are 'readable', you'll sell more. Here are 9 simple steps for your marketing team—no coding needed.
Images That Humans and AI Understand
Visual search is finding a product by its image, like using Google Lens. An AI agent is a shopping assistant on your phone or browser that understands both images and text. Provide them with images that leave no questions unanswered.
Google has been developing image search for years (like Google Images and Lens). Now is a great time to organize your photos and metadata so that both people and agents can understand your offerings.
- 1. Required shots: front, back, side, close-ups, and a lifestyle shot (showing the product in use). If possible, add a 360° view or a short rotating clip. Conclusion: show the full picture without guesswork.
- 2. Background and lighting: use a neutral (white or light gray) background, no patterns, and even lighting with a 10% margin around the product. Conclusion: nothing competes with the product.
- 3. Scale in the frame: add a reference point (like a hand or ruler) or include a measuring tape in the shot. Conclusion: fewer returns due to 'I thought it was bigger'.
- 4. Variants and colors: show each color and significant variant as separate images. Conclusion: both the agent and customer can match the exact model.
Descriptions, Files, and Data – So AI Can Understand
Alt text (alternative text) is a short description of an image that is visible to screen readers and when an image fails to load. For AI, it serves as a hint about what the image contains—like a caption under a picture in a newspaper.
Structured data (schema.org) consists of labels in your webpage's code that clearly state: 'this is a Product, and this is an Image of that product'. Support for this varies by platform and theme: some offer ready-made settings or plugins, while others may require manual addition. If your platform doesn't support it or needs tweaking, ask for configuration help from a developer.
AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization) involve preparing content so that assistants and generative AI engines can easily understand and use it in their responses.
- 5. File names and alt texts 'for humans': instead of 'IMG_1234.jpg', use 'brown-leather-mens-shoes-model-alpha-side-view.jpg'. Alt text: 'Brown leather men's shoes, model Alpha, side view, with non-slip sole'.
- 6. Structured data for Product and ImageObject: ensure each product card has schema.org/Product, and images have schema.org/ImageObject with captions, dimensions, and links. In Shopify and WooCommerce, this is usually a
- 7. Image sitemap: this is a list of images for search engines and agents. Most platforms generate this automatically or via a plugin (like Yoast SEO). Conclusion: nothing gets 'lost'.
Delivery and Trust – For Speed and Security
Crawlers (these are 'spiders' or robots that scan webpages) and agents need to download your images quickly. The file format matters here.
Also, ensure you have rights to the images and clear labeling—lack of licenses or misleading images can limit visibility in visual search results and shopping channels. If an image was created using AI, clearly label it—this builds trust and may be required in advertising.
- 8. Format and size: use WebP or AVIF (these are modern image formats—smaller files with the same quality). For the longer side, aim for 1200–1600 px and a file size usually up to 300 KB. Leave an automatic fallback to a
- 9. Rights and AI labeling: ensure you have licenses (stock, photographer, model release for faces). If an image is generated or heavily modified by AI—label it in the description or tag. Conclusion: zero disputes.
- Tip: check in your protection tools and CDN (like Cloudflare) to ensure no rules are hindering trusted services from downloading thumbnails and metadata. Test the product in Google Lens and with a popular AI assistant to
Apply these 9 steps and treat your images like a first-contact salesperson: they should clearly show, describe, and be easily found. Want to go through the checklist with examples from your store? Schedule a brief consultation—we'll evaluate 10 random product cards and point out quick wins.
Frequently asked questions
Do I need to take 360° photos?
No. Standard shots (front, back, side, details, and 'in use') are sufficient. 360° helps with spatial products (like shoes, bags), but it's a nice addition, not a requirement.
How long should alt texts be?
1-2 short sentences, specifically: product type, brand/model, color, view ('side', 'top'), and a key feature. Avoid keyword stuffing.
Will WebP/AVIF work in my store?
In most Shopify and WooCommerce themes, yes. If in doubt, enable conversion in your CDN or plugin and leave an automatic backup format of JPG for older browsers.
Do AI agents really read structured data?
Many assistants and search engines use HTML and schema.org to understand context. It's not a guarantee of ranking, but it genuinely helps with accurate matching to queries and responses.