More and more shoppers skip the search box and use their camera instead. They point Google Lens at a lamp in a café, circle a jacket in an Instagram video, or snap a friend's sneakers. Then they buy whatever the search engine matches.
That's visual search, and for e-commerce it changes what your product photos are for. They're no longer just there to convince someone already on your page. They're how a camera-first shopper finds you at all. This guide covers how big visual search is, how Google Lens shopping works, and how to optimise your product images so the match is yours.
How big visual search is
- Google Lens: nearly 20 billion visual searches a month, and 20% of them are shopping-related, according to Google. That's roughly 4 billion shopping searches a month that start with a picture.
- Circle to Search: Google says shopping-related searches are among its top uses, and since February 2026 it can search every item in an outfit at once.
- Google AI Mode runs visual searches against a Shopping Graph of more than 50 billion product listings, about 2 billion of which are refreshed every hour.
- Amazon Lens Live matches what your camera sees against billions of Amazon products in real time.
And Google now sells ads against it: Shopping ads appear above and alongside visual search results, and your product listing ad can show up for a visual search the same way it does for a typed one.
How Google Lens shopping works
When someone runs a Lens search on a product, Google tries to work out what's in the frame and then shows matching products from the Shopping Graph: listings from Merchant Center feeds and product pages it has crawled. Shoppers see image cards with price, store and often reviews, and they tap the one that looks most like what they pointed at.
Two things follow from that:
- Your images are the match. If the shopper photographs your product from the side, at an angle, on a shelf, your listing needs images that look like that, not just one perfect front-on studio shot.
- Your images are also the pitch. Lens shows several lookalikes side by side. The clearest, most accurate image gets the tap.
How to optimise product images for visual search
1. Show the product from the angles people see it
Google's Merchant Center guidance for additional images says to "show the product from different angles or the product being used", and allows up to 10 extra images per product. For visual search, this is the single most useful change you can make. A shopper photographing a chair in a restaurant sees it from three-quarters, not from straight ahead.
A practical minimum per product:
- front view on a clean background (your main image)
- three-quarter or side view
- a detail shot of the most distinctive feature (texture, hardware, label)
- the product in use or in a real setting
2. Make the main image clean and large
Merchant Center image requirements: at least 500×500 px, ideally 1500×1500 or larger; the product should fill 75–90% of the frame; no watermarks, promo text, borders or placeholder images. A plain or neutral background makes the product's shape easy to read. Google doesn't require white, but a cluttered main image hurts matching. Our white background guide covers the basics.
3. Add lifestyle images in their own field
Google has a separate lifestyle_image_link attribute for up to 5 in-context images per product (minimum 600×600, no promo text or logos). In-context shots matter for visual search because shoppers' photos are in context too: a vase on a table, a bag on a shoulder.
4. Keep every image true to the product
Visual search matches what the product actually looks like. If your images show a warmer colour, a cleaner label or a slimmer shape than the real item, you get two problems: worse matches, and returns from buyers who did get matched. This is where AI images go wrong most often. We found label text and colour drift were the top failures across 111 flagged renders.
5. Do the basic image SEO
From Google's image SEO guidance:
- Alt text is "the most important attribute". Google uses it "along with computer vision algorithms". Describe the product plainly: "sage green linen duvet cover on a queen bed". Don't stuff keywords.
- File names should be "short, but descriptive". Use
sage-linen-duvet-queen.jpg, notIMG_4821.jpg. - Put images near relevant text on the product page, and include them in an image sitemap.
- Use modern formats (WebP and AVIF are supported) at high resolution.
6. Label AI images properly
Google allows AI-generated product images in Merchant Center, including as the main image, but they must carry the IPTC DigitalSourceType metadata (TrainedAlgorithmicMedia). Many compression and export tools strip metadata, so check yours keeps it.
Why visual search is going to need more images per product
Visual search isn't the only thing pushing image counts up. The newest shopping features generate personalised visuals from your catalog:
- Meta's Muse Image lets someone upload a photo of their room and have Meta AI redesign it with real products from businesses' catalogs, then click through to buy.
- Google's "try it on" puts shirts, pants, dresses and shoes from Shopping listings onto a shopper's own full-length photo.
Both work better with clear, accurate images from several angles. One hero shot per SKU isn't enough any more. The realistic target is 4–8 accurate images per product, every product, every variant.
How to get there without reshooting everything
For most stores, the bottleneck is producing those images, not knowing what's needed. A studio day per collection doesn't scale to every variant.
ProductAI generates the extra images from the product photos you already have: clean main shots on neutral backgrounds, plus lifestyle and in-use scenes in a consistent style across the catalog. Because visual search punishes inaccurate images, every render goes through a fidelity check that compares it with your original photo and flags changed labels, colours or missing parts before you publish.
Want to know how your current images stack up? Get a free visual audit of your store, or try the free AI background generator on one product.
Bottom line
Visual search for e-commerce isn't a future trend. Google Lens alone handles billions of shopping searches a month, and Google already sells ads against them. The stores that get matched are the ones with several clear, accurate images per product, shot from the angles shoppers actually see. Start with your best sellers: add angles, add an in-context shot, fix your alt text, and make sure every image looks exactly like what you ship.
