Your next order will start with a photograph of something that is not for sale.
A stranger's jacket on a train. A chair in a hotel lobby. A bowl on a friend's shelf. Those moments used to end in a vague memory and a failed keyword search two days later. In spring 2026, Shopify's Catalog API added image and multimodal search, which means a photo can now be a query against your catalog directly.
The build is smaller than the idea. A camera-first experience, either a lightweight web app or a Shop Mini, opens straight to a viewfinder. The customer shoots. You return matches from your catalog, plus near neighbors when there is no exact hit. No keywords, no filters, no product name the customer never knew.
The interesting decision is where the camera lives. Put it in your customers' hands and you get visual reorder: shoot the empty bottle, get the refill. Put it behind a QR code on a hangtag and every unit you have ever sold becomes an entry point. Put it in a store associate's phone and a competitor's product in a customer's hand becomes a comparison you get to answer on the spot.
The catch is your imagery. Image search is only as good as the reference set behind it, and most catalogs are built for a white-background grid, not for matching a blurry photo shot at an angle in bad light. If your line is 200 SKUs of near-identical black tees, expect confident wrong answers. Categories with real visual variance do far better: furniture, prints, patterned apparel, packaging. Budget for reshoots, multiple angles, and a failure state that says "closest matches" instead of pretending certainty. Test it against photos your customers would actually take, not the ones your studio produced.
Search boxes assume the shopper can name the thing. Half the time they can only point at it.
