Amazon’s search bar will invent AI-generated products you can’t buy

Amazon has updated its mobile search bar to display AI-generated images of products in real time as users type their queries. The feature currently covers clothing and home goods, and it is designed to surface visual approximations of what a shopper is describing rather than returning actual listings straight away. Users can then tap an image that closely matches what they want and proceed to search for real products with a similar appearance.

The company frames this as a solution for situations where shoppers know what something looks like but cannot recall the specific terminology. Amazon's example is a shopper typing 'shirt with a draped collar' when they cannot think of the term 'cowl neck.' In those cases, having a generated image to anchor the search could reduce the friction of browsing through loosely matched text results.

The practical value of the feature seems somewhat narrow. When a user already knows what they want and how to describe it, generated imagery adds little to the experience - a direct text search would work just as well or better. The feature is more useful as a visual bridge for vague or imprecise descriptions, essentially letting the AI interpret an approximate description and offer a reference point before a real product search begins.

This sits within a broader pattern of Amazon integrating generative AI tools across its shopping experience, including AI-written product summaries and a conversational shopping assistant called Rufus. Adding image generation to the search bar extends that effort into the discovery phase of shopping, though the feature still depends on existing product inventory to deliver anything a customer can actually purchase. The gap between a generated image and what is genuinely available in the catalog remains a real limitation.