Amazon announced that it will introduce a new artificial intelligence feature into the search interface of its shopping application: when users enter search terms, the system will generate a series of AI-synthesized product images based on search intent, which will be displayed below the search association results to help users "better find the products they want." This approach is regarded as a new attempt by e-commerce platforms to apply generative AI in search scenarios, but it is also considered to be one of the most controversial AI use cases currently.
According to Amazon’s introduction on its official blog, this feature is mainly for users who “have specific needs in mind but can’t tell the professional name.” For example, people who don’t know the clothing style terminology of “cowl neck” or are not familiar with the concept of furniture materials such as “rattan”. Under the existing search mechanism, it is often difficult to return accurate results with such vague descriptions. Amazon hopes to use AI to automatically generate a batch of product diagrams that roughly match the search description, allowing users to gradually narrow the search scope by "looking at the pictures and clicking on them."
In actual interaction, when the user enters a search term such as "blue plaid dress", the system will display multiple AI-generated dress images of different styles below the automatic completion suggestions, including different sleeve types, skirt lengths, cuts and other options. After the user clicks on an image, he will be directed to a product result page that is closer to that style. At the back end, Amazon's visual search technology completes the matching and sorting of similar styles.
However, doubts surrounding this feature quickly surfaced. Critics pointed out that for a retail platform with physical transactions as its core, displaying "non-existent product photos" at the key decision-making entrance can easily lead to misleading. Some users may not notice that the image is synthesized by AI, and mistakenly believe that they can buy the exact same product in the picture. Eventually, they will be disappointed when they cannot find the corresponding product on the actual product page. What’s even more confusing is that Amazon itself already has a large number of real product images, but still chooses to batch generate “fictitious products” during the search stage. This logic itself is considered “somewhat outrageous.”
This AI product image search function is another example of Amazon’s intensive application of generative AI in its retail business in the past year or two. Among the more recognized uses, Amazon has launched a review summary function based on large models, which can automatically extract the main advantages and disadvantages of product reviews, allowing users to quickly grasp key information without reading long reviews one by one. In comparison, Amazon also launched a more "unexpected" short audio product summary function last year, in which AI verbally summarizes the selling points of a certain product to users in a few sentences in a format similar to a podcast explanation.
In addition to text and voice summaries, Amazon is also exploring more AI shopping experiences combined with visual search. For example, it launched "shoppable collages" generated by AI, which combine products around a specific outfit or style into a page of selected images and text to guide users to complete purchases on the theme page. In addition, the "Amazon Lens Live" function uses the camera to scan physical objects in the user's field of vision and find similar-looking products on the platform; users can also overlay text descriptions in visual searches, or directly initiate image searches with the help of iOS lock screen widgets.
In the direction of conversational shopping, Amazon’s previously launched Rufus AI chatbot has recently been replaced by the new “Alexa for Shopping”. This upgrade means that users can interact with Alexa in a more natural language through voice or text in the search bar, and put forward complex requests such as "Help me find a blue dress suitable for summer commuting", and the AI assistant will comprehensively understand and return matching results. Now superimposed on AI-generated image search, Amazon is trying to transform its shopping path through multiple entrances such as "Look at Pictures", "Conversation Search" and "List Browsing".
In a larger industry discussion, Amazon's attempt to use AI synthesized images as a search entry once again touches on the boundary issues of artificial intelligence application in the field of e-commerce: how to improve search efficiency and interface "visualization" while avoiding excessive fiction, misleading expectations, and even blurring the boundaries between real products and imaginary products. For regulatory agencies, consumer protection organizations and ordinary users, how to identify, explain and restrict such functions in the future may become a new proposition that must be answered jointly between platform innovation and user trust.