Our Take
Showing fabricated products to guide real shopping is a solution to a problem that doesn't need solving—Amazon already has millions of real photos.
Why it matters
Amazon is betting that synthetic images help customers navigate search better. But the risk is higher than the payoff: misleading users about product availability damages trust faster than better search terms build it.
Do this week
Product leads at retail platforms: audit your visual search features for friction points before adding generative UI layers—the gap may not be image generation, it's search ranking.
Amazon adds AI-generated product photos to search results
Amazon announced Wednesday it will display AI-generated product images in its shopping app when users search for items without knowing the exact terminology. If you search "blue gingham dress," you'll see several synthetic dress variations (short or long sleeves, different lengths) below autocomplete suggestions. Clicking one redirects you to real product listings matching that style, powered by Amazon's visual search capabilities.
The company frames this as solving a user problem: shoppers often have something in mind but lack the right search terms. Amazon's examples include "cowl neck" for shirt styles or "rattan" for furniture.
This follows a series of AI features Amazon has rolled out on its retail site and app. The company already summarizes customer reviews via AI and last year introduced AI-narrated product summaries in podcast style. More recent additions include AI-generated "shoppable collages" for fashion curation, Amazon Lens Live for visual product matching, text-augmented visual search, and a Lock Screen visual search widget for iOS. Amazon also replaced its Rufus AI chatbot with Alexa for Shopping earlier this month to handle natural language shopping queries.
The UX case doesn't align with the actual friction point
The problem Amazon identifies is real: terminology gaps do affect search. But the solution conflates two separate issues.
First, displaying fake products risks misleading customers. A user who doesn't read carefully may assume the synthetic image represents an available item, click it, and land on a results page without an exact match. That's disappointment, not guidance.
Second, and more fundamental: Amazon owns a catalog of millions of real product photographs. Those real photos are what online shoppers actually want to see. If search ranking or autocomplete is the friction (which it likely is), the fix is better ranking or smarter suggestions—not inventing products that don't exist to illustrate a style.
The feature assumes customers benefit from seeing generative mockups. What the feature actually reveals is that Amazon's search experience for unfamiliar terms is weak enough that it needs synthetic intermediaries. That's a search problem, not an image problem.
Audit what customers actually need before adding generative UI
If your platform has high search friction around terminology gaps, generate better search suggestions, not fake products. Run user research on what happens when someone lands on a results page after clicking a synthetic image. If the conversion is lower than baseline (real product click-through), you've added cognitive load instead of removing it.
For teams considering similar features: the cost of user disappointment and potential trust loss outweighs the benefit of a prettier search funnel. Solve the search ranking problem first. If you can't, synthetic images won't fix it.