Journal / Kirill Naumov
Target’s 2026 AI shopping features bring photo search, review summaries and reordering together. The real test is whether each helps a shopper make a sound decision.
By Kirill Naumov ·
A shopper takes a photo of a lamp in a hotel lobby. Ten minutes later, she is looking at similar lamps on her phone. Wonderful. But “looks like it” is only the first question. Will it fit the room? Is the colour accurate? Can she pick it up nearby? If the answer to the last three is unclear, a smart search has simply helped her reach a disappointing product page faster.
Target’s September 2026 announcement describes several AI-powered shopping features in its app: Photo Search, Review Insights, Buy Again and Continue Shopping. Target says Photo Search turns an image into product discovery; Review Insights groups common points in customer feedback; Buy Again helps people reorder familiar items; and Continue Shopping reconnects them with products they viewed earlier. These are Target’s descriptions of its own features. They do not, by themselves, tell us how often a particular shopper gets the right answer.
The interesting part is how these features cover different jobs. One starts with “I saw something like this.” Another asks “Will this work for me?” A third says “I already know what I need; please do not make me hunt for it again.” Treating all three as one generic personalization problem would miss the point.
Photo Search can be delightful when the photographed object is distinctive. A retailer should also test the unglamorous cases: a colour changed by warm lighting, a product partly hidden behind furniture, a close-up with no sense of scale. When confidence is low, a useful result explains the uncertainty and gives the shopper a way to narrow the choice. A neat grid of vaguely similar objects is not the same as an answer.
Target says Review Insights identifies themes such as fit, comfort or durability. That can save time. But a summary should not turn five conflicting experiences into one confident verdict. A shopper buying a vacuum wants to know whether the complaints concern suction on thick carpet or on hard floors, and whether those reviews apply to the current model. The original reviews need to remain available. AI is useful here as a map to evidence, not a replacement for it.
Buy Again and Continue Shopping can remove repetitive work. The risk is treating every past click as a lasting preference. Someone may have researched a gift, compared a one-off appliance or abandoned an item because it was wrong. “Continue” should make it easy to dismiss the suggestion, see why it appeared and correct the memory. Reordering should be especially careful with substitutions, current price and availability.
Target reports encouraging early engagement and add-to-cart signals for some features. Those are company-reported observations, not proof that every recommendation improved the final purchase. For another retailer, I would watch completed purchases alongside returns, corrections, customer-service contacts and repeat use. If photo search raises clicks but sends more people to the wrong size or finish, the apparent win is expensive.
The practical lesson is to design AI around three different customer promises: discovery, confidence and convenience. Test each promise against its own failure case. That sounds less exciting than “an AI-powered shopping journey”. It is also far closer to what a person actually needs when standing in a room with a lamp she likes.
Explore the English AI for Retail workshop → · Discuss a retail AI consulting brief
Primary source: Target’s September 2026 overview of its AI shopping features. Descriptions and early engagement signals are company-reported; the evaluation framework here is an independent proposal, not a Target project.