AI Shopping Is Starting to Replace the Search Box — What That Means Before You Buy
SHOPPING · AI · EXPLAINER
Online shopping is quietly moving from “search, filter, compare” toward “ask, shortlist, buy.” One useful signal comes from British retailer John Lewis: AI-agent-driven shopping now accounts for 2.5% of its product searches, up from just 0.3% a year earlier.
That is still a small share, but the growth rate matters. Retailers are beginning to redesign product information for machines that may choose the shortlist before a human ever opens a product page.
What is AI-agent shopping?
Traditional ecommerce asks you to translate a need into keywords: “quiet cordless vacuum small apartment,” for example. An AI shopping agent can start with the need itself — apartment size, pets, budget, storage constraints — then compare products and surface a shortlist.
The difference sounds subtle, but it changes who controls discovery. Search rankings, category pages and filters matter less if an AI system reads specifications, reviews, availability and policies across multiple sources before recommending three products.
Why retailers are paying attention now
John Lewis told Reuters that AI-agent shopping has expanded across age groups, not just younger early adopters. The retailer is investing more heavily in content creation as it adapts to this new discovery layer.
That makes sense. Products with vague specifications, inconsistent model names or missing dimensions are difficult for both humans and AI systems to compare. In an agent-driven store, structured information becomes part of the merchandising.
Final price including delivery · return window · exact model/variant · whether the recommendation is sponsored or commercially influenced.
The convenience problem
An AI agent can reduce the exhausting part of shopping — opening 20 tabs to compare nearly identical products. But compression creates a new risk: the reasons behind the shortlist can disappear.
A recommendation that feels objective may depend on which retailers expose clean data, which reviews the system can access, and which commercial relationships are integrated into the platform. Convenience is useful; invisible selection criteria are not.
A better way to use AI for shopping
Use an agent to narrow a large category, not to outsource the final decision. Ask it to state the criteria it used, identify what information is missing, and compare the final two or three products against manufacturer specifications and current retailer policies.
The best use of AI shopping may not be “buy this for me.” It may be “remove 90% of the options I never needed to see.”
Sources: Reuters, Sep. 3, 2026. The 2.5% and 0.3% figures are John Lewis figures reported by Reuters.
댓글
댓글 쓰기