AI now handles the research phase of shopping for most American consumers. A January 2026 Clutch survey found 70% of US shoppers use AI somewhere in the buying process, and 65% use it before spending anything. Klaviyo's survey of nearly 8,000 consumers across eight countries found 41% had bought something an AI recommended in the prior six months. The tool is no longer a novelty at the edge of commerce. It is the front door.
The mechanism behind this shift is specific and worth understanding. The piece traces how cognitive load, automation bias, and the paradox of choice combine to make a single AI verdict feel better than a messy list of links, regardless of whether the verdict is correct. The jam experiment from Columbia is here: 24 options drew bigger crowds but a 3% purchase rate, while 6 options produced a purchase rate near 33%. That same dynamic now plays out at scale inside chat windows. Researcher Chiara Longoni's 'word-of-machine effect' adds another layer: people trust algorithmic picks for functional purchases like snow shovels, but resist them for sensory or personal ones like scarves. The line between helpful recommendation and paid placement is where this gets genuinely uncomfortable.
This article is worth reading in full because it does not stop at behavior. It maps what happened to the people building the content that feeds these systems, who now write for crawlers as much as for humans, and it asks who is waiting in the wings when judgment gets offstaged. Only 4% of shoppers will let AI complete a purchase autonomously. We kept the wallet. We gave away everything else.
[READ ORIGINAL →]