For the past eighteen months, anyone arguing that AI was eating ecommerce search had to fight for it. The SEO industry is large, conservative and well-paid, and most of it spent the year telling clients to keep doing what they were doing — publish content, build links, optimise for Google, ride out the LLMs.
That argument is now effectively over — and it was the SEO industry’s own paper of record that ended it. In April, Search Engine Land covered a survey of just over 1,000 US consumers run by Exploding Topics. The headline numbers:
- 77% have used AI to shop in the past six months.
- 44% start their purchase journey inside an AI assistant — not Google, not Amazon.
- Another 44% start elsewhere and bring AI in before they buy.
- Only 2% actually check out through the AI.
- 31% say they wouldn’t authorise an AI agent to spend a single dollar on their behalf.
Most of the coverage read those numbers as agentic commerce is here. That’s the headline-friendly read, and it isn’t quite right.
The useful read: discovery, not checkout
The far more useful read is this: AI is now a discovery channel, not a checkout channel. Roughly half of shopping journeys begin in an assistant. Almost none of them end there.
That distinction matters because much of the AI-commerce industry has been building for the wrong half of it. The high-profile bets — instant checkout inside chat, agent-payment protocols, wallet integrations — are all checkout bets, moving the transaction inside the AI surface. The consumer data says shoppers haven’t agreed to that yet, and early launches keep running into the wall of distrust that the 31% figure represents.
Meanwhile the thing consumers are doing — asking an assistant “what’s the best X for Y?” — has already become the way half of America starts shopping. And most ecommerce teams aren’t optimised for it at all.
The line from the Search Engine Land piece worth pinning above your desk: “a huge base of potential customers are using AI as a starting point, so it is imperative that your brand gets organically mentioned.”
The publication that built the SEO playbook, naming the new game.
Two things this means in practice
First: the surface area is wider than most teams think. ChatGPT leads AI shopping usage, but Gemini, Perplexity, Copilot and Claude each carry a meaningful share of shoppers. If you’re auditing visibility in one engine only — and most agencies audit in one, usually ChatGPT — you’re flying blind on a large share of the surface. The work has to be cross-engine or it isn’t really being done.
Second: the mechanism is different from ranking. SEO operates against a ranked list — page one, position three, climb the SERP. AI search operates against a single recommendation. The model picks one or two products, names them, and moves on. There is no page two. The discipline that gets you into that slot is closer to engineering than to content marketing: product information clean, structured, and comparable enough that the model can pick you accurately. We call that practice Reference Engineering; the industry calls the broader category GEO. Same problem, different vocabulary.
The fifteen-minute exercise
If you sell online, the most useful thing you can do this week is stop debating whether this matters and check whether your product information is actually answerable. Run the questions your customers ask before they buy through ChatGPT, Gemini and Perplexity. See whether you turn up. See whether you turn up accurately.
That fifteen-minute exercise will tell you more than any agency proposal you’ll receive this quarter. The argument is over. The work has only just started.
Source: Search Engine Land, “New data: 77% use AI to shop. Nearly 1 in 3 won’t let it spend,” April 2026.
See how you show up across the engines. Get your free GEO Score.