51% of US shoppers now use AI tools to shop, NIQ says

AI shopping adoption just crossed half the US market — on a survey of about 500 people. NIQ published new Agentic Commerce Tracker data today showing 51% of US consumers used at least one AI-powered tool to support shopping in the past month. In May, the same tracker reported 42%. Nine points in roughly four months. Both numbers deserve your attention and neither deserves your full trust — and that combination is the actual story for anyone selling online.

What actually happened

NIQ's tracker is a monthly US study run through its Quick Question research product, with a sample of roughly 500 consumers. This month's read: 51% used at least one AI shopping tool in the prior 30 days. Broken out, AI-powered product recommendations lead at 20% adoption, and AI-powered personal shopping assistants follow at 16%. Respondents described using AI to compare options, evaluate value, and narrow choices — not to complete purchases.

Liz Buchanan, NIQ's North America president, called crossing 50% "a defining moment for the industry" and framed the shift as the path to purchase moving "from search-driven to AI-driven."

Read the structure of that 51% carefully. It is an umbrella figure covering any AI touch in a month. The two named behaviors underneath it sit at 20% and 16%. That gap is where most of the hype lives.

Why AI shopping adoption matters for your business

The honest read is that discovery moved and checkout mostly did not. That is a narrower change than "AI is buying things now," and it is far more actionable.

Your product data is now your storefront. When a recommendation engine narrows a customer's list, it reads structured attributes: specs, dimensions, materials, compatibility, return terms. Thin descriptions and missing fields do not lose you a ranking — they lose you a mention. We spend more client hours on attribute completeness than on any front-end work, because that is what the machine reads.

Measure the referral, not the vibe. A 51% headline is not a line in your analytics. Assistant and chat traffic arrives with inconsistent referrers and often none. Tag it deliberately, segment it, and watch conversion separately from organic search. Without that, you will spend on AI discovery and never learn whether it worked.

Treat a 500-person monthly panel as a direction, not a forecast. At n=500, the margin of error on a topline is a few points either way, and a self-reported "used an AI tool" is a soft definition. The trend from 42% to 51% is credible and worth planning around. Any single decimal in it is not. If a vendor quotes this number at you as justification for a six-figure platform, the number is not the argument.

The buying step is still yours to lose. Respondents compared, evaluated, and narrowed. Then they landed somewhere. Make sure the somewhere is a page that loads fast, shows stock honestly, and does not ask an already-decided customer to re-decide.

Sources: NIQ, BusinessWire.