IBM-NRF Study: 45 Percent of Consumers Now Use AI During Their Shopping Journey — But 72 Percent Still Want to Touch the Product in a Store

Nearly half of the world's consumers — 45 percent — now use AI during their shopping journeys, according to a global study of 18,000 shoppers across 23 countries by IBM's Institute for Business Value and the National Retail Federation. They are using AI to research products (41 percent), interpret reviews (33 percent), and hunt for deals (31 percent). Yet 72 percent still shop in physical stores. AI is not replacing where people buy. It is reshaping how they decide what to buy before they get there.

The study, published January 7, draws on consumer surveys across 23 markets and a parallel survey of 200 retail executives. It has been covered by more than ten independent sources including PRNewswire, NRF's own research blog, Fibre2Fashion, and AI Journal. The scale — 18,000 respondents — makes this one of the largest consumer AI adoption datasets in retail.

The decision has already been made before the consumer arrives

The 41 percent figure for product research is the number that should concern every retailer without an AI strategy. It means that when a consumer walks into a store or opens a retailer's website, the comparison phase is increasingly already finished. AI did it.

The 33 percent using AI to interpret reviews represents a qualitative shift from how reviews worked for the past decade. Consumers are not scrolling through individual reviews anymore. They are asking AI to synthesize patterns across hundreds of them: "Do reviewers consistently mention noise issues?" "Is the sizing accurate or does it run small?" "What do people who bought this for outdoor use say after six months?" The AI extracts the signal from the noise in seconds. A consumer who has done this arrives at the purchase with a level of pre-research that no in-store sales associate can match.

The 31 percent using AI for deal-finding collapses what used to be a time-intensive process. Price comparison across a dozen retailers, coupon aggregation, cashback stacking, historical price tracking — tasks that once required multiple browser tabs and dedicated deal sites now happen in a single AI conversation.

Retailers know this is happening — and most are not ready

The same IBM study surveyed 200 retail executives. The results suggest a significant readiness gap. Fifty-four percent report persistent challenges integrating data across channels and systems. Fifty-one percent cite limited AI expertise within their organizations.

Caroline Reppert, NRF's Senior Director of AI and Technology Policy, framed the stakes directly: retailers that understand this shift "will be best positioned to earn trust, relevance, and long-term customer loyalty." The implication is clear. Retailers without machine-readable product data, structured and accessible review databases, and AI-discoverable inventory information risk losing the consumer's decision before the consumer even visits. If AI cannot find your product during the research phase, you do not exist during the research phase.

IBM's recommendations to retailers include redesigning the customer journey around AI decision moments, ensuring data readiness for AI consumption, deploying agents to reduce consumer uncertainty earlier in the funnel, and investing in internal AI skills. These are not incremental improvements. They are structural changes to how retail operates.

Trust remains the hard constraint on AI commerce

IBM found 45 percent of consumers use AI to help decide. Acosta Group's research from December 2025 found that only 12 percent trust AI enough to make a purchase on their behalf. That 33-point gap between "I will use AI for research" and "I will let AI buy for me" is consistent across consumer verticals. In travel, a Dune7 survey from April 8 found 71 percent want an AI booking assistant, but Skift reported on April 3 that only 2 percent want that assistant to book with full autonomy.

The pattern holds everywhere researchers measure it. Consumers adopt AI enthusiastically for research, comparison, and discovery. They resist it firmly for the final transaction — the moment money changes hands. The bottleneck is not technology. It is trust. And trust accrues to platforms that demonstrate independence from the transaction they are helping the consumer evaluate.

Mubboo's Take

The IBM-NRF finding that 45 percent of consumers use AI in shopping but only 12 percent trust AI to buy validates a specific platform position. When 41 percent of consumers use AI to research products, the question is: what content does that AI find? If it surfaces recommendations authored by the retailer selling the product, the advice is not independent. If it surfaces content from a comparison platform with no inventory to sell, the recommendation carries a different kind of weight. Independent comparison platforms do not become less valuable as AI adoption grows. They become the editorial layer that AI-assisted consumers rely on precisely because the platform has no stake in which product wins.