AI shopping has moved beyond a niche group of early adopters, at least according to a recurring US consumer tracker. NielsenIQ says 51% of US consumers in its September findings reported using at least one AI-powered tool to support shopping during the previous month.
That does not mean half of Americans handed a credit card to an autonomous agent. The category includes several kinds of assistance across the shopping journey. NIQ says AI-powered product recommendations were the most widely used application at 20%, while personal shopping assistants were reported by 16%.
AI shopping now covers more than asking a chatbot what to buy
The practical meaning of AI shopping is widening. A shopper may encounter AI when searching for a product, comparing options, receiving a recommendation, summarising reviews, checking whether something fits a requirement or narrowing a large catalogue to a handful of choices.
Those activities happen before the transaction, but they can shape which products ever reach the shortlist. This is why the shift matters to retailers even when the AI does not press the final purchase button. If an assistant becomes part of product discovery, it effectively becomes another shelf on which brands compete for visibility.
NIQ says its Agentic Commerce Tracker is a monthly study covering the US and Canada. The September US finding comes from its ongoing Quick Question research, with a monthly sample of approximately 500 US consumers during 2026. The result is therefore a self-reported snapshot from a modest recurring sample, not a census of all shopping behaviour.
The distinction is worth keeping in view because people can interpret the same tool differently. One person may consider an ordinary recommendation widget to be AI while another reserves the label for a conversational assistant. The survey’s value is in tracking a behaviour over time with the same research programme, rather than treating 51% as a precise measure of every AI-mediated purchase.
The commercial contest is shifting towards the recommendation layer
Traditional online retail optimised for search boxes, category pages and paid placements. AI assistants create a different path. A person can state an intent in natural language, add constraints and receive a much shorter answer than a search-results page.
That compression has consequences. When the assistant recommends three products instead of returning hundreds, appearing in the answer can matter more than ranking a few positions higher on a conventional results page. Product descriptions, structured attributes and reliable availability data become inputs to the assistant’s judgement.
NIQ has separately announced work with Similarweb on measuring the agentic shopping journey. The companies say the planned system will connect AI-driven discovery with product visibility, traffic and sales conversion. An initial version is expected in the fourth quarter of 2026.
That announcement is itself evidence of how retailers are thinking about the problem. Once AI becomes a discovery channel, businesses want the equivalent of search analytics: which questions people ask, whether a product appears, whether its description is accurate and whether that exposure eventually produces a sale.
Recommendations create a new form of influence
An AI shopping assistant feels personal because the conversation can include budget, preferences and constraints. That does not automatically make the recommendation neutral. The system still relies on product data, ranking choices, commercial relationships and whatever information is available to it.
LiveAIWire has examined how AI product recommendations can alter the shopping experience. As assistants become more common, the important consumer question is not only whether they save time, but why a particular product was surfaced and what alternatives were left out.
There is also a risk of false confidence. A fluent explanation can make a recommendation sound carefully reasoned even when a specification is missing or outdated. A shopper comparing an expensive appliance, insurance product or piece of equipment may need to verify details at the retailer or manufacturer rather than assuming the assistant’s summary is current.
This makes data quality a consumer issue as well as a retailer issue. If product information is inconsistent across sources, the AI can reproduce that inconsistency at the moment a person is trying to decide.
Older shoppers and occasional users may change the shape of adoption
Early technology stories often focus on young enthusiasts, but broad adoption depends on people who do not think of themselves as AI users. Shopping is one of the activities where the technology can become almost invisible because the value is practical: find the right size, compare two models or explain an unfamiliar specification.
LiveAIWire has previously looked at AI shopping assistance among older travellers. The more these features are embedded inside familiar retail or travel services, the less adoption depends on somebody deliberately opening a general-purpose chatbot.
That could make future measurement harder. People may use an AI-ranked shortlist, summary or assistant without knowing which model produced it. Self-reported surveys may then undercount some forms of assistance while overcounting others that are more visibly branded as AI.
Agentic commerce is the next step, not the same thing as today’s result
The phrase agentic commerce is often used for systems that can move from recommendation towards action, including completing parts of a purchase. That is a more consequential capability than helping a shopper compare products.
LiveAIWire’s background coverage of agentic shopping in retail shows why the distinction matters. An assistant that suggests a kettle can be wrong without spending money. An agent that places the order needs stronger controls around price, quantity, delivery details and returns.
NIQ’s 51% finding should therefore be read as evidence that AI-supported shopping is becoming familiar, not that autonomous purchasing has already become normal. The current mainstream behaviour appears to be assistance across discovery and decision-making.
Even so, that is a substantial change. Once consumers get used to asking an AI to narrow the market for them, the assistant becomes part of the purchase path before autonomous buying is required. Retailers may discover that the most important new customer is not an AI agent with a wallet, but a human shopper who increasingly lets an AI decide what is worth looking at.
What shoppers should still check for themselves
AI is particularly good at reducing a large choice to a manageable shortlist, but that is also the point where a hidden mistake can have the most influence. If a shopper starts with hundreds of products and an assistant recommends only a few, an omitted model may never receive any further consideration. The convenience comes from trusting the first filter.
That makes verification more important when the decision has lasting consequences. Prices can change, warranties can differ by seller, product names can hide regional variations and a recommendation may be based on information that is no longer current. An assistant can help formulate the right questions, but the retailer’s current listing and the manufacturer’s specification remain the sensible place to confirm the answer before buying.
The same applies to personalisation. A useful assistant needs enough information to understand what the shopper values, but giving it more context can also reveal preferences, budgets and habits. Consumers may increasingly have to make a trade-off between a more tailored recommendation and the amount of information they are willing to share.
The next useful signal will be whether AI changes actual purchases
Usage is only the first stage of the story. A person can experiment with an AI shopping tool without allowing it to change what they eventually buy. The more important commercial question is whether AI-assisted discovery starts moving sales, changing brand choice or reducing the number of retailers a shopper visits before deciding.
That is why measurement efforts around agentic commerce matter. Retailers will want to separate curiosity from influence. Consumers will want to know whether the assistant is simply organising information or steering them towards a narrower commercial path.
NIQ’s result is therefore best treated as an adoption signal. AI has become familiar enough to enter ordinary shopping behaviour for a substantial share of respondents. The harder question now is what happens after it enters the decision: whether people verify its suggestions, whether brands can understand why they were selected, and whether the convenience ultimately improves the purchase or merely makes the route to it shorter in practice.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.