23% of AI Shoppers Say AI Advised Against a Brand

• ICERTIAS surveyed 8,000 adults with internet access across all 57 OSCE participating States during July and August 2026.

• 55% of respondents said they had already used AI while shopping, including product research, comparison, evaluation, or choice.

• 23% of AI shopping users said AI had advised against buying a brand they were already seriously considering.

• 61% of respondents preferred AI selecting three suitable products rather than personally comparing 20 alternatives themselves.

• 49% of AI shopping users were willing to buy without first visiting the recommended manufacturer’s official website.

• Purchase likelihood reached 65% with independent recognition, versus 53% when an unfamiliar brand received only an AI recommendation.

• 58% of AI shopping users check another source before acting, while 83% consider seeing recommendation sources important.

• 60% of AI shopping users discovered previously unconsidered brands through AI, while 33% had purchased such brands.

For decades, companies have competed to influence the moment when consumers choose between brands. Artificial intelligence is beginning to change where that competition takes place. The most consequential shift is not that AI is replacing the buyer. It is that AI increasingly influences which products the buyer seriously considers before making the final decision.

In July and August 2026, ICERTIAS conducted an international online study among 8,000 adults with internet access across all 57 OSCE participating States. The research examined how consumers use AI to discover products, compare alternatives, evaluate brands, assess prices, verify information, and decide what deserves a place in their consideration set.

The ICERTIAS survey found that 55% of respondents had already used AI while shopping. Among those AI shopping users, 30% said they probably would have bought a different brand without AI, while 23% said AI had advised them against buying a brand they were already seriously considering.

Taken together, these findings suggest that AI is no longer simply supporting product research. It is increasingly shaping which brands consumers consider, reject, and ultimately choose.

That figure requires careful interpretation. It is a self-reported counterfactual, not experimental proof that AI caused each purchase to change. Consumers cannot know with certainty what they would have bought in an alternative scenario. Yet the result remains strategically important because it measures something businesses cannot afford to ignore: consumers themselves perceive AI as influencing brand choice.

Three additional findings help explain how that influence develops. Sixty-one percent of respondents preferred having AI identify three products that best matched their needs rather than comparing 20 alternatives themselves. Among AI shopping users, 49% were willing to buy an AI-recommended product without first visiting the manufacturer's website. Independent evidence also substantially increased the appeal of an unfamiliar AI-recommended brand.

Taken together, these findings point to a change in the architecture of consumer choice. AI is becoming an intermediary between companies and customers, influencing what is discovered, what is compared, which information receives attention, and which products disappear from consideration. Consumers still choose what to buy, but increasingly they choose from a field that AI has already helped construct.

The New Scarcity Is Consideration

The internet dramatically increased the amount of information available to buyers. Search engines produced pages of results. Marketplaces created almost unlimited shelf space. Comparison platforms helped people navigate expanding assortments. For most businesses, the resulting challenge was visibility: how to appear frequently enough, prominently enough, and persuasively enough to enter the customer's consideration set.

Generative AI introduces a different proposition. Rather than helping consumers navigate hundreds of choices, it can eliminate most choices before the consumer examines them. When ICERTIAS asked respondents whether they would prefer three products selected by AI as the best fit or 20 alternatives they could evaluate themselves, 61% preferred the smaller, AI-curated set.

It would be easy to interpret this as a finding about convenience. Strategically, it is more important than that. When consumers allow AI to reduce a large market to three options, the system gains influence over one of the most economically valuable decisions in commerce: determining which companies receive meaningful customer attention.

Traditional search distributed visibility unevenly but broadly. A brand could rank second, seventh, or twentieth and remain discoverable. AI can compress the competitive field much more aggressively. If a customer asks for the three products that best fit a particular need, finishing fourth may increasingly produce the same commercial result as never being considered at all.

This creates a new form of scarce commercial real estate. Companies have spent decades competing for supermarket shelves, television exposure, search rankings, retailer pages, recommendation widgets, and social feeds. AI creates another position that may become equally consequential: a place inside the answer.

The strategic question therefore changes. It is no longer enough to ask whether customers know the brand, like the brand, or can find the brand. Companies increasingly need to ask whether an AI system evaluating a customer's needs has sufficient reason to place their product among the few alternatives that deserve serious consideration.

The Customer Journey May No Longer Lead to Your Website

The second shift concerns where persuasion occurs. Among consumers who use AI while shopping, 49% said they would be willing to buy an AI-recommended product without first visiting the manufacturer's official website. That finding challenges a deeply embedded assumption in digital strategy: serious prospects will eventually arrive somewhere the brand controls.

Historically, even when customers discovered products through advertising, search engines, retailers, recommendations, or social platforms, manufacturers could reasonably expect many serious prospects to visit the company website before buying. That visit created another opportunity to frame the product, explain its advantages, justify the price, answer objections, build confidence, and influence the final choice.

AI can shorten that sequence considerably. Thirty-seven percent of AI shopping users in the ICERTIAS study said they often or always begin product research with AI before going to a search engine, marketplace, or brand website. The journey may therefore become less dependent on repeated visits across multiple digital properties and more concentrated around a smaller number of mediated decisions.

Consider a customer searching for a refrigerator for a family of four, below a specified budget, with low energy consumption, strong reliability, and sufficient capacity. If AI identifies three models, explains their differences, summarizes relevant evidence, compares their value, and identifies where each can be purchased, the customer may see little reason to visit ten manufacturers' websites.

That does not make brand websites irrelevant. It changes their function. Information can no longer be designed only for people who eventually arrive on a company's own pages. Product data, specifications, pricing logic, warranty terms, service information, and evidence of performance increasingly need to remain understandable and persuasive when they are encountered outside a branded environment.

The uncomfortable implication is that a company may lose the customer before it has an opportunity to make its own case. A brand no longer needs to lose at checkout. It can lose when an AI system concludes that competing products better match the customer's needs and never places the brand among the alternatives receiving serious consideration.

Being Recommended Is Only the First Test

Appearing in an AI-generated shortlist does not guarantee a sale. Once AI introduces a product, the consumer still needs sufficient confidence to act. The ICERTIAS study tested this distinction by presenting respondents with an unfamiliar Brand X under different information conditions and measuring the proportion who expressed high purchase likelihood.

When AI recommended Brand X without additional supporting evidence, 53% of respondents expressed high purchase likelihood. When strong verified customer reviews accompanied the recommendation, the figure rose to 62%. When independent third-party consumer recognition based on market research supported the recommendation, purchase likelihood reached 65%.

The contrast with the brand's own communication is particularly revealing. When Brand X itself claimed that it offered superior value for money, purchase likelihood reached 54%, almost identical to the 53% recorded for the AI recommendation without additional supporting evidence.

The responsible interpretation is not that one form of external evidence necessarily outperforms another. The broader pattern is more important. Verified customer experience and independent third-party evidence produced substantially stronger purchase appeal than either the AI recommendation alone or the brand's own value claim.

That distinction matters because consumers evaluate the source of a claim as well as its content. Companies are expected to advocate for their own products. A manufacturer's statement that its product offers exceptional quality or superior value carries information, but it also carries an obvious commercial incentive.

Independent evidence answers a different question: is there a credible reason outside the seller's own communication to believe the claim? In an AI-mediated buying environment, that distinction may become increasingly valuable because systems comparing products need evidence that can be retrieved, evaluated, and explained rather than merely promotional language that can be repeated.

A company can claim reliability. Verified customer experience can provide evidence of reliability. A company can claim superior value. Independent market evidence can provide a reason to believe that proposition. As AI increasingly organizes product comparisons, the difference between assertion and substantiation becomes not merely a communications issue, but a competitive one.

Consumers Want Help, Not Control

AI's growing influence should not be confused with unconditional trust. Fifty-eight percent of AI shopping users in the ICERTIAS study said they check at least one additional source before acting on an AI recommendation. Eighty-three percent said seeing the sources behind a recommendation was important, while 53% considered it important to understand why AI recommended a particular product.

These findings describe a consumer who is willing to delegate work without fully delegating judgment. People may allow AI to search a large market, summarize specifications, compare prices, organize reviews, and reduce complexity. But efficiency does not eliminate skepticism. A recommendation becomes more useful when consumers can understand the evidence behind it and verify the conclusion.

This produces a customer journey that is more compressed but not necessarily less demanding. A consumer may happily allow AI to eliminate 17 products from a field of 20 because doing so saves time. The same consumer may scrutinize the final three more carefully precisely because another system performed the initial elimination.

For companies, this creates two distinct competitive tests. The first is eligibility: does the product have enough relevant, accessible, and understandable information to enter the AI-generated shortlist? The second is credibility: once the product appears there, is there enough trustworthy evidence for the consumer to accept the recommendation rather than reopen the search?

The distinction has significant implications for how companies think about content and evidence. Information that once appeared secondary to advertising may become central to consideration. Specifications, warranties, service terms, verified reviews, independent tests, safety information, certifications, and credible third-party recognition can all help explain why a product deserves to survive comparison.

AI Opens Doors for New Brands

One of the most consequential ICERTIAS findings concerns unfamiliar brands. Sixty percent of AI shopping users said AI had introduced them to a brand they had not previously considered seriously. Thirty-three percent said they had actually purchased a previously unconsidered brand after AI brought it into their consideration set.

For challenger brands, this creates an important opportunity. Historically, entering the same consideration set as an established competitor often required significant advertising investment, broad distribution, repeated exposure, and time. AI can potentially compress part of that process by introducing an unfamiliar product together with an explanation of its price, specifications, reviews, features, and fit.

A brand that lacks decades of accumulated awareness may therefore gain consideration if the available evidence gives AI a compelling reason to recommend it. That does not eliminate the advantages of familiarity, reputation, or distribution. It does mean those advantages may no longer be sufficient to keep lesser-known competitors outside the customer's field of view.

For incumbent brands, the same mechanism creates a corresponding risk. Awareness remains valuable, but awareness no longer guarantees consideration. A consumer can know a brand extremely well and still receive a shortlist composed entirely of competitors whose available evidence appears to match the request more closely.

The ICERTIAS study also found evidence of AI functioning as an elimination mechanism. Twenty-three percent of AI shopping users said an AI system had advised them against buying a brand they had previously been seriously considering.

Among consumers who reported such an experience, 45% cited better value from a competitor, 39% cited quality or reliability concerns, 34% pointed to weaker specifications, 31% cited negative reviews, and 30% cited excessive price. Respondents could identify more than one reason, so these percentages should not be interpreted as mutually exclusive categories.

What matters is what those reasons have in common. Most can be compared, documented, and explained. AI does not need to develop a negative opinion of a brand. It only needs enough evidence to conclude that another product offers better value, stronger reliability, more appropriate specifications, better customer feedback, or a closer fit with the stated need.

Companies have always invested heavily in understanding why customers choose them. The emerging environment requires a second question that may prove equally important: why might an AI system remove our brand before the customer seriously evaluates it?

AI Could Make Brand Loyalty More Conditional

Price reveals the same dynamic. ICERTIAS asked respondents to imagine that their preferred brand cost 10% more than an alternative and that AI concluded the competitor offered better value for money. Only 24% said they would simply remain with their preferred brand. Fifty-one percent would seriously consider the competitor, while 25% said they would switch.

The finding does not suggest that brand loyalty is disappearing. It suggests that loyalty becomes easier to challenge when consumers receive a clear explanation of what they are sacrificing to remain loyal. AI can make differences in price, specifications, performance, reviews, and long-term value easier to compare than they were in a fragmented information environment.

That does not necessarily favor the cheapest product. Thirty-two percent of AI shopping users said AI had at some point led them to spend less than they originally planned. Nineteen percent said AI had persuaded them to pay more because a more expensive product appeared to represent the better long-term choice.

For premium brands, that distinction is critical. AI can expose an unjustified premium, but it can also help defend a justified one. Durability, safety, warranty protection, service quality, energy efficiency, reliability, performance, and total lifetime value can all provide rational explanations for why paying more may be the better decision.

The strategic vulnerability is therefore not premium pricing itself. It is premium pricing that cannot be explained convincingly. A higher price can survive comparison when the reasons behind it are concrete, credible, and visible. A premium supported mainly by familiarity or vague superiority claims may become more vulnerable when AI makes alternatives easier to evaluate.

This changes the discipline of premium positioning. The question is no longer simply whether customers perceive the brand as prestigious or superior. Companies also need to ask whether the underlying reasons for that superiority are sufficiently explicit that an AI system can identify, compare, and communicate them to a customer considering a less expensive alternative.

Assisted Commerce Comes First

The prospect of autonomous AI agents buying products without human involvement attracts attention because it represents the most dramatic possible version of AI commerce. The ICERTIAS findings suggest that the more immediate change is less spectacular but commercially more important: consumers are adopting AI assistance much faster than they are surrendering control over the transaction.

Fifty-five percent of respondents had already used AI during shopping, yet only around one in five expressed willingness to permit some form of autonomous purchasing within predefined rules. The gap is strategically important because it shows that businesses do not need to wait for fully autonomous commerce before AI begins influencing competitive outcomes.

Consumers remain cautious about giving software final authority, but they are considerably more comfortable allowing AI to perform the work that precedes the decision. That includes finding alternatives, comparing specifications, evaluating prices, summarizing evidence, identifying trade-offs, introducing unfamiliar brands, and eliminating products that appear to be weaker choices.

The immediate question is therefore not whether AI will soon buy most products without human approval. It is whether AI is already reorganizing the competitive field before humans make those purchases themselves.

A machine does not need permission to complete the transaction in order to influence its outcome. It needs only enough influence to determine what the customer sees, which alternatives appear credible, which weaknesses become visible, and which products never reach the stage where a human being considers buying them.

The Metric Companies Do Not Yet Have

Most companies can measure brand awareness, consideration, search visibility, website traffic, distribution, conversion, repeat purchase, customer satisfaction, and market share. Few can yet answer with comparable precision a question that may become increasingly important: when a customer asks AI what to buy, does our brand appear?

That question quickly leads to others. For which customer needs does the brand appear? Which competitors appear instead? What evidence supports their inclusion? What causes the brand to be excluded? Can its most important advantages be independently verified? Can the system explain why its price is higher or why its performance is better?

A useful way to think about this emerging metric is shortlist visibility: the probability that a brand appears when an AI system is asked to recommend products for a relevant customer need. Unlike conventional awareness, shortlist visibility does not ask whether consumers recognize a brand. It asks whether that brand survives AI-assisted comparison long enough to receive serious consideration.

The distinction matters because a company can possess high awareness and low shortlist visibility at the same time. Consumers may know the brand, remember its advertising, and recognize its identity, yet an AI system may still favor competitors if the evidence available for the specific buying question points elsewhere.

This does not mean companies should abandon traditional brand building and optimize everything for machines. Consumers still respond to familiarity, emotion, identity, design, reputation, experience, and meaning. Those forces remain powerful because humans still make the final decision in most purchases.

The stronger strategy is to compete in both systems. Build a brand that people remember, trust, and want, while making the rational reasons for choosing it sufficiently accessible and verifiable that AI systems can find them, compare them, understand them, and explain them accurately.

That elevates information once treated as secondary into a strategic asset. Product specifications, verified reviews, warranties, independent tests, service standards, availability, pricing logic, safety information, certifications, customer-experience evidence, and credible third-party recognition increasingly influence not only what consumers know about a product, but whether the product enters the decision at all.

Adoption Will Move at Different Speeds

The ICERTIAS data indicates that AI-assisted shopping is not developing at the same speed across the population. Younger adults reported greater use of AI during shopping than older respondents, suggesting that the transition toward AI-mediated consideration will initially be more pronounced in categories where younger consumers represent a larger share of demand.

Income also influences AI-assisted shopping, although the findings do not suggest that AI shopping belongs primarily to affluent consumers. Higher-income consumers may use AI to evaluate complex products or premium alternatives, while consumers with tighter budgets have an equally strong incentive to use it to compare prices and maximize value.

For companies, the strategic implication is that AI should not be treated as a single channel with a uniform adoption curve. Its importance will differ by category, geography, age profile, income distribution, purchase complexity, and the degree to which consumers already rely on digital information before making a decision.

Because the ICERTIAS study spans 57 very different markets, detailed demographic subgroup estimates should be interpreted within the study's sampling and weighting framework. The more important strategic conclusion is directional: AI-assisted shopping is broad enough to matter today, while its commercial impact will spread unevenly across consumer groups and categories.

Competition Is Moving Earlier in the Buying Journey

The most important conclusion from the ICERTIAS study is not that AI is replacing consumers. It is that a meaningful part of competition is moving earlier in the buying journey, into a stage where companies historically had less visibility and where an intermediary can increasingly influence who receives serious consideration.

Consumers still want control. They continue to verify recommendations and remain cautious about fully autonomous purchasing. Yet they are increasingly willing to delegate much of the difficult work that precedes choice: searching, comparing, summarizing evidence, identifying trade-offs, evaluating prices, introducing alternatives, and eliminating weaker options.

The change is therefore better understood as a movement from human-only consideration toward AI-mediated consideration, not as a transfer from human choice to machine choice. That distinction matters because AI can redistribute commercial opportunity long before autonomous purchasing becomes widespread.

A system does not need authority to press the buy button to change market outcomes. It only needs sufficient influence over which three products the consumer examines after starting with 20, which evidence the consumer sees, and which competitor never reaches the final comparison.

That brings the argument back to three findings at the center of the study. Sixty-one percent preferred three AI-selected products to reviewing 20 alternatives. Forty-nine percent of AI shopping users were willing to buy without first visiting the manufacturer's website. And external evidence substantially strengthened purchase likelihood for an unfamiliar AI-recommended brand compared with the brand's own unsupported value claim.

Together, those findings describe an emerging competitive sequence. A brand must first qualify for the AI-generated consideration set. It must then survive comparison against alternatives. Finally, it must provide enough credible evidence for a human being to trust the recommendation.

For business leaders, this adds a question upstream of traditional marketing priorities. Companies have long asked how to create demand, win attention, strengthen preference, defend pricing, and convert customers. They increasingly need to ask something that comes before all five: what gives an intelligent intermediary sufficient reason to recommend us?

The title of this study asks a deceptively simple question: Who chooses the brand now?

The consumer still makes the final choice.

Increasingly, however, AI helps determine what the consumer gets to choose from.

That may prove more consequential to the future of consumer competition than whether an AI agent ever presses the buy button.

Methodology

The ICERTIAS 2026 Study on AI and Consumer Purchase Decisions was conducted in July and August 2026 among 8,000 adults aged 18 and over with internet access across all 57 OSCE participating States. Data were collected using CAWI through online research panels. Quotas were applied by country, age, gender, and region, with additional controls where panel availability permitted. Results were weighted to reflect the demographic structure of the adult online population across the surveyed markets. Percentages refer to the relevant respondent base for each question unless otherwise stated. Country-level sample sizes were not designed to provide equally precise standalone national estimates. Findings should therefore be interpreted primarily at the pan-OSCE level.