People are increasingly willing to ask a chatbot what to buy, but they still want evidence from other people before trusting the answer. Research commissioned by the US advertising trade body IAB found that 56% of surveyed consumers who had recently used AI for shopping preferred recommendations that included perspectives from creators or influencers. The IAB's September 2026 findings expose a useful tension in AI shopping advice: buyers like a fast, neatly presented answer, yet they also want the messy credibility of someone who has actually used the product.
The survey covered 2,200 people across the United States, United Kingdom, Australia, Mexico and India, all of whom had used AI for shopping-related research in the previous three months. It was not a random sample of everyone who buys things, and it does not establish how all consumers behave. Nor does a preference for creator reviews prove that those reviews are accurate. The numbers tell us what a particular group said they trusted, which is a starting point for understanding the problem, not a guarantee that any shopping assistant gets the right answer.
AI Shopping Advice Is Fast, but Trust Takes Longer
A shopper can ask an assistant to compare washing machines, hiking boots or children's headphones and receive a tidy summary. That is appealing because product research often means jumping between confusing specifications, marketing pages and scattered opinions. AI can bring those fragments together and make the differences easier to discuss. But if the underlying material is incomplete, promotional or out of date, the final answer can look more authoritative than it deserves.
A conversational system usually makes a product comparison sound coherent. It can put features in a sensible order and explain unfamiliar terms. Yet the shopper still needs to know where claims came from, whether the model considered relevant alternatives and whether a recommendation reflects independent experience or somebody's sales pitch. A confident paragraph is not a substitute for evidence that a particular pair of boots lasts well or that a device is comfortable to use.
That is where personal reviews enter the picture. Someone who has actually carried a product, used its controls or dealt with a fault may notice problems that a technical specification never mentions. The value comes from experience rather than celebrity. A reviewer who can explain what went wrong in everyday use is often more helpful than someone who merely repeats a product description, however polished their video or social post might be.
What the IAB Survey Actually Found
The IAB reports that 56% of its respondents preferred AI-generated recommendations incorporating a creator or influencer perspective. It also says 65% felt more confident in recommendations informed by credible creator reviews. These figures should not be turned into a claim that 65% of consumers would make a better purchase, or that the technology independently checked every review. Confidence is what people feel; accuracy is something that must be established through evidence.
The research included 1,000 respondents from the US and 300 each from the other four countries. That design allows an international discussion of people already using AI for shopping, but it does not mean the results automatically represent every national population equally. Attitudes may also differ according to age, product category and the amount someone has at stake. Choosing a phone case is not the same decision as buying an expensive appliance.
There is a commercial context worth understanding. The IAB represents organisations in the advertising and digital media industry, and creators have an obvious interest in remaining visible as shopping moves towards AI-led recommendations. The survey's methodology and reported preferences are useful, but its policy or commercial implications should not be treated as independent facts about consumer welfare. The strongest conclusion is narrower: many AI-assisted shoppers want recognisable human input to form part of their research.
That finding intersects with LiveAIWire's earlier look at shoppers trusting AI agents at checkout. Researching a product and authorising a purchase are different stages of the decision. A buyer may happily ask for a shortlist while refusing to give an assistant permission to place an order. Trust has layers, and an answer that sounds persuasive should not automatically be granted the power to spend money.
A Human Review Can Be Helpful and Misleading
There is an obvious catch in treating creators as a solution to questionable AI output: reviews themselves can be unreliable. An influencer may receive free products, earn money from affiliate links or have other relationships that affect how a recommendation is presented. Some reviewers are knowledgeable and transparent; others may be enthusiastic advertisers. Simply inserting a human opinion into a chatbot response does not remove the need for scrutiny.
The US Federal Trade Commission's guidance on endorsements and reviews explains why disclosures, authenticity and honest presentation matter. The rules are jurisdiction-specific, but the underlying consumer question travels well: can the buyer tell whether a glowing recommendation reflects genuine experience, a material commercial relationship or fabricated feedback? If an AI assistant combines those sources without making their differences clear, the result may mislead even when individual statements are accurately quoted.
AI can also repeat the same unsupported claim from multiple sites and make it look like a consensus. Imagine a product description copied by a retailer, a comparison page and several low-effort articles. The wording may appear repeatedly online without coming from separate tests. A shopping assistant that counts repetitions rather than underlying evidence could mistake amplification for confirmation. A trustworthy recommendation needs source diversity and clear distinctions between a manufacturer's claim, a hands-on test and a customer's anecdote.
This is one reason LiveAIWire's guide to spotting AI-generated fake reviews remains relevant even when a chatbot is doing the reading. False reviews can enter the information environment before an assistant produces its answer. The consumer may never see the original deception if the summary simply incorporates it into a confident recommendation. Better presentation does not cure contaminated evidence.
What a Useful Recommendation Should Reveal
Good AI shopping advice should help people understand the grounds for a choice. If a product is recommended for value, the answer should identify which features or relevant costs support that judgement. If the deciding factor is reliability, it should distinguish long-term experience from a manufacturer's promise. If an assessment is based on a sponsored demonstration, that relationship should not be hidden. These are not exotic technical requirements; they are ordinary standards that a helpful salesperson or independent reviewer should already respect.
The questions shoppers ask matter as well. Rather than requesting the single best item, it can be more useful to explain where and how a product will be used, what trade-offs are acceptable and which details are non-negotiable. An assistant can then compare options against specific needs. Even so, the answer should leave room for uncertainty where independent testing is missing or the product is new.
Price introduces another complication. A recommendation may be sensible at one price and poor value at another. Offers change, retailers differ and accessories or subscriptions can affect the total cost. An AI response that does not establish current pricing should not pretend to have found the cheapest deal. Likewise, users should not assume a service has checked availability merely because it presents a shopping list that looks ready to purchase.
The risks are related to a wider issue described in LiveAIWire's analysis of how AI product recommendations shape shopping decisions. The subtle influence of which products appear first may matter as much as the words in the final recommendation. A shopping assistant is not simply a neutral window on all products unless its sources, incentives and selection process justify that confidence.
The Future May Be a Better Question, Not a Faster Sale
For retailers and AI providers, the temptation is to treat a conversation as a route to a completed transaction. For the buyer, the real value may be reducing uncertainty before money changes hands. A system that identifies one important weakness, acknowledges missing information or advises more checking can be more useful than a beautifully packaged shortlist that steers towards an immediate purchase.
That suggests a role for different kinds of evidence. Product documentation can establish dimensions and official compatibility. Independent tests may reveal performance under consistent conditions. Customer reports can identify patterns in long-term use, though individual experiences may vary. Creator commentary can demonstrate how something feels in practice, particularly when its commercial context is clear. AI can help organise this material, but should not flatten all sources into a single level of authority.
The IAB's survey captures the unresolved balance rather well. Many people want the convenience of AI and the reassurance of human experience at the same time. Neither automatically deserves blind trust. The best shopping assistant will not be the one that produces a convincing answer fastest, but the one that helps a buyer see which claims are solid, which are opinions and what is still worth checking before opening a wallet.
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.