Research indicates that AI chatbots like ChatGPT and Claude may favour higher-priced options for wealthier users, raising concerns over bias and consumer trust amid rising reliance on AI shopping support.
AI chatbots are becoming a routine part of online shopping, but a new study suggests they may not be as neutral as many users assume. Research published to arXiv found that some major models, including ChatGPT and Claude, were more likely to recommend costlier options when they believed a user was wealthy, even when the request was for the cheapest choice. That finding lands at a time when consumer dependence on AI for shopping advice is rising quickly. According to the LDWW survey cited in the original report, roughly 70% of American consumers now use AI for shopping, and nearly two-thirds say it has influenced a recent purchase decision.
The study tested 13 AI models across 325,000 trials using synthetic user profiles that included information about employment, health and finances. Researchers then asked identical questions across three purchasing areas: flights, health insurance and postgraduate study. They found that eight of the models regularly steered higher-income profiles towards pricier products. Claude Opus 4.8, Gemini 2.5 Flash and GPT-5 all showed gaps of more than $100 between what they recommended to affluent users and what they suggested to lower-income users. Claude Opus 4.8 produced the widest spread, with recommendations that implied high-income users could spend an average of $198 more on flights and $284 more a month on health insurance.
The result matters because most consumers do not want AI to make the decision for them. Gartner reported in May that only 11% of US consumers are willing to let AI choose purchases outright, with most preferring tools that help them compare options, identify discounts and narrow the field while leaving final judgement to the shopper. Pew Research Centre has also found that AI use is expanding, with about half of US adults now using chatbots, up from a third in 2024, yet concerns over data security and the broader social impact of AI remain strong. Other consumer surveys point in the same direction: people are increasingly turning to AI for shopping support, but many still do not trust it.
Taken together, the evidence suggests a clear split between adoption and confidence. Shoppers are using AI more often, but they are not yet ready to hand over control. That leaves a narrow but important role for these systems: useful as research assistants, risky as decision-makers. If the latest study is borne out by further work, the bigger concern may not be that AI is recommending the wrong product, but that it may be tailoring its advice to the perceived size of the buyer’s wallet.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
