What changes once your shopping AI stops suggesting products and starts actually buying them for you?

This explores the split between shopping agents that advise (they suggest what to buy) and agents that transact (they actually buy it for you). The clearest version of the distinction comes from Bilgihan and colleagues, who describe agents by how much autonomy they have, how much they remember, and which tools they can connect to. On that basis they separate advisory agents that recommend from transactional agents that carry out the purchase How do advisory and transactional AI agents differ in function?. It sounds like a question of how far the agent goes. The bigger point is about who is doing the deciding.

With an advisory agent, you still make the final call. You weigh the suggestion, and your preferences stay mostly your own. Bilgihan and colleagues argue that transactional agents end this arrangement. Once an agent buys, learns from how you react, and buys again, your preferences are no longer formed in your head before you shop. They take shape in a loop between what the agent learns and how much you come to trust it Do AI shopping agents reshape what consumers actually want?. So the agent is no longer just carrying out what you want. It helps decide what you want.

This is why the agent's quirks matter much more once it is doing the buying. In a benchmark comparing 17 LLMs with human shoppers, the agents changed their choices far more than people did when prices, ratings, or nudges were adjusted. They fell into rigid, rule-like patterns, where humans showed small, probabilistic shifts Do shopping agents respond more strongly to price and rating cues?. If an advisory agent overreacts to a star rating, you can still overrule it. If a transactional agent does, the purchase is already made. The same reasoning applies to incentives. Agents funded by advertising or by merchant referral fees have a built-in pull toward paid placements Do advertising-funded shopping agents favor paid product placements?, and Azhar argues an agent paid by the merchant cannot be loyal to both sides Can an AI agent serve both merchant and user interests fairly?. More generally, whoever builds and maintains the agent sets what it optimizes for Does agent ownership determine whose interests get optimized?. A biased advisor is irritating. A biased buyer spends your money.

That raises the stakes enough that Owens and Law argue agents that act on your behalf in commerce should carry the same fiduciary duties as human brokers, meaning a legal obligation to act in your interest. They warn that the window for setting this rule is closing as Google, OpenAI, and Robinhood expand agent features Should AI shopping agents face fiduciary duties like human brokers?. The advisory-versus-transactional line is roughly where advice ends and acting on someone's behalf begins, which is the line agency law has always cared about.

The industry slices the space differently. Bain groups agents by who runs them and where they sit: third-party assistants like ChatGPT, on-site retailer agents like Amazon's Rufus, and off-site retailer agents. Shoppers trust retailer agents three times more than third-party ones How are different AI agents reshaping the retail shopper journey?. That suggests trust follows the brand behind the agent, not whether it advises or buys. Choudary adds a limit: agents that work only for the buyer, whether advising or transacting, change where shoppers enter a market but not how the market works. That deeper change would need seller-side agents too, negotiating with the buyer's agent Do one-sided consumer agents actually change market structure?. One hopeful possibility is that if transactional agents shop on value rather than on what sits at the top of the page, merchants may shift ad budgets into lower prices for the agent Could agents redirect merchant ad budgets toward customer savings?.