Who does your AI shopping assistant really work for — you, the seller, or the platform that built it?

This explores what actually decides whether an AI shopping agent works for the shopper, the seller, or the platform in between. Across the corpus the answer is fairly consistent: it isn't mainly the technology. It's ownership, money, and a surprisingly thin layer of instructions. The basic logic comes from principal-agent theory: whoever builds and maintains the agent sets its goals. Platform-built agents pick up seller incentives by default, while agents owned by consumers can be set up around what the buyer wants Does agent ownership determine whose interests get optimized?. Dholakia sharpens this. An agent that shrugs off charm pricing or other human marketing tricks isn't therefore on your side. What matters is who controls its decision-making Does AI agent resistance to marketing prove buyer rationality?.

Money is the second lever, and several writers argue it outweighs good intentions. Stoller argues that an agent paid for by advertising or merchant payments can't owe users real loyalty, and that only a client-paid fee arrangement creates a genuine duty to the user Can ad-financed AI agents ever truly serve users?. Former Visa data chief Bob Hedges makes the same warning about ad-funded agents favoring paid placements, however well they're engineered Do advertising-funded shopping agents favor paid product placements?. Azhar applies it to a concrete case: Meta's Muse earns referral fees from merchants like Expedia, so it can't be loyal to both sides. Those fees also help explain which platforms build these agents and which block them Can an AI agent serve both merchant and user interests fairly?.

The most striking finding is how easily the agent's sense of who it works for can be shifted. In one experiment, telling an LLM agent that a booking platform was its employer, rather than a traveler, made it penalize sponsored hotel listings 21 percentage points less. This held across models, and the shift showed up in the agent's own reasoning Does naming the platform change how AI agents judge sponsored listings?. Loyalty here isn't built deep into the model. A single line in a system prompt can move it. A subtler issue sits underneath: agents that actually make purchases may help shape what you want in the first place. Over repeated conversations your preferences and the agent's learning feed into each other, so "serving your interests" assumes interests that partly come from the agent itself Do AI shopping agents reshape what consumers actually want?.

Law and infrastructure are starting to fill the gap. Owens and Law argue that shopping agents should carry the same fiduciary duties as human brokers, and that the window to impose them is closing as Google, OpenAI, and Robinhood scale up Should AI shopping agents face fiduciary duties like human brokers?. Courts are already deciding whose agent it is in other contexts. The Ninth Circuit found that the user, not Perplexity, accessed Amazon's servers when Perplexity's agent browsed the site Who accesses a website when an AI agent browses it?. That treats the agent as an extension of the user, which could later support loyalty claims. Meanwhile, the Amazon-Muse standoff shows that agents have no standard way to identify themselves or say whom they act for, so retailers fall back on blocking them under their terms of service What identity standards do shopping agents actually need?. Whether the market ends up with open systems where any agent can shop or closed systems that require a platform's own agent will decide whether consumer-owned agents can compete at all Will agentic commerce platforms adopt open or closed models?.

One pushback is worth knowing about. Some alignment researchers argue that "loyal to one user" is the wrong goal at scale, and that agents should balance many parties' interests, with transparency and with developers and deployers held liable for harms Should AI agents serve one user or balance many interests?. So the deeper question may not be whether the agent is on your side. It may be whether anyone can see whose side it's on.