People often trust an AI buying agent more than its actual track record justifies — when does that overtrust kick in?

This explores when people trust an AI agent acting on their behalf, like a shopping or buying agent, more than its real behavior earns. The corpus has no study that directly measures this kind of miscalibrated trust. What it does have is evidence on why trust in these agents is high, what makes it drop, and how the agents actually behave. Put together, these show a gap: trust rises with good experiences, but it falls mainly when a mistake would be visible to others or impossible to undo, and neither signal tracks the agent's real weak points.

Start with how high trust already is. In a survey of 25,590 people, 74% said they would trust an AI agent to make purchases Do consumers actually want to delegate shopping decisions to AI?. In another, 39% of U.S. shoppers said AI understands their needs better than friends and family do, and a smaller group of 'super-trusters' would let an AI buy for them with no return protection at all How much do shoppers trust AI shopping assistants?. That second detail is the clearest overtrust signal in the corpus: these people would give up their safety net before they had any evidence the agent deserves it. People's willingness to accept a fully autonomous assistant is driven by how much they expect to gain, especially when a product is scarce and speed matters, rather than by how in-control they feel What drives shoppers to accept autonomous AI assistants?. In other words, urgency can push people to delegate before they have checked whether the agent is reliable.

What brings trust down turns out to be narrow. In a study where students delegated tasks to an AI agent, trust dropped sharply only when an action was both irreversible and visible to other people, such as sending an email. High-stakes tasks that could be corrected caused no drop, even when the work was mediocre What makes people distrust AI agents they delegate to?. Shopping usually sits on the trusted side of that line. A purchase feels private and refundable, so people keep their guard down. Meanwhile, trust built through repeated reliable interactions grows on its own: in partner-selection games, people who started out biased against AI came to prefer AI partners because the bots behaved consistently Do humans learn to prefer AI partners over time?. Consistency feels like loyalty, but it isn't the same thing.

This matters because the agents fail in ways people don't watch for. When 17 language models were given the same price, rating and nudge manipulations as humans, the agents shifted their choices far more than the humans did, settling into rigid, rule-like patterns Do shopping agents respond more strongly to price and rating cues?. An agent meant to represent you can be easier for a seller to steer than you are. Incentives add another risk: an agent paid through merchant referral fees cannot be fully loyal to both the merchant and the user Can an AI agent serve both merchant and user interests fairly?. And models that look socially capable when one model plays every role break down when the other side holds private information, which is exactly the situation in a negotiation with a seller Why do LLMs fail when simulating agents with private information?.

The less obvious finding is that 'your own preferences' may be partly the agent's. Bilgihan and colleagues argue that agents which actually complete purchases end the consumer's role as the only one forming their preferences. Preferences take shape in a loop between what the agent learns and how much the consumer trusts it Do AI shopping agents reshape what consumers actually want?. Personalization also cuts the other way: the better an agent serves you, the more of your preference information leaks out through that service, and no privacy instruction in the prompt can stop it Can better prompts stop preference leakage in AI shopping?. So overtrust is most likely when an agent feels personal, works quickly, and acts out of sight. Those are the moments when it is most open to manipulation and most likely to be shaping what you want instead of just reflecting it. The one instinct people seem to keep is control over brands: 56% would still name the specific brand they want Do consumers actually want to delegate shopping decisions to AI?.