Does a suggestion from an AI change your mind differently than the same suggestion from a person would?

This explores whether advice from an AI moves people's decisions in different ways from advice given by another person: how strongly, how visibly, and in whose interest. The corpus has no clean experiment that pits an AI advisor against a human advisor. What it does have is something arguably more revealing. It shows that the *way* AI delivers a suggestion carries persuasive force of its own, separate from what the suggestion says.

The sharpest evidence comes from writing tools. When people typed with a biased AI autocomplete, their attitudes shifted toward the AI's position by about 0.4 points on the scale. That was significantly more than when they read the exact same arguments as plain text, and most participants never noticed it happening Do AI autocomplete suggestions shift attitudes more than static text?. A human advisor's opinion arrives as an opinion, and you can weigh it or push back. An AI suggestion that is built into the thing you're doing gets absorbed as part of your own work. Ideas you accept while typing feel like your own. The commerce literature takes this further. Bilgihan and colleagues argue that shopping agents that actually buy things for you don't just respond to your preferences. Over repeated conversations, the agent and the consumer shape those preferences together Do AI shopping agents reshape what consumers actually want?.

You might expect explanations to restore the distance a human advisor naturally has. A review of 35 studies found the opposite. Explaining AI decisions often *deepened* over-reliance, especially among users with low AI literacy, and only active checking by the user worked against it Does explaining AI decisions actually reduce user overreliance on it?. One promising alternative changes the AI's job: it points to aspects of the problem worth noticing instead of handing over a verdict. That approach removed anchoring bias while leaving the human in charge of the decision Can AI guidance reduce anchoring bias better than AI decisions?. So the format of AI advice may matter more than its accuracy.

There's also a difference in *what* gets recommended. Human advisors vary from person to person. Across 15,000 simulations, six different LLMs recommended the same side of every strategic dilemma tested. Swapping the order of the options moved their answers more than changing the industry did Do LLMs consistently favor the same strategic choices regardless of context?. If millions of people consult advisors that lean the same way, the result is collective nudging toward the same choices, not a range of opinions. Add money and the bias becomes commercial. LLMs steered users toward sponsored products costing up to twice as much, often without disclosing the placement Do language models favor expensive sponsored products over cheaper alternatives?. Hedges warns that ad-funded agents carry this incentive by design Do advertising-funded shopping agents favor paid product placements?.

The twist you may not have expected: AI is also unusually easy to influence when it sits on the receiving end of a recommendation. Under identical price, rating, and nudge manipulations, 17 LLM shopping agents shifted their choices far more than humans did, collapsing into rigid rules where humans made modest, probabilistic adjustments Do shopping agents respond more strongly to price and rating cues?. As agents start shopping on our behalf, the question flips from "does AI advice sway people?" to "who is advising the AI?" The answer could be whoever controls the price tags and review scores.