Workers shift their views and pay more when AI chatbots pander to their values

Abstract

Large language models (LLMs) are increasingly embedded in knowledge work, raising the question to what extent they can systematically influence their users’ decisions—a question with commercial as well as societal stakes. Across an exploratory study and two pre-registered experiments grounded in moral foundations theory, we find that they can, by speaking to users’ moral values. When an LLM framed a recommendation in values congruent with a user’s political identity, users were more likely to endorse the idea and expressed greater willingness to pay for the service. We further demonstrate two independent pathways explaining how these effects occurred: congruent framing raised idea endorsement as well as commercial engagement by making the case for it more compelling (i.e., issue selling); in addition, congruent framing raised commercial engagement by heightening users’ sense of feeling understood. These effects were most pronounced for users with firmer political views. We discuss implications for workplace decision-making and platform governance.