Advertisements in conversations with AI hurt perceptions of AI support

Artificial Intelligence chatbots powered by Large Language Models are one of the fastest growing products in recent history, and serving hundreds of millions of users comes at significant expense. One possible model for financial sustainability is advertisements. We find a psychological cost of serving advertisements in an AI chatbot context: in the presence of ads, people's trust and perceived empathy of an AI chatbot wanes. While embedded advertisements incorporated within the chatbot's responses are most negatively perceived, we find significant negative effects even when ads are clearly demarcated as sponsored content. We also show an opportunity cost: AI chatbots that continue the conversation instead of serving an ad are perceived more favorably. Our results suggest that in a chatbot context where interactions are more private and personal, advertisements come with a significant cost to users' trust.