Decision gatekeepers: how hedonic vs. utilitarian products shape differential persuasion effects in LLM-mediated recommendations

As consumers increasingly rely on AI for purchase guidance, large language models (LLMs) have become key mediators of buying decisions, processing marketing content before generating recommendations. Understanding how these AI-based decision support systems respond to persuasive techniques becomes critical for consumers and system designers. This study maps LLM response patterns to persuasive techniques across product types. Using a controlled experiment with 13,500 observations across three LLMs (GPT-4.1 Mini, GPT-5 Mini, Kimi K2-0905), the research tested hedonic versus utilitarian products under four persuasion conditions based on Cialdini’s principles. Results reveal differential persuasion sensitivity by product type: social proof increased the rate at which LLMs expressed high recommendation certainty for concert tickets from 9.6% to 65.6%, while authority appeals reduced certainty for utilitarian products. This asymmetric pattern persisted across all models and partially mirrors human consumer psychology, suggesting LLMs internalized decision frameworks from training data. The study contributes to IS research by extending the hedonic-utilitarian framework to AI-mediated decision support and connecting LLM persuasion susceptibility to debates on technology trust. These findings have practical implications for how persuasive content shapes AI-mediated consumer decision support.