A new study from Penn State’s School of Hospitality Management reveals that customers are more likely to choose indulgent fast food options when ordering through voice-based AI systems compared to interacting with human employees. The research, set to be published in the April issue of the International Journal of Hospitality Management, highlights how subtle design choices in AI interfaces can significantly influence consumer behavior. The study involved 117 participants who were randomly assigned to either a voice AI or human interaction scenario. After watching a simulated drive-thru video, participants chose between a high-calorie combo—cheeseburger and fries—or a healthier option, such as a grilled steak salad and fruit cup. Results showed that those interacting with voice AI were significantly more likely to select the indulgent meal. The researchers then explored why this shift occurs. In a follow-up study with 123 participants, they measured cognitive depletion—the mental fatigue that comes from exerting extra effort to communicate clearly with an AI system. Unlike human interactions, which include natural social cues like facial expressions and tone, voice AI requires users to focus more intensely on articulating their orders, ensuring clarity, and remembering details. This increased cognitive load leads to mental tiredness, which in turn reduces self-control and makes people more likely to choose immediate gratification over healthier, more deliberate options. To test whether design improvements could mitigate this effect, a third study involved 164 participants who interacted with voice AI either with or without a human-like avatar. The results showed that adding a friendly avatar reduced cognitive depletion and led to a greater likelihood of choosing healthy meals. The avatar provided visual cues and a sense of connection, helping ease the mental burden of the interaction. Chandler Yu, associate professor of hospitality management at Penn State and co-author of the study, emphasized that AI is not neutral. “Even small design features—like whether a customer interacts with a human, a voice AI, or an AI with an avatar—can change decision-making in meaningful ways,” he said. “This helps explain how AI systems may unintentionally encourage indulgent choices.” The findings have important implications for businesses. Fast-food chains that prioritize sales of comfort foods may benefit from using voice AI, as it can subtly increase demand for high-calorie items. However, companies aiming to promote healthier eating should be cautious about how they implement AI systems and consider incorporating visual elements like avatars to support better decision-making. Ethical concerns also arise, as customers may not realize their choices are being influenced by technology. With rising public health concerns around diet-related illnesses, the study underscores the need for responsible AI design that considers consumer well-being. Yu noted that these insights extend beyond the food service industry. “As AI becomes more embedded in everyday decisions—from shopping to healthcare—understanding how these systems shape behavior is critical,” he said. “Designers and policymakers must ensure AI is not only efficient but also transparent, fair, and aligned with human values.”