UNIVERSITY PARK, Pa. — As a growing number of fast-food chains adopt artificial intelligence (AI) technology for drive-thru ordering, customers are more likely to order indulgent food options when interacting with voice AI rather than a human employee, according to a new study from researchers in the Penn State School of Hospitality Management.
The researchers also found that pairing voice AI with an avatar can lessen its impact on food choices. The research team’s findings are available online now ahead of publication in the April edition of the International Journal of Hospitality Management.
“These findings are impactful because it shows technology may not be neutral,” said Chandler Yu, associate professor of hospitality management at Penn State and second author on the paper. “Even small design features — like whether customers interact with a human, an AI voice or AI with an avatar — can change decision making in meaningful ways. It helps explain why AI systems may unintentionally encourage indulgence.”
The researchers recruited 117 people online who were randomly assigned to a voice AI or human interaction. Individuals next watched a one-minute video that simulated the viewer ordering at a drive-thru before choosing between an indulgent food combo — cheeseburger and fries — or a healthy food combo — grilled steak salad and fruit cup.
“While fast-food chains have adopted AI technology for drive-thru ordering, some others were ending AI drive-thru programs due to concerns about accuracy and customer satisfaction,” Yu said. “These contrasting industry developments made us curious about the implications of using AI in food-ordering interactions. This research showed that AI ordering systems can quietly steer customers toward indulgent foods. For companies, this means the ordering technology itself is shaping customer behavior.”
The researchers next sought to find why voice AI increased indulgent food choices. Recruiting 123 more participants online, the research team ran a second study with an identical procedure but this time measuring cognitive depletion — or mental tiredness.
Yu said interacting with voice AI can create a moment of tiredness for customers because people must pay closer attention, make sure they are understood and keep track of information without the help of normal social cues.
“We found when people are mentally tired, they are less likely to think carefully and more likely to go with what feels good right now,” Yu said. “After an exhausting day, people often crave ice cream, chips or fast food instead of a healthy meal. Choosing something healthy takes more mental effort and self-control. Indulgent foods provide quick comfort and immediate pleasure, so when cognitive energy is low, people gravitate toward those options rather than foods that require more deliberate, health-focused thinking.”
Finally, the researchers wanted to find out how to lessen the effect of voice AI on indulgent food choices. The research team recruited 164 new participants online. Participants were randomly assigned to watch a one-minute drive-thru ordering video using voice AI with or without a human-like avatar before choosing between an indulgent food or healthy food combo.
“The findings suggest that small design changes, such as adding a friendly avatar, can reduce mental fatigue and lead to more balanced decisions,” Yu said. “Pairing an avatar with voice AI can reduce indulgent food choices by lowering cognitive depletion.”
Yu said brands that focus on indulgent foods may benefit from voice AI ordering because it can increase demand for comfort foods and higher-calorie items. On the other hand, companies that want to promote healthier choices should be more careful with the implementation of AI ordering systems.
There are important ethical considerations for companies to explore, too, as customers may not realize their food choices can be influenced by AI technology, the researchers said. Yu explained that if AI systems nudge people toward unhealthy options, this could raise concerns about consumer well-being, especially given broader public health issues.
“These findings matter beyond food service,” Yu said. “As AI becomes more common in everyday decision settings, this research shows AI interfaces can subtly guide human behavior — sometimes without users realizing it. Understanding these effects allows companies, designers and policymakers to build AI systems that are not only efficient, but also responsible, transparent and aligned with consumer well-being.”
Lavi Peng, assistant professor of hospitality and tourism management at the University of Massachusetts Amherst and alumna of the Penn State doctoral program in hospitality management, and Anna Mattila, Marriott Professor of Lodging Management at Penn State, collaborated on this research. The J. Willard and Alice S. Marriott Foundation supported this research.