ISCAP Proceedings: Abstract Presentation
Suspicion Follows Detectability, Not Harm: Consumer Response to Generative AI in Food Delivery Communities
Minjae Kim
Georgia State University
Frank Lee
Georgia State University
Abstract
Generative artificial intelligence has drastically reduced the cost of producing photorealistic product imagery, spawning a new vendor industry that supplies AI-generated food photographs to restaurants on delivery platforms. Theoretically, if a listing photograph serves as a signal of actual product quality, the proliferation of deceptive imagery should degrade this signal and prompt consumer backlash. To investigate this, we analyzed a random sample of 321,811 Reddit comments from 2019 to 2026 across six online communities. Isolating 143,931 comments within three delivery-platform communities, we found virtually no expressed suspicion: exactly one customer questioned the authenticity of a food photograph. By contrast, applying a comparable procedure to the same corpus revealed a sharply rising trend of customers suspecting and complaining about AI-generated text, beginning in the third quarter of 2023 and escalating through 2026. We argue this asymmetry stems from detectability rather than the absence of harm. While AI-generated text exhibits linguistic markers that readers notice without effort, AI food photography remains largely covert given the brief attention consumers give menu thumbnails. This presents a significant challenge to market-based quality mechanisms, which assume consumers will eventually detect deception and adjust their purchasing behavior. Our findings suggest that highly convincing, low-detectability deceptions may systematically bypass consumer scrutiny, neutralizing market corrections entirely.