Assessing GPT's capabilities in consumer food survey analysis: A comparative approach for understanding food technophobia and novel protein perceptions

This study explores the application of GPT models for automating consumer food survey analysis, focusing on Chinese consumers' acceptance of plant-based foods, cultured meat, insect-based proteins, and microbial proteins. Traditional survey analysis methods face limitations in handling large-scale, open-ended responses, whereas GPT's natural language processing capabilities offer efficient, bias-reduced alternatives. Employing Partial Least Squares Structural Equation Modeling (PLS-SEM), we investigate how food technophobia (FTN) and food values (FV) affect perceived benefits (PB) and perceived risks (PR), ultimately influencing consumer acceptance. Results show that acceptance is highest for plant-based foods and lowest for cultured meat, with PB positively and PR negatively impacting consumer willingness to these foods. Mediation analysis reveals that PR and PB mediate the effects of FTN and FV on acceptance, indicating that attitudes toward food safety, naturalness, and production processes shape consumer choices. The findings underscore the value of using GPT for comprehensive, real-time survey analysis and suggest marketing strategies and policies that emphasize product safety, environmental benefits, and consumer education to enhance acceptance of alternative proteins.