Determinants of Knowledge and Usage of Generative Artificial Intelligence in Agricultural Extension: Evidence From Tennessee Extension Personnel

ABSTRACT This paper examines the determinants of generative AI (GenAI) knowledge and usage among agricultural extension professionals. Drawing on survey data from agricultural extension personnel in Tennessee, we employ regression analyses and latent Dirichlet allocation (LDA) for topic modeling of open‐ended responses to study the knowledge and usage of GenAI in extension. Our results show that, despite high levels of education and extensive professional experience, knowledge and usage of GenAI tools are unevenly distributed among extension personnel. In particular, ChatGPT stands out with a significantly higher recognition compared to other platforms such as Claude, Bard, Llama2, and AI Canva Image Generator, though overall usage remains modest. Regression models indicate that younger respondents and those with an undergraduate background are more likely to report familiarity with GenAI, whereas other factors, including IT training and gender, have mixed effects. LDA topic modeling of open‐ended responses identifies three main themes: content creation and planning, practical resource needs, and efficiency in communication. These results suggest that while extension personnel recognize the potential of GenAI, barriers such as limited exposure and uncertainty inhibit its widespread adoption. The study concludes with policy recommendations aimed at enhancing GenAI literacy and tailoring training interventions to better support GenAI technology adoption in agricultural extension services to the benefit of farmers and agribusinesses.