Teaching People to Spot Deepfakes: Investigating User Acceptance of LLM-Powered Chatbots

Deepfakes are an emerging form of disinformation that threatens the integrity of digital media. Research shows that people often struggle to recognize them, with accuracy slightly above chance, demonstrating the need for user education. Our study investigates user acceptance of Deepfake Fighter, a Large Language Model chatbot designed to enhance individuals’ understanding of deepfakes and their detection skills. We conducted an online survey with 196 participants, and the data were analyzed using Partial Least Squares Structural Equation Modeling. The analysis focused on how trust, efficiency and enjoyment shaped perceptions of usefulness, ease of use and satisfaction, which then influenced attitudes and adoption intention. Findings showed that trust was the strongest predictor of perceived usefulness, efficiency improved ease of use, while enjoyment increased satisfaction. Our study introduces a theoretical framework for understanding the acceptance of deepfake educational chatbots and highlights their potential to foster deepfake literacy and guide effective design of learning tools.