The User Agency Trap in Media Literacy Chatbots: Interactive Evaluation Format, Imperfection Disclosure, and Visual Misinformation Discernment

Although media literacy interventions are increasingly used to counter misinformation, many remain text-centered and non-interactive, limiting engagement during real-time evaluation. This limitation is particularly consequential for visual misinformation, where realistic synthetic images can encourage intuitive judgments. This study examines how two chatbot design features—interactive evaluation format and chatbot (im)perfection disclosure—shape perceived user and AI agency and, in turn, truth discernment. In an online experiment (N = 348), participants evaluated visual misinformation in a chatbot-supported task varying these two features. Interactive evaluation increased perceived user agency but decreased perceived AI agency, whereas imperfection disclosure modestly increased user agency without affecting AI agency. Critically, interactive evaluation indirectly reduced truth discernment through increased perceived user agency. These findings reveal a potential user agency trap: chatbot interactivity can make users feel more empowered while paradoxically undermining their ability to discern visual misinformation.