Tommy Hastomo, Utami Widiati, Francisca Maria Ivone, Evynurul Laily Zen
The integration of Artificial Intelligence (AI) into English as a Foreign Language (EFL) instruction presents new opportunities and challenges. While research often focuses on student outcomes, teacher self-efficacy with these tools, particularly across different educational levels, remains underexplored. Addressing this gap, this study investigated the perceived AI self-efficacy of Indonesian EFL teachers at the primary, secondary, and tertiary levels using a sequential explanatory mixed-methods design. Initially, a quantitative survey (N = 150) using the AI Tool Teacher Self-Efficacy Scale (AIT-SES) was administered. Analysis revealed a distinct hierarchy, with tertiary-level teachers reporting the highest self-efficacy, followed by secondary and then primary teachers. Subsequent semi-structured interviews with a purposively selected sub-sample explored the reasons behind these differences, which are linked to factors like professional autonomy and divergent role demands. Qualitative findings suggest that teachers perceive self-efficacy as both a catalyst for pedagogical innovation and a tool for administrative efficiency. A mature form of efficacy also emerged, striking a balance between confidence and a critical awareness of AI’s risks. Key influencing factors identified were direct mastery experiences, the quality of professional development, and Indonesia's national "Merdeka Belajar" policy, which serves as a significant top-down driver for technology adoption. The study concludes that fostering teacher AI self-efficacy requires more than just providing technology; it demands differentiated, practical training and professional environments that support teacher autonomy and innovation.