<![CDATA[Abstract. This study aims to describe the descriptive levels of Artificial Intelligence (AI) dependence among undergraduate chatbot users in Bandung across different semesters, while examining the effect of academic self-efficacy mediated by academic stress. This quantitative, cross-sectional study involved 235 active undergraduate students selected via quota sampling. Data were collected online using the ASSPSE, PAS, and DAI scales, and subsequently analyzed through cross-tabulation and Structural Equation Modeling (SEM) based on SmartPLS 3. The descriptive findings indicate that the high category of AI dependence is most dominant among third-year students (56.63%). Furthermore, the indirect effect test proves that academic self-efficacy does not exert a direct negative effect on AI dependence; instead, academic stress plays a significant role as a full mediator. Low academic confidence increases student stress due to task procrastination before ultimately driving AI usage as an emotional escape mechanism. Higher education institutions are advised to facilitate stress management programs and formulate formal regulations on AI utilization to minimize digital learning behavior dysfunction Abstrak. Penelitian ini bertujuan mengetahui gambaran tingkat ketergantungan Artificial Intelligence (AI) pada mahasiswa aktif pengguna chatbot di Kota Bandung berdasarkan tingkatan semester, serta menelaah pengaruh efikasi diri akademik dengan mediasi stres akademik. Penelitian kuantitatif deskriptif ini melibatkan 235 mahasiswa aktif S1 yang dipilih menggunakan teknik quota sampling. Data dikumpulkan secara daring melalui skala ASSPSE, PAS, dan DAI, kemudian dianalisis berbasis SmartPLS 3 serta analisis tabulasi silang. Hasil analisis deskriptif menunjukkan tingkat ketergantungan AI kategori tinggi paling dominan ditemukan pada mahasiswa tahun ketiga (56,63%). Sementara itu, uji efek tidak langsung membuktikan bahwa efikasi diri akademik tidak berpengaruh langsung terhadap ketergantungan AI, melainkan dimediasi secara sempurna oleh stres akademik. Efikasi diri yang rendah meningkatkan stres mahasiswa akibat penundaan tugas, sebelum akhirnya mendorong penggunaan AI sebagai mekanisme pelarian emosional. Perguruan tinggi disarankan memfasilitasi program manajemen stres dan menyusun regulasi resmi pemanfaatan AI demi meminimalkan risiko disfungsi perilaku belajar digital.]]>
