<![CDATA[ABSTRACT: The aim of this study was to determine the differences in the effects of counseling carried out entirely by AI and counseling carried out by counselors assisted by AI on the mental health of adolescents. The measuring instruments used to reveal mental health are the General Health Questionnaire (GHQ-12) to measure general health, the GAD-7 Anxiety Scale to measure general anxiety, the Positive and Negative Affect Schedule (PANAS) to reveal positive and negative feelings, and the Mood Scale to reveal participants' moods. The study involved 56 adolescent participants (female, n = 43; male, n = 13) who were randomly assigned to three experimental groups. The procedure included a pre-test, a 30–45 minute counseling session, a post-test, and a concluding interview. Overall, chat counselling decreased negative feelings (p=0.031) and improved mood (p=0.018). AI-only counselling had no significant effect, human-only improved mood (p=0.018), while human-AI collaborative decreased perception of general mental health (p=0.038). AI-only counselling had a significant effect on anxiety level (p=0.047) and human-only had a significant effect on mood (p=0.004) compared to human-AI collaborative. Significant differences were identified only between the human-AI collaborative and human-only groups (p=0.033), with the human-only group achieving higher mood scores. Given the distinct capabilities of AI and human counselors, these results suggest the potential for developing hybrid models and Digital Therapeutic Alliance (DTA) frameworks to support existing counseling services. Keywords: adolescents, artificial intelligence, counselling, mental health. ABSTRAK: Tujuan penelitian ini adalah mengetahui perbedaan pengaruh konseling yang sepenuhnya dilakukan AI dan konseling yang dilakukan konselor dibantu oleh AI terhadap kesehatan mental remaja. Alat ukur yang digunakan untuk mengungkap kesehatan mental adalah General Health Questionnaire (GHQ-12) untuk mengukur kesehatan umum, GAD-7 Anxiety Scale untuk mengukur kecemasan umum, Positive and Negative Affect Schedule (PANAS) untuk mengungkap perasaan positif dan negatif, dan Skala Suasana Hati untuk mengungkap suasana hati partisipan. Terdapat 56 partisipan (perempuan, n = 43, laki-laki, n = 13) remaja yang dibagi secara acak ke dalam tiga kelompok eksperimen. Prosedur meliputi prates, sesi konseling selama 30-45 menit, pascates, dan wawancara. Secara keseluruhan, konseling via chat menurunkan perasaan negatif (p=0.031) dan meningkatkan suasana hati (p=0.018). Konseling AI-only tidak memberikan pengaruh signifikan, human-only dapat meningkatkan suasana hati (p=0.018), sedangkan human-AI collaborative menurunkan persepsi atas kesehatan mental umum (p=0.038). Konseling AI-only berpengaruh signifikan terhadap tingkat kecemasan (p=0.047) dan human-only berpengaruh signifikan terhadap suasana hati (p=0.004) dibandingkan human-AI collaborative. Perbedaan signifikan hanya ditemukan antara kelompok human-AI collaborative dan human-only (p=0.033), dengan kelompok human-only memiliki skor suasana hati yang lebih baik. Berdasarkan potensi dari konseling dengan AI dan konselor, dapat dilakukan pengembangan model konseling hibrida dan konsep DTA untuk saling melengkapi layanan kesehatan mental, khususnya konseling. Kata kunci: remaja, artificial intelligence, konseling, kesehatan mental.]]>
