<![CDATA[ABSTRACT Tuberculosis requires long-term treatment adherence, yet many patients experience low adherence and self-efficacy, which can lead to treatment failure. The use of technology-based interventions such as chatbot AI offers potential support in improving patient outcomes.This study aimed to analyze the effect of HeriChat-TB web-mobile AI on medication adherence and self-efficacy among tuberculosis patients.This study used a quantitative quasi-experimental design with a pretest–posttest with control group approach. A total of 147 respondents were selected using simple random sampling and divided into three groups: intervention 10 days (n=49), intervention 20 days (n=49), and control (n=49). Data were collected using MMAS-8 and TBSES instruments and analyzed using Wilcoxon, Kruskal-Wallis, and Mann-Whitney tests. The 20-day intervention significantly improved medication adherence and self-efficacy compared to the 10-day intervention and control group. The 10-day intervention improved self-efficacy but did not significantly affect adherence. No significant change in adherence was found in the control group. The correlation between changes in self-efficacy and adherence was positive but not significant. HeriChat-TB web-mobile AI is effective in improving medication adherence and self-efficacy in tuberculosis patients, particularly with longer intervention duration. Keywords: Tuberculosis, Adherence, Self-Efficacy, Chatbot AI, Digital Health. ABSTRAK Tuberkulosis memerlukan kepatuhan pengobatan jangka panjang, namun masih banyak pasien yang mengalami rendahnya kepatuhan dan self-efficacy sehingga berisiko terhadap kegagalan terapi. Pemanfaatan teknologi berbasis artificial intelligence seperti chatbot menjadi salah satu alternatif untuk mendukung peningkatan hasil pengobatan.Penelitian ini bertujuan untuk menganalisis pengaruh HeriChat-TB berbasis web-mobile AI terhadap kepatuhan minum obat dan self-efficacy pasien tuberkulosis. Penelitian ini menggunakan desain kuantitatif quasi-experimental dengan pendekatan pretest–posttest with control group. Sampel berjumlah 147 responden yang dipilih menggunakan simple random sampling dan dibagi menjadi tiga kelompok, yaitu intervensi 10 hari (n=49), intervensi 20 hari (n=49), dan kontrol (n=49). Data dikumpulkan menggunakan instrumen MMAS-8 dan TBSES, kemudian dianalisis menggunakan uji Wilcoxon, Kruskal-Wallis, dan Mann-Whitney. Intervensi selama 20 hari terbukti meningkatkan kepatuhan minum obat dan self-efficacy secara signifikan dibandingkan kelompok 10 hari dan kontrol. Intervensi 10 hari meningkatkan self-efficacy tetapi belum memberikan pengaruh signifikan terhadap kepatuhan. Kelompok kontrol tidak menunjukkan perubahan kepatuhan yang bermakna. Hubungan antara perubahan self-efficacy dan kepatuhan bersifat positif namun tidak signifikan. HeriChat-TB berbasis web-mobile AI efektif dalam meningkatkan kepatuhan minum obat dan self-efficacy pasien tuberkulosis, terutama dengan durasi intervensi yang lebih lama. Kata Kunci: Tuberkulosis, Kepatuhan, Self-Efficacy, Chatbot AI, Kesehatan Digital.]]>
