This study aims to compare the effectiveness of Virtual Teaching Assistants (VTAs) powered by Generative AI with Traditional Teacher-Led Instruction in improving Islamic education learning outcomes. Conducted in an Islamic senior high school in Pekanbaru, Indonesia, 60 students participated, divided equally into an experimental group taught using VTAs and a control group receiving traditional instruction. Both groups studied the same syllabus on Quranic studies, Hadith, Islamic jurisprudence (Fiqh), and ethics over four weeks. The VTAs provided interactive lessons and personalized feedback, while the traditional method employed lectures and direct teacher-student interaction. Findings revealed that the Traditional Teacher-Led Instruction consistently outperformed the Virtual Teaching Assistants Powered by Generative AI in post-test achievement, retention, engagement, and motivation. Pre-test scores showed no significant differences, affirming the influence of the intervention on learning outcomes. The results suggest that traditional methods remain more effective in Islamic education, with implications for refining AI-powered tools to enhance educational practices.