<![CDATA[Effective assessment of EFL academic writing in Indonesian universities is still difficult because lecturers have heavy workloads and provide inconsistent feedback. While AI tools like Grammarly, ChatGPT, and Gemini promise to improve efficiency, most research focuses on single platforms or Western contexts. This leaves a significant gap in understanding how different AI systems compare with human assessment across various writing aspects in Indonesia's specific EFL environment. This mixed methods study addresses this gap by comparing lecturer assessments with three AI platforms in five writing areas: grammar, coherence, organization, vocabulary, and mechanics. It also explores stakeholder perceptions. A quantitative analysis of 30 students' essays showed that AI consistently gave higher scores in technical aspects, such as grammar and mechanics (p<0.05), but lower scores in holistic dimensions like coherence and organization. There were strong correlations in grammar (r=0.85) and weak correlations in coherence (r=0.38). Qualitative findings revealed that 70.0% of participants felt lecturer assessments were fairer because of their understanding of cultural context. Although AI showed efficiency, it lacked sensitivity to Indonesian rhetorical norms. The study suggests a culturally responsive hybrid assessment model where AI handles initial technical screening, and lecturers focus on contextual evaluation. This approach balances AI's efficiency, which could reduce workloads by 60%, with human expertise in culturally relevant feedback, providing a practical framework for Indonesian EFL institutions undergoing digital transformation while maintaining educational integrity.]]>
