With the rapid advancement of AI, its integration into EFL education has become increasingly prominent. This small-scale study provides insights from a Taiwanese higher education context, drawing on both EFL teachers’ and students’ perspectives. Guided by Cognitive Load Theory (CLT), the study examines how AI is used to support language learning tasks, involving varying levels of cognitive load, and explores its potential role in managing learners’ cognitive processing demands. A mixed-methods approach was employed, with quantitative and qualitative data collected through questionnaires. The study also explores patterns of AI usage, perceptions of its effectiveness, challenges encountered, and attitudes toward its future role in English education. The findings indicate that AI has become a widely used support tool in EFL learning. Students primarily use AI tools, such as ChatGPT and Grammarly, for writing, information searching, and grammar support, i.e., tasks that typically impose high intrinsic cognitive load on EFL learners. In contrast, teachers mainly employ AI for lesson preparation, material design, and efficiency enhancement, thereby reducing extraneous cognitive load. Although both groups hold positive attitudes toward AI integration, teachers demonstrate a more cautious approach due to concerns regarding usability, pedagogical alignment, data privacy, and potential students’ overreliance. From a CLT perspective, the results suggest that while AI can reduce extraneous cognitive load and provide procedural support, excessive dependence on AI may limit learners’ germane cognitive load by reducing their active engagement in language processing. Students largely agree that AI cannot replace human teachers, particularly in areas involving emotional support, interaction, and guidance. The data indicate that effective AI integration requires institutional support, professional training, and the development of digital literacy to ensure a balanced, human-centered approach to AI-enhanced EFL education.