Driven by educational digital transformation and core literacy education, English interdisciplinary teaching has become vital for breaking subject barriers in primary and secondary schools. Traditional English teaching overemphasizes mechanical language drills and neglects interdisciplinary integration and students' problem-solving abilities. AI, big data and cloud teaching platforms can address the prominent drawbacks of conventional interdisciplinary teaching such as fragmented resources, rigid teaching modes, simplistic evaluation and insufficient personalized learning support.This study adopts a mixed-method design combining questionnaires, semi-structured interviews and eight-week classroom action research. It collects empirical data from 168 English teachers and 1,247 students across 32 primary and secondary schools in central and eastern China via quantitative statistics and qualitative coding. Four major bottlenecks limiting AI-interdisciplinary teaching integration are identified: incomplete localized resource systems, teachers' insufficient compound digital literacy, students' ingrained passive learning habits, plus superficial AI application with cultural bias and student data privacy risks.Drawing on two replicable AI classroom cases with specific tool schemes and pre-post test data, this paper proposes five feasible optimization strategies: building regional intelligent interdisciplinary resource banks, developing targeted digital training for teachers, designing layered inquiry learning frameworks, creating differentiated AI interdisciplinary homework systems, and setting norms for technical application and educational risk management. The findings enrich localized empirical evidence for CALL and CLIL theories in China's basic education and offer practical intelligent interdisciplinary teaching frameworks for frontline English teachers.