The bulk of recent research on generative AI (GenAI) in education mainly focused on its potential pedagogical uses as well as how students actually exploit such tools during learning, while little has been done to systematically investigate how teachers integrate those tools in different parts of their teaching process. To bridge this gap, the present study conducted in-depth interviews with 17 university English as a foreign language (EFL) teachers in mainland China, and scrutinized their different phases of teaching for any GenAI involvement. Through the lens of the ADDIE model (i.e., analysis, design, development, implementation, and evaluation), it was found that GenAI tools were mainly involved in the development stage (e.g., devising teaching activities and developing teaching materials). During the implementation stage, teachers merely introduced GenAI tools to their students, through means such as demonstrations and workshops, indicating a lack of organic integration of those tools into curriculum-based activities. GenAI involvement in the other three stages of instruction was reported to be none to minimal. The participants also shared both positive and negative experiences with GenAI in their EFL instruction. These findings highlight the urgent need for providing stage-specific professional training on how to integrate GenAI into different instructional stages and developing specialized educational GenAI programs that can produce accurate and high-quality outputs. Other implications for incorporating GenAI into education are also discussed.