This study explores the integration of ChatGPT, OpenAI's conversational AI, into English as a Foreign Language (EFL) classrooms at Korean universities, focusing on student interactions and language learning strategy preferences. It categorises interactions using the Strategy Inventory for Language Learning (SILL) and Strategic Self-Regulation (S2R) frameworks to evaluate the pedagogical effectiveness of AIassisted learning. Ninety-nine university students participated in training sessions with ChatGPT prompts tailored to different learning strategies. Data were collected through surveys, chat transcripts, and qualitative feedback. Results indicate frequent student interactions with ChatGPT, averaging 4.49 strategies in initial training chats. Compensatory strategies like error correction and adaptive difficulty received high ratings, while social strategies were rated lowest. Metacognitive strategies, especially planning and summarising, were also well-received. The study concludes that ChatGPT supports diverse learning strategies, enhancing linguistic competence and promoting self-regulated learning. However, limitations such as AI accuracy and authenticity issues highlight the need for continued human interaction in language education.