Researchers have thoroughly examined language learners' perspectives on ChatGPT using technology acceptance models, particularly considering its integration into computer-assisted language learning (CALL). Nevertheless, further investigations need to implement a theoretical framework that has a pedagogical-oriented perspective. To address this gap, this research applied learners' approaches to the learning environment (SAL). Moreover, it expanded SAL by integrating a multilevel perspective that includes contextual, individual, and ChatGPT-related factors. In the current research, three universities in Ardabil integrated ChatGPT into their language syllabus, enabling 192 language learners to learn language with it and respond to the study questionnaire. The results of partial least squares modeling (PLS-SEM) indicated that ChatGPT leadership, in case the university authorities provide the atmosphere conducive to the norms of ChatGPT integration, could mainly determine language learners’ organizing approach to using it in their daily academic schedule. Furthermore, the study revealed that personalization and anthropomorphism, two key ChatGPT-related factors, significantly influenced learners' deep approach to using ChatGPT as a source for meaningful, cross-referenced CALL tools. On the other hand, it was reported that low feedback reliability, privacy concerns, and ChatGPT's perceived value supported language learners' surface approach that limited its use as a ChatGPT-related factor. Considering the results, the study brought forward a new conceptual framework for CALL and Artificial Intelligence Language Learning (AILL) and proposes that ChatGPT leadership should be encouraged at a macro-contextual level that may encompass other micro-contextual, personal, and ChatGPT-related factors.