Artificial intelligence (AI) has brought new ways for implementing learning analytics in computer-supported collaborative learning (CSCL). However, there is a lack of literature reviews that focus on AI-driven learning analytics applications and tools in CSCL contexts. To fill the gap, this systematic review provides an overview of the goals, characteristics, and effects of existing AI-driven learning analytics applications and tools in CSCL. According to the screening criteria, out of the 2607 initially identified articles between 2004 and 2023, 26 articles are included for final synthesis. Our results show that existing tools primarily focus on students’ cognitive engagement. Existing tools primarily utilize communicative discourse, behavioral, and evaluation data to present results and visualizations. Despite various formats of feedback are provided in existing tools, there is a lack of design principles to guide the tool design and development process. Moreover, although AI techniques have been applied for presenting statistical information, there is a lack of providing alert or suggestive information in existing tools or applications. Compared with the positive impacts on collaborative learning, our results indicate a lack of support for instructional interventions in existing tools. This systematic review proposes the following theoretical, technological, and practical implications: (1) the integration of educational and learning theories into AI-driven learning analytics applications and tools; (2) the adoption of advanced AI technologies to collect, analyze, and interpret multi-source and multimodal data; and (3) the support for instructors with actionable suggestions and instructional interventions. Based on our findings, we provide further directions on how to design, analyze, and implement AI-driven learning analytics applications and tools within CSCL contexts.