This study aims to develop a comprehensive competency framework for artificial intelligence (AI) literacy, delineating essential competencies and sub-competencies. This framework and its potential variations, tailored to different learner groups (by educational level and discipline), can serve as a crucial reference for designing and implementing AI curricula. However, the research on AI literacy by target learners is still in its infancy, and the findings of several existing studies provide inconsistent guidelines for educational practices. Following the 2020 PRISMA guidelines, we searched the Web of Science, Scopus, and ScienceDirect databases to identify relevant studies published between January 2012 and October 2024. The quality of the included studies was evaluated using QualSyst. A total of 29 studies were identified, and their research findings were synthesized. Results show that at the K-12 level, the required competencies include basic AI knowledge, device usage, and AI ethics. For higher education, the focus shifts to understanding data and algorithms, problem-solving, and career-related competencies. For general workforce, emphasis is placed on the interpretation and utilization of data and AI tools for specific careers, along with error detection and AI-based decision-making. This study connects the progression of specific learning objectives, which should be intensively addressed at each stage, to propose an AI literacy education pathway. We discuss the findings, potentials, and limitations of the derived competency framework for AI literacy, including its theoretical and practical implications and future research suggestions.