This qualitative study aimed to investigate the mechanisms, effects, and implications of Artificial Intelligence (AI) application in representing and facilitating the understanding of abstract phenomena and concepts in elementary science education. The research design followed a qualitative approach, employing a systematic review and thematic synthesis methodology, guided by the PRISMA statement. The scope of the research encompassed credible studies published between 1390-1404 Solar Hijri (2010-2025 AD). A purposive sampling strategy, based on predefined criteria, was utilized to select the study sample. The search strategy, adhering to PRISMA guidelines, involved searching, screening, and evaluating relevant articles from domestic databases including Magiran, IranDoc, Civilica, and Noormags, as well as international databases such as Web of Science, Scopus, ScienceDirect, and Google Scholar, using specific keywords. After excluding duplicates and applying inclusion/exclusion criteria, 10 articles were selected for final analysis. The data were analyzed using the thematic analysis method to extract the main themes. Findings revealed that AI, through mechanisms such as multimodal representations, interactive simulations, knowledge modeling, and personalized learning, significantly aids in better comprehension of abstract phenomena and concepts in elementary science. Specifically, AI reduces cognitive load, enhances causal reasoning, increases learner motivation and engagement, and elevates learning from rote memorization to deep conceptual and exploratory understanding. Consequently, AI can serve as an effective tool for transforming elementary science education. However, its successful implementation necessitates appropriate instructional design and readiness in terms of infrastructure and professional development for teachers.