Indonesia has a high vulnerability to disasters, but the integration of disaster education in the curriculum is often still sporadic so that it has an impact on students' low science literacy. Conventional methods are considered inadequate, so a transition to a Problem-Based Learning (PBL) model is needed which is strengthened with a Deep Learning approach. This study aims to analyze and synthesize the current scientific literature landscape regarding the effectiveness of PBL integration with Deep Learning in disaster mitigation science learning. Using the Systematic Literature Review (SLR) method which refers to the PRISMA 2020 guidelines, this study examines empirical articles from the Scopus database for the period 2015–2025 which were analyzed with the help of VOSviewer. The results show that the current research trend is dominated by the social science education domain with a focus on Active Learning to improve higher level thinking skills (HOTS), but significant methodological gaps were found in the form of a lack of studies that integrate PBL and Deep Learning syntaxin its entirety in the context of science. This study concludes the need for a new conceptual framework that synergizes problem-solving structures with deep information processing to build adaptive and disaster-responsive science literacy.