<![CDATA[Objective: This study aims to analyze the trends, developments, and characteristics of global scientific publications regarding the deep learning approach in elementary schools over an eleven-year period (2015–2025) in order to overcome the limitations of systematic scientific mapping using computational bibliometric approaches in this field. Methods: This study used a quantitative bibliometric design, collecting metadata from 985 deep learning articles in elementary schools indexed through Harzing's Publish or Perish application. Statistical data processing was carried out using Microsoft Excel, followed by computational mapping and analysis of network visualization, density, and overlays using the VOSviewer application to test the relationship between keyword co-occurrence. Findings: The analysis revealed that there are 985 publications with fluctuating trends that tend to increase, peaking at 149 articles in 2025 with an average annual growth of 89.5 publications. The VOSviewer mapping identifies 5 interconnected keyword clusters where the terms "deep learning", "elementary schools", "machine learning", and "implementation" form the core with the highest density. Meanwhile, topics that emerged as cutting-edge trends include Project-Based Learning (PBL) and learning outcomes (outcomes). Research Implications: These findings provide a comprehensive reference for researchers to determine future research directions, in particular highlighting pedagogical applied issues driven by digital technology change, and offer a foundation for stakeholders to integrate Artificial Intelligence technology in the active learning ecosystem in primary schools. Originality: This research contributes to presenting a unique eleven-year computational mapping of deep learning publications in elementary schools, as well as successfully uncovering research gaps that are still rarely touched, such as aspects of cognitive transfer ability and students' creativity in this discipline.]]>
