This study presents the first comprehensive dual-database bibliometric analysis of game-based learning research, investigating its evolution between 2015 and 2024 through analysis of 34,125 documents from Scopus (18,487) and Web of Science (WoS) (15,638) using advanced visualization techniques. The findings reveal consistent field expansion with notable database variations: Scopus demonstrates steady growth while WoS experienced a significant 2020 decline. Citation analysis shows a higher impact on 2015-2018 studies, with Hwang, G. J. as the most influential author and "Computers and Education" as the top journal. Co-keyword analysis reveals distinct conceptual structures--Scopus exhibits two clusters (human-centered studies and game-based learning/serious games) versus three in WoS (education-motivation, healthrehabilitation, technology-design). Thematic analysis demonstrates that game-based learning, serious games, and virtual reality occupy different maturity positions across databases, indicating varied theoretical consolidation. Collaboration networks show US, China, and European dominance, with weaker ties involving developing countries. These findings provide critical insights for educational policymakers and curriculum designers, identifying research clusters to guide targeted funding, collaboration gaps requiring international partnerships, and evidence-based guidance on established versus emerging game-based learning approaches. The dual-database methodology offers comprehensive research landscape mapping that supports strategic decisions in educational technology implementation and research prioritization.