Risk and Artificial Intelligence Adoption: A Scientometric and Thematic Evolution Analysis Based on Scopus and Web of Science (1990-2025)

The integration of risk into artificial intelligence (AI) adoption research has evolved from a peripheral behavioral inhibitor into a structurally defining governance construct. This study provides a PRISMA-compliant scientometric synthesis of 612 peer-reviewed documents indexed in Scopus and Web of Science between 1990 and 2025. Using performance analysis, thematic mapping (centrality–density), thematic evolution (Sankey), co-citation network analysis, and collaboration mapping, the study reconstructs the intellectual, conceptual, and social structure of the risk–AI adoption nexus. Results indicate three structural phases: an early reliability-oriented stage; a normative turn emphasizing privacy, bias, and accountability; and a consolidation phase characterized by compliance and risk-based governance architectures. Thematic migration reveals a shift from perceived individual risk toward system-level accountability logics. The co-citation backbone demonstrates continued anchoring in technology acceptance theory while increasingly integrating governance scholarship. The findings clarify the cumulative development of this interdisciplinary field and identify explainability, auditability, and regulatory compliance capability as emergent research frontiers.