Artificial intelligence in education: How has research evolved over time? A longitudinal bibliometric analysis

Objective. A longitudinal approach based on author keywords was used to analyze the thematic evolution of research on artificial intelligence (AI) in education. This approach enabled the identification of changes in the field’s vocabulary, established and emerging themes, and shifts in their relationships, as well as their behavior across five analytical dimensions (AI technologies/techniques, algorithms/models, platforms/systems, educational level, and methodological approach). Design/Methodology/Approach. A comprehensive search was conducted on Scopus (1973–2025) using an advanced title search algorithm that combined terms from the field of AI with educational terms. The analysis was based exclusively on the authors’ keywords. A global co-occurrence network and a temporal visualization (overlay) were constructed in VOSviewer. The minimum threshold for terms within each cluster was set at 15, resulting in a final map comprising 1,197 keywords, which were subsequently organized into 20 clusters. Results/Discussion. The field of AI in education exhibited a robust technical foundation, with a focus on machine learning, deep learning, and neural networks. This foundation was complemented by a more nascent trend that was predominantly characterized by generative AI. During the 2024–2025 period, there was an escalation in the use of vocabulary related to ChatGPT, language models, and applied practices. This escalation was accompanied by a notable increase in terms related to ethics, integrity, and governance. The cluster-based structure reflected a broad thematic organization, with connections across technological, pedagogical, and institutional fronts. Conclusions. The field of AI research in education has undergone a significant shift in focus, transitioning from a more “traditional” approach, which was dominated by techniques such as machine learning and deep learning, toward a new phase marked by the emergence of generative AI and language models, such as ChatGPT. This transition has been most evident in the domain of higher education. Concurrently, subjects such as AI literacy, academic integrity, bias, and governance have emerged as the most pressing issues in the current discourse. Originality/Value. The study provided a longitudinal perspective on AI research in education and a thematic analysis by dimension, which facilitated the identification of topics that have persisted, topics that have emerged, and the evolution of the field over time.