Global stock markets represent vast repositories of wealth, and as economies worldwide become increasingly market oriented, the ability to forecast stock performance has gained critical importance for investors, policymakers, and researchers alike. In accordance with 380 research papers indexed in the Scopus database between 2020 and 2026, this study delivers a bibliometric impression of enquiry on stock market prediction exploiting machine learning (ML) and artificial intelligence (AI) methods. The review of the most-cited studies highlights the predominance of deep learning designs particularly Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Convolutional Neural Network (CNN), and hybrid models while ensemble and reinforcement learning methods demonstrate superior predictive accuracy and adaptability. Moreover, fusion models integrating numerical and sentiment data consistently outperform single-source approaches. The bibliometric mapping, conducted using VOSviewer, delivers all-inclusive impression of the field’s intellectual structure, research impact, and global collaboration patterns. Expert Systems with Applications emerged as the most productive and influential journal. A strong China-United States–India collaboration axis was observed, with most research contributions originating from computer science departments. Thematic mapping revealed key research clusters and underscored gaps in risk-adjusted returns and portfolio-level optimization, indicating promising directions for future research and practical implementation...