Artificial Intelligence is fundamentally transforming and reshaping the way we live and work. Al can also be adopted in libraries for technical services such as classification and cataloguing, and for library management, such as staffing and decision-making. This study is guided by the PRISMA framework to analyse studies on the use of machine learning or artificial intelligence in library classification. A quantitative approach was adopted to assess the applicability of machine learning or artificial intelligence (AI) for automating the classification of DDC, UDC, LCC, and CC. The research articles were retrieved from the SCOPUS, Web of Science, and Google Scholar databases. The search results found 535 documents in Scopus, 99 in WOS, and 473 in Google Scholar. The researcher identified 27 documents as closely related and retrieved their full texts for review. This analysis revealed that the DDC scheme is commonly used to automate classification. SVM and BERT are the most popular models among the studies. Only 20 of the 27 studies provided formal evaluation metrics such as accuracy, precision, recall, and F1-score. Highest accuracy found in Dewey Decimal Classification (DDC) with Support Vector Machine (SVM) model, with F1 scores of 0.80 to 0.85. The challenges in achieving fully automated classification included insufficient datasets, the complexity of classification schemes, and the technical limitations of machine learning models. This research has made a valuable contribution to the existing body of knowledge. It is the first systematic review to examine machine learning in library classification quantitatively. It has also provided real-world problems for the application of machine learning or Al technologies in Classification.