The number of court decisions is increasing daily. Manual collection and analysis of data from available court decisions is a time-consuming process prone to errors, especially considering the volume of data and the complexity of document structures. Advances in artificial intelligence within the field of natural language processing are enabling the exploration and implementation of automated tools for extracting and analyzing large amounts of data. This paper compares and evaluates the application of the ChatGPT model in combination with machine learning and deep learning models, such as SVM, CNN, and LSTM, on court decisions from New Zealand courts with the aim of classifying documents into corresponding legal areas and extracting relevant information. Three approaches to text preparation for classification are compared against the results of traditional preprocessing methods. By using constructed instructions, the ChatGPT model successfully extracts accurate data in most cases. Although a relatively small number of errors and hallucinated data are observed, it proved as the most effective solution for classification of court decisions.