<p>Reliable prediction of future land use and land cover change is essential for sustainable spatial planning in rapidly transforming African landscapes, including Ghana. Ghana has experienced accelerated urbanisation, agricultural expansion, and forest degradation. Yet the predictive performance of machine learning and cellular automata-based models applied in this context have not been systematically synthesised. This study conducts a systematic review and meta-analysis to evaluate the effectiveness of machine learning models, cellular automata models, and hybrid machine learning–cellular automata approaches for land use and land cover change prediction in Ghana between 2010 and 2025. Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, eighty-five peer-reviewed studies were selected from an initial pool of 12,008 records. Quantitative synthesis focused on standardised accuracy metrics, including overall accuracy, the Kappa coefficient, and the figure of merit. Results demonstrate that hybrid models consistently outperform standalone approaches across diverse land use contexts. Cellular automata–machine learning frameworks, including cellular automata-Markov-artificial neural network and future land use simulation models, achieved overall accuracy values exceeding 85 percent and Kappa coefficients commonly ranging from 0.80 to 0.89 in urban, forest, wetland, and river basin applications. Standalone machine learning classifiers attained high classification accuracies of up to 95–98 percent but showed limited capacity for spatial–temporal forecasting when applied without cellular automata integration. In contrast, standalone cellular automata–Markov models exhibited moderate predictive performance, with Kappa values typically between 0.75 and 0.82, nationally relevant.</p>