Land-use change is an important representation of urban-rural system evolution. In rural areas, the modern agriculture industry promotes spatial pattern changes. This study introduces a multi-agent system into the spatial simulation and reorganisation of rural settlements. The cellular automata and agent-based model (CA-ABM), a type of simulation and prediction model, was constructed using the geographic information system platform and CA tool. Based on data from 2000 to 2020, the parameters of the model were revised, and the model was used for spatial prediction in 2030. Using six rural settlements from plain areas, hilly areas, and mountainous areas as examples—combined with field investigation, remote sensing images, official land use planning, and other data—this study analysed land-use evolution under the influence of farmers, developers, and government agents. From the perspective of the three kinds of agents’ decision-making, land use prediction and optimisation suggestions were proposed. Compared with previous research, this study shows that the CA-ABM can reflect the dynamic behaviour of farmer agents, government agents, and developer agents in the simulation system. With the help of the CA tool, the model successfully fuses rural settlement attribute information and a spatially accurate simulation. This study can provide a scientific basis for the quantitative study of rural settlement land use prediction. Simultaneously, the proposed spatial optimisation mechanism provides a reference for the spatial reorganisation of agricultural rural settlements (ARS) worldwide.