Concrete carbonation is one of the key factors affecting the durability of structures. Accurate simulation of the carbonation process is of great significance for structural durability design and service life prediction. Mesoscopic simulation of carbonation can describe the influence of aggregates on carbonation depth, but considering mesoscale heterogeneity of concrete usually leads to increased computational complexity, especially when dealing with engineering-scale components where computational efficiency is often unacceptable. In this research, we proposed a mesoscopic simulation method based on cellular automata for concrete carbonation. In this method, the concrete aggregate content, carbon dioxide concentration, calcium hydroxide concentration, and porosity are defined as cell state parameters. The cell evolution rules are defined based on the phenomenon of carbon dioxide diffusion with reaction. Finally, through the evolution process of the CA model, the prediction of the spatiotemporal distribution of the key substance concentrations involved in the carbonation process is achieved. By comparing with available experimental results, the accuracy of the CA model in simulating carbonation depth is validated. Compared to traditional mesoscopic simulation solution method, the CA model can provide accurate simulations of average calcium hydroxide concentration distribution and carbonation front. Besides, this method demonstrates a computational efficiency improvement of approximately 11 times compared to traditional solution method. This is of significant importance for simulating carbonation in engineering-scale components of further study.