Abstract: Chloride ion diffusion in reinforced concrete (RC) is widely recognized as one of the primary causes of reinforcement corrosion and exhibits significant uncertainty and randomness. Accurate prediction of chloride concentration profiles is therefore essential for durability assessment, structural safety evaluation, and maintenance decision-making of concrete structures. Based on the traditional stochastic chloride diffusion model derived from Fick’s second law, a modified stochastic model incorporating an error term with unknown variance was established to characterize the nonlinear dynamic diffusion process of chloride ions. Subsequently, a sequential Bayesian updating method for posterior parameter distributions was developed within a Bayesian network (BN) framework using Bayes’ theorem and Markov chain Monte Carlo (MCMC) algorithm, in which the newly acquired experimental data were continuously incorporated to update the model parameters. The updated parameters were then substituted into the probabilistic model to realize dynamic prediction of the spatiotemporal distribution of mean chloride concentration in concrete. The chloride concentration data obtained from RC beam tests subjected to drying–wetting cycles were employed to illustrate the application of the proposed sequential Bayesian updating approach for dynamic chloride concentration prediction. The results show that, compared with analytical models and conventional stochastic models, the proposed BN-based dynamic prediction model effectively reduces parameter uncertainty by incorporating prior information and continuously correcting model parameters using the newly measured data. Consequently, the predicted chloride concentration profiles exhibit improved objectivity, accuracy and reliability, indicating that the proposed model is more suitable for long-term dynamic prediction of chloride ion diffusion in concrete structures.