An increasing attention has been directed to applying the artificial neural network (ANN) method in civil engineering. This paper examines the possibility of artificial neural network (ANN) to predict the chloride diffusion coefficient of concrete. A total 653 available sets of data from 13 literatures was used for establishing the network model. The developed ANN model used as many as 13 input variables, including water/cement ratio; the dosage of cement, superplasticizer, fly ash, granulated blast furnace, silica fume, aggregate; compressive strength; curing mechanism; testing method; testing time and environment to achieve one output parameter, referred to as chloride diffusion coefficient. The research results show that ANN is feasible in predicting the chloride diffusion coefficient in offshore concrete structures and the selected input variables are all correlated parameters.