Artificial Neural Network (ANN) models have found many applications in hydrology; however, the need for uncertainty quantification of their results becomes more and more obvious. In this work the bootstrap percentile method is combined with an ANN model, already developed to predict a karstic aquifer’s response, to provide confidence intervals for the results of the model. The percentile method can be used for automatic confidence intervals generation, though the procedure is quite time consuming. The 95% confidence intervals were first computed and the actual coverage of the intervals was calculated and compared to the theoretical ones. In a second step the confidence intervals and their success rate for different degrees of nominal certainty levels were also computed. The presence of outliers in the training and testing data, which cannot be adequately modeled by the ANN network, resulted in a reduced agreement between actual and nominal coverage of the percentile intervals.