Different environmental issues such as carbonation and corrosion due to chloride threaten aging reinforced concrete (RC) bridges that are in service in areas highly prone to corrosion and earthquakes. Significant experimental and numerical efforts have been put into scrutinizing the effect of corrosion on nonlinear behavior of structural elements. With the rapid development of artificial intelligence, useful methods are now provided to allow for the assessment of such bridges without the drawbacks and limitations of the experimental and numerical methods. In this paper, four machine learning (ML) algorithms are employed; linear regression (LR), decision tree (DT), random forest (RF), and XGBoost for data fitting of the models, and Bayesian search is used for optimization of hyperparameters. Numerical models of RC piers with stochastic parameters defining geometry, loading, and materials are built, and the degradation due to corrosion is applied with a randomly determined level of corrosion. Then, the corroded models are nonlinearly analyzed with random ground motions scaled to design-based and maximum credible earthquake spectra, and maximum drift ratios are stored. Using the created database, different ML models are compared to find the most accurate one. R-squared, mean absolute error, mean squared error, and root mean squared error metrics are considered as the criteria for the selection of the most accurate model. LR model with R2 = 0.53, MAE = 0.0026, mean squared error (MSE) = 1.4 × 10−5, and root mean squared error (RMSE) = 0.0036 has the lowest accuracy while XGBoost with R2 = 0.8, MAE = 0.0015, MSE = 5 × 10−6, and RMSE = 0.0028 is the most accurate model. DT and RF models with R2 = 0.7 and R2 = 0.73, respectively, are in between.