The aim of this study was to develop a simple, rapid, and reagent-free method for predicting the content of typical monocyclic aromatic hydrocarbons, benzene and toluene, in water using fluorescence spectroscopy data, chemometric methods, and random forest (RF) regression models. This was achieved through the reproducibility testing of fluorescence spectra configured at various concentrations of benzene and toluene, where the relative standard deviation was approximately 5%. An extensive study of the spectral characteristics of a large dataset of fluorescence spectra revealed that the benzene spectrum produced three peaks, with the strongest peak at 277 nm; the toluene spectrum produced a single peak at 284 nm. Subsequent spectral specificity analysis based on the substituents of benzene and toluene was conducted. A random forest regression model was established using the relationship between concentrations of benzene and toluene and fluorescence intensity. The optimal random forest regression models achieved a coefficient of determination (R2) of 0.997 and root mean square error (RMSE) of 3.06 for high concentrations of benzene, and an R2 of 0.994 and RMSE of 0.04 for various concentrations of toluene. This study demonstrates that fluorescence spectroscopy, distinct from conventional mass spectrometric and chromatographic methods, provides a convenient and novel approach for the detection and online monitoring of typical monocyclic aromatic hydrocarbons, such as benzene and toluene, in water.