The present work is devoted to the study of the spectrofluorimetric behavior of the herbicide bentazone in the presence of selected organized assemblies, with the purpose of optimizing the working conditions for its determination in environmental samples. Bentazone is one of the most applied herbicides in countries of profuse agriculture, and belongs to the group of relevant persistent pollutants. Among the studied media, methyl-β-cyclodextrin produced the largest fluorescence signals for bentazone, and was selected as auxiliary reagent for the quantitative analyses. The selectivity of the method, i.e., its capability to measure accurately an analyte in the presence of interferents that may be present in the sample matrix, was improved through second-order multivariate calibration, measuring excitation–emission fluorescence matrices for both artificial and real samples, and processing them with the algorithm parallel-factor analysis (PARAFAC). A detection limit of 3 ng mL−1 and a relative error of prediction of 5.6% for bentazone were obtained in samples without interferents. In samples containing seven additional agrochemicals acting as interferences, the limit of detection and the relative error of prediction were 5 ng mL−1 and 9.9%, respectively. The method was successfully applied to spiked natural water samples, with statistical parameters in these cases which were similar to those calculated in samples with interferences. In this study, the advantages of coupling fluorescence measurements in organized media and second-order chemometric analysis are exploited in order to rapidly quantify a frequent pollutant at low concentrations in a very interfering medium and using green chemicals.