<p>The quality and price of wine depend on raw materials, geographical origin, specific production processes, and product aging. The prediction of the age of wine based on chemical composition is laborious and time-consuming, so spectroscopic methods are studied as a faster and cheaper but still reliable alternative. In this study, UV-Vis absorption and synchronous fluorescence (SF) spectral data sets for undiluted and diluted sweet wines were obtained, and each data set was then evaluated separately by partial least squares (PLS) regression to predict the age of the wines, using the entire data set or significant spectral variables selected by forward selection of predictors and Variable Importance in Projection-PLS (VIP-PLS). The VIP-PLS model based on the SF spectral data of diluted samples had the best performance characteristics (coefficient of determination of cross-validation (R²CV) of 0.992, root mean square error of cross-validation (RMSECV) of 0.5 years, coefficient of determination of prediction (R²P) of 0.990, root mean square error of prediction (RMSEP) of 0.8 years, and relative predictive deviation (RPD) of 9.2). To assign the selected spectral variables to chemical compounds, phenolic compounds (catechin, caffeic acid, caftaric acid, gallic acid, protocatechuic acid, and <i>p</i>-coumaric acid) were determined by the HPLC-DAD method, and their contents were correlated with the age of the wine. A high correlation was found between the age of the wine and the content of caftaric acid, catechin, and gallic acid (<i>p</i> &lt; 0.0001). Therefore, and considering their characteristic spectral features, these compounds can be related to some of the selected variables. Overall, SF spectroscopy combined with chemometric methods could establish an advanced analytical tool for wine age prediction that significantly reduces analytical time, cost, and environmental damage compared to chemical and chromatographic methods.</p>