The determination of food authenticity is a crucial issue for food quality and safety. Midinfrared spectroscopy provides rapid chemical profiling of agricultural products and could become an effective tool for authentication when coupled to chemometrics. This study developed a simple protocol for classifying commercial juices using attenuated total reflectance infrared spectroscopy. Spectra from a total of 52 juices together with their extracted sugar-rich and phenol-rich fractions were obtained to construct multivariate models [hierarchical cluster analysis (HCA) and soft independent modeling of class analogy (SIMCA)] for pattern recognition analysis and prediction. Spectra of the sugar-rich fraction, comprised primarily of sugars and simple acids, almost superimposed the whole juice spectra. Solid-phase extraction enriched phenol compounds and provided signature-like spectral information that substantially improved the SIMCA modeling power over the whole juice or sugar-rich fraction models and allowed for the differentiation of juices with different origins. Zero percent misclassification was achieved by the phenol-rich fraction model. HCA successfully recognized the natural grouping of juices based on ingredients similarity. The infrared technique assisted by a simple fractionation and chemometrics provided a promising analytical method for the assurance of juice quality and authenticity. © 2007 American Chemical Society.