Species identification of botanical products is a crucial aspect of research and regulatory compliance; however, botanical classification can be difficult, especially for morphologically similar species with overlapping genetic and metabolomic markers, like those in the genus Ocimum. Untargeted LC–MS metabolomics coupled with multivariate predictive modeling provides a potential avenue for improving herbal identity investigations, but the current dearth of reference materials for many botanicals limits the applicability of these approaches. This study investigated the potential of using greenhouse-grown, authentic Ocimum to build predictive models for classifying commercially available Ocimum products. We found that three species, O. tenuiflorum, O. gratissimum, and O. basilicum, were chemically distinct based on their untargeted UPLC-MS/ MS profiles when grown in controlled settings. Combining non-targeted fingerprints with high-performance thin-layer chromatography (HPTLC) approach reveled two distinct O. tenuiflorum chemotypes. Three predictive models (partial least squares, LASSO regression, and random forest) were employed to extrapolate these findings to commercially available products. Discrepancies between controlled and commercial samples led to inconclusive classifications. This suggests that environmental and processing variance exceed each model’s tolerance, and an expanded reference library is necessary for successful predictions. Overall, this study highlights that variance in metabolomic profiles is driven by genetic differences (species and cultivar level), environmental factors (greenhouse vs commercial growth), and processing conditions (post-harvest treatment). Future studies may establish variance thresholds for these factors to be included in sample selection and study designs for robust and reliable authentication workflows.