The effect of extraction methods on detecting hydrocarbon oxidation products (HOPs) in groundwater remains unclear. HOPs are polar, water-soluble byproducts of petroleum biodegradation. Our previous work showed that liquid–liquid extraction (LLE), a method commonly used in regulatory monitoring, has a significantly lower extraction efficiency for HOPs compared to solid-phase extraction (SPE). In this study, we evaluate the analytical limitations and compositional selectivity of LLE and SPE using groundwater samples from the Bemidji, MN, crude oil spill site. Optical properties were characterized using excitation–emission matrix spectroscopy (EEMs), and a three-component PARAFAC model was validated, showing consistent trends across both extracts and whole water samples. Ultrahigh-resolution mass spectrometry (UHR-MS) revealed that LLE selectively recovered aliphatic-like compounds but underrepresented more polar oxygenated HOPs. In contrast, SPE methods were more effective at isolating highly oxidized compound classes. These differences were consistent across a gradient of contamination. Overall, the LLE was less precise and less representative of polar HOPs, introducing bias in the characterization of HOPs. This study is the first to quantitatively demonstrate the compositional selectivity and analytical bias of LLE versus SPE for HOPs using combined EEM-PARAFAC and UHR-MS techniques, with implications for long-term monitoring and site assessment protocols.