The automated identification and quantification of Li-bearing phases in core samples and hand specimen has gained interest due to increasing industrial demand for lithium. Despite the fact, that Li cannot be directly detected by µEDXRF, the spectral signatures of a number of Li-bearing phases such as spodumene, lepidolite, amblygonite and others are characteristic enough to differentiate them from non-Li phases. By applying hyperspectral ENVI software and its spectral angle mapper (SAM) algorithm a supervised classification based on an endmember data bases can be used to image the phase distribution. SAM considers all ratios of available band based on total signal intensity per pixel per selected spectral region to offer a measure to compare an unknown pixel with all available endmembers of a database. A good knowledge of the paragenesis of each investigated system is key to limit ambiguities between Li phases, and non-Li phases carrying elements, such as H, Be, B, which cannot be detected either. This will be achieved by tailoring the endmember database according to these premises. Quantitative µEDXRF-based mineralogy is based on interpretation of chemical signatures, but shows some limitations regarding isochemical phases, isochemical mixtures at grain boundaries and sporadically strong diffraction signals of one or the other grain. Results which can be obtained are modal mineralogy, grain area size distribution, grain aspect ratio and sample and grain area chemistry.