As the geometallurgy concept gains more visibility, the importance of parameters, in particular, ore texture, in downstream processing performance is increasingly recognized, yet methodologies for fast, unbiased, and automated texture classification remain limited, particularly for complex and low-grade deposits. This study proposes an alternative approach with potential for automated ore texture classification by combining micro-X-ray fluorescence (μ-XRF) imaging with unsupervised machine learning. Drill core samples from northern Sweden iron ore deposits were analyzed using μ-XRF to produce high-resolution mineral and X-ray intensity maps, which were converted to grayscale, divided into patches, and processed with Gray Level Co-Occurrence Matrix (GLCM) for feature extraction. Additionally, Principal Component Analysis (PCA) was applied for dimensionality reduction and k-means clustering for textural classification. The integration of mineral and X-ray maps improved classification accuracy, with clustering results effectively distinguishing major textural groups, despite some misclassifications attributed to pixel intensity variations. Evaluation of possible correlation between the classified textures and Leeb hardness measurements was carried out. Promising results were obtained, however, future advancements, such as the application of deep learning and alternative clustering algorithms, could further enhance the accuracy and applicability of this technique.