The intra-sphere and inter-sphere structural attributes of controlled release microsphere drug products can greatly impact their release profile and clinical performance. In developing a robust and efficient method to characterize the structure of microsphere drug products, this paper proposes X-ray microscopy (XRM) combined with artificial intelligence (AI)-based image analytics. Eight minocycline loaded poly(lactic-co-glycolic acid) (PLGA) microsphere batches were produced with controlled variations in manufacturing parameters, leading to differences in their underlying microstructures and their final release performances. A representative number of microspheres samples from each batch were imaged using high resolution, non-invasive XRM. Reconstructed images and AI-assisted segmentation were used to determine the size distribution, XRM signal intensity, and intensity variation of thousands of microspheres per sample. The signal intensity within the eight batches was nearly constant over the range of microsphere diameters, indicating high structural similarity of spheres within the same batch. Observed differences in the variation of signal intensity between different batches suggests inter-batch non-uniformity arising from differences in the underlying microstructures associated with different manufacturing parameters. These intensity variations were correlated with the structures observed from higher resolution focused ion beam scanning electron microscopy (FIB-SEM) and the in vitro release performance for the batches. The potential for this method for rapid at-line and offline product quality assessment, quality control, and quality assurance is discussed.