The corpus callosum (CC) plays an important role in interhemispheric cerebral communication, facilitating the integration and coordination of many brain signals. Its division into subregions, called parcellation, is crucial for studying many brain conditions. However, prevalent parcellation methods rely on fixed geometrical partitioning, failing to adapt to the known variability of the CC across individuals. Data-driven methods based on diffusion MRI have been proposed, but they lack adequate validation. In this study, we used tractography-based analysis to investigate the consistency and compare four CC parcellation methods: Witelson’s and Hofer’s geometrical approaches, and Cover’s and Santana’s diffusion MRI-based data-driven methods. Whole-brain tractograms of one hundred subjects from the Human Connectome Project were segmented using the parcellation masks from each method. We then calculated the cortical surface coverage of each segmented tractogram and their inter-subject density correlation coefficients. The average number of streamlines passing through different cortical areas and their relationship with each CC subregion was also assessed. The results revealed significant differences between the methods. Geometrical approaches showed inconsistencies with the expected cortical connections in their subregions. Santana’s method presented higher consistency in density correlations and cortical connections, particularly in CC Regions III and IV, which are primarily connected to motor and somatosensory cortical areas. The findings suggest that diffusion MRI-based methods, especially those incorporating directional information, may produce more consistent CC parcellations. Nonetheless, all methods exhibited inconsistencies, resulting in CC subregions that are connected to multiple cortical areas simultaneously.