To investigate the reliability of sacroiliac radiographic grading based on the modified New York criteria, both through grading pelvic radiographs alone and in combination with sacroiliac magnetic resonance imaging and the influence of alone and multimodal interpretation on the categorization of radiographic and non-radiographic axial spondyloarthritis (r-axSpA and nr-axSpA). Radiographs of the SI joints from patients with axSpA were graded by two rheumatologists independently in two sessions (radiographs only and MRI + radiographs) and followed by consensus sessions with the participation of a radiologist. Overall, anteroposterior pelvic radiographs and T1 weighted MR scans of the sacroiliac joints from 138 patients with axSpA were eligible. Intra- and inter-reader reliability was fair to moderate for both techniques regardless of whether they were examined alone or in conjunction with MRI, however confidence levels of readers were relatively higher in the MRI + radiographs combined technique compared to radiographs only. Regarding the classification, both readers’ intra and inter reader agreements were moderate to substantial. The kappa values and agreement were notably higher for the combined reading technique. In the consensus sessions, 10.2% of patients were recategorized. Merging the information from MRI scans with pelvic radiographs yielded no notable enhancement in the accuracy of radiographic grading of SI joint. Categorizing axSpA as either radiographic or non-radiographic based on pelvic radiography falls short in reliability even after combining with MR images and may likely call for a more reliable definition, axSpA with or without structural changes, relying on more advanced imaging techniques than radiographs.