Concerns have been raised that randomized placebo-controlled trials (RCTs) in non-radiographic axial spondyloarthritis (nr-axSpA) might be failing to identify patients that best show differences in clinical response rates between those receiving active drug and those receiving placebo therapies; in addition, some studies might even be showing spurious differences in responses to TNF and IL-17 inhibitor therapies. In particular, the most recent phase III RCTs in nr-axSpA have reported variable and generally lower response rates than observed in phase III trials of patients with ankylosing spondylitis and in trials conducted a decade ago in patients with early axSpA who were selected on the basis of axial inflammation evident on MRI scans. We argue that these observations at least partly reflect an RCT design that does not take full advantage of MRI to select patients who are responsive to therapy because the current MRI-based inclusion criteria cannot identify patients with axSpA with sufficient specificity. We propose that future studies should be designed using revised patient inclusion criteria based on expanded MRI evaluation and the application of data-driven definitions of a positive MRI for inflammatory and structural lesions typical of axSpA reported in an international multicentre analysis of MRI scans from the Assessment of SpondyloArthritis International Society (ASAS) classification cohort.