A proportion of patients with myasthenia gravis (MG) without acetylcholine receptor (AChR) antibodies have antibodies to muscle-specific kinase (MuSK). MG with MuSK antibodies (MuSK-MG) is often associated with persistent bulbar involvement, including marked facial weakness and tongue muscle wasting. The extent of muscle wasting in MuSK-MG, and whether it is also found in the few acetylcholine receptor (AChR-MG) patients who have persistent bulbar involvement, is not clear. We studied 12 MuSK-MG patients and recruited 14 AChR-MG patients matched broadly for age, sex ratio, duration of disease and degree of ocular, bulbar and facial weakness. We used coronal and sagittal T1-weighted (T1W) and T2-weighted (T2W) magnetic resonance imaging (MRI) to assess muscle wasting in facial and tongue muscles. Hyperintense signal on T1W MRI and comparison of axial T1W sequences with cUTE sequences were used to assess fibrous/fatty tissue in the tongue. We compared the results with those of four patients with myotonic dystrophy and 12 healthy individuals. We correlated the changes with clinical and treatment histories, and established a new ocular-bulbar-facial-respiratory (OBFR) score. At the time of study, none of the clinical measures, including the OBFR score, differed between the two MG groups. MRI demonstrated thinning of the buccinator, orbicularis oris (O.oris) and orbicularis oculi (O.oculi) muscles in MuSK-MG patients compared with healthy controls, whereas thinning of these muscles was not significant in AChR-MG. Tongue areas with T1W high signal were increased in MuSK-MG patients and the intensity of the signal on axial T1W sequences was greater in MuSK-MG than in controls. To look for possible correlations between imaging and clinical findings, we pooled results from all MG patients. The duration of treatment with prednisolone at >40 mg on alternate days (AD) correlated positively with the percentage of tongue area with high signal (P = 0.006) and negatively with MRI measurements of individual muscles and with the mean muscle dimensions (P = 0.001). The new OBFR score correlated positively with current Myasthenia Gravis Foundation of America grades and with the percentage of high signal (P = 0.004) and negatively with the mean muscle dimensions (P < 0.001). The results show that bulbar and facial muscle weakness and wasting are associated with significant muscle atrophy and fatty replacement in MuSK-MG, which was not found in the AChR-MG patients. MuSK antibodies per se may predispose to muscle thinning, but the difficulties in obtaining clinical remission under steroid therapy in some patients, resulting in long duration of treatment with higher doses (>40 mg AD), may be an additional factor.