The spectrum of MOG-IgG-associated disease (MOGAD) includes optic neuritis (ON), myelitis (MY), acute disseminated encephalomyelitis (ADEM), brainstem encephalitis, cerebral cortical encephalitis (CE) and AQP4-IgG-negative neuromyelitis optica spectrum disorder (NMOSD). In MOGAD, MOG-IgG are usually detected in sera (MOG-IgGSERUM), but there have been some seronegative MOGAD cases with MOG-IgG in CSF (MOG-IgGCSF), and its diagnostic implications remains unclear. In this cross-sectional study, we identified patients with paired serum and CSF sent from all over Japan for testing MOG-IgG. Two investigators blinded to MOG-IgG status classified them into suspected MOGAD (ADEM, CE, NMOSD, ON, MY and Others) or not based on the current recommendations. The MOG-IgGSERUM and MOG-IgGCSF titres were assessed with serial 2-fold dilutions to determine end point titres [≥1:128 in serum and ≥1:1 (no dilution) in CSF were considered positive]. We analysed the relationship between MOG-IgGSERUM, MOG-IgGCSF and the phenotypes with multivariable regression. A total of 671 patients were tested [405 with suspected MOGAD, 99 with multiple sclerosis, 48 with AQP4-IgG-positive NMOSD and 119 with other neurological diseases (OND)] before treatment. In suspected MOGAD, 133 patients (33%) tested MOG-IgG-positive in serum and/or CSF; 94 (23%) double-positive (ADEM 36, CE 15, MY 8, NMOSD 9, ON 15 and Others 11); 17 (4.2%) serum-restricted-positive (ADEM 2, CE 0, MY 3, NMOSD 3, ON 5 and Others 4); and 22 (5.4%) CSF-restricted-positive (ADEM 3, CE 4, MY 6, NMOSD 2, ON 0 and Others 7). None of AQP4-IgG-positive NMOSD, multiple sclerosis or OND cases tested positive for MOG-IgGSERUM, but two with multiple sclerosis cases were MOG-IgGCSF-positive; the specificities of MOG-IgGSERUM and MOG-IgGCSF in suspected MOGAD were 100% [95% confidence interval (CI) 99-100%] and 99% (95% CI 97-100%), respectively. Unlike AQP4-IgG-positive NMOSD, the correlation between MOG-IgGSERUM and MOG-IgGCSF titres in MOGAD was weak. Multivariable regression analyses revealed MOG-IgGSERUM was associated with ON and ADEM, whereas MOG-IgGCSF was associated with ADEM and CE. The number needed to test for MOG-IgGCSF to diagnose one additional MOGAD case was 13.3 (14.3 for ADEM, 2 for CE, 19.5 for NMOSD, infinite for ON, 18.5 for MY and 6.1 for Others). In terms of MOG-IgGSERUM/CSF status, most cases were double-positive while including either serum-restricted (13%) or CSF-restricted (17%) cases. These statuses were independently associated with clinical phenotypes, especially in those with ON in serum and CE in CSF, suggesting pathophysiologic implications and the utility of preferential diagnostic testing. Further studies are warranted to deduce the clinical and pathological significance of compartmentalized MOG-IgG.