Malformation disease (MD) is one of the major constraints in the mariculture of the commercially important seaweed Saccharina japonica, which can cause severe losses of the kelp seedlings in the production process. However, the pathogenesis of MD remains largely undetermined. In this study, using cultivation experiments under laboratory conditions, MD severity was estimated for the young sporophytes of S. japonica, which were separately cultivated under different treatment conditions, including light intensity, duration of light exposure, and maturity level of the parent kelp. The associations between MD severity and epibacterial community divergence and assembly patterns were characterized. Higher MD severity significantly correlated with longer light exposure, and with both unmatured and overmatured parent kelp. The bacterial classes γ-Proteobacteria and α-Proteobacteria dominated all samples, but different MD severities were associated with different epibacterial communities. Comparisons of predicted functions for epibacterial communities revealed that longer light exposure led to a depletion in development and regeneration, while overmatured parent kelp resulted in a depletion in glycan biosynthesis and metabolism. Functional comparisons of the epibacterial communities between normal and unmatured parent kelp-generated seedlings revealed significantly different categories related to metabolism, environmental information processing, cellular processes, drug resistance, and bacterial infection. The significantly different pathways between groups, notably those related to bacterial infection and cellular processes, were partly consistent with the differences in bacterial compositions. The assembly of epiphytic bacterial communities was predominately governed by deterministic processes, and less impact was determined when there was significantly higher MD severity except when using overmatured parent kelp. Co-occurrence networks of the epibacterial communities associated with higher MD severity contained fewer nodes and exhibited lower modularity, but had higher graph density and degrees compared with those of seedlings with lower MD severity, indicating more complicated interactions. Nesterenkonia, Glycocaulis, Halomonas, Pseudoalteromonas, Pseudomonas, Loktanella, and Cobetia were frequently determined keystone taxa in communities associated with higher MD severity. The present study enhances our understanding of the factors significantly associated with MD severity and the potential roles of epimicrobiome in determining the disease severity, which will be useful for disease management in the future.