Background: To explore the changes of gut microbiota in Graves' orbitopathy (GO) patients of different severity grades and to identify the pathogenic bacteria of GO and the associated mechanism.Methods: A total of 18 healthy controls and 62 GO patients were recruited. The baseline information and faecal samples of all subjects were collected for gut microbiota analysis and metabolic function prediction analysis. 16SrDNA sequencing was used for microbial diversity detection. The operational taxonomic unit (OTU) was divided using the Mothur software, and the dominant microbiota was analysed. OTU number, Chao1 index, ACE index, and Shannon index of microbiota in faecal samples were analysed using the QIIME1.9.0 software. The relative abundance of microbiota in faecal samples was analysed through principal component analysis (PCA) using the Canoco Software 5.0. The metabolic function of microbiota in faecal samples was predicted using PICRUSt 2.0.Results: There was no remarkable difference in gut microbiota diversity between groups; however, the gut microbial community and dominant microbiota significantly differed among groups. Klebsiella_pneumoniae was deemed the potentially pathogenic bacteria of GO, and its abundance was positively correlated with disease severity. The metabolic prediction results revealed that inorganic nutrition metabolism, fatty acid and lipid degradation, electron transfer, aromatic compound degradation, and alcohol degradation were notably different between groups with high and low abundance of Klebsiella_pneumoniae and among groups with different GO severity grades, thereby showing a positive correlation with GO clinical risks.Conclusions: Klebsiella_pneumoniae was a potential GO-related pathogen, which may regulate the metabolic pathways to affect GO progression.