At the immediate interface between the cell and the often-hostile environment, the bacterial capsule (CPS) provides protection against external assaults including the host immune response, antimicrobials and bacteriophage. As the specific structure can influence phage and monoclonal antibody specificity, robust prediction of CPS composition underpins the application of these therapies for treating resistant infections. For many bacterial species, one of the most significant determinants of CPS topology is the Wzy polymerase. This enzyme links sugars of adjacent glycan units to form mature CPS polymers, for which the type of Wzy can determine whether a CPS has a linear or branching pattern. However, despite their central role in CPS biosynthesis, Wzy are one of the most ill-defined proteins in bacterial systems and remain largely uncharacterised due to an extraordinary level of sequence diversity. Even the specific linkage they form is often unknown. Here, we conduct the first large-scale analysis of 245 Wzy identified in Acinetobacter baumannii, for which we have determined linkages formed by 80 of them. Based on shared Hidden Markov Models (HMMs), 136 were found to fall into one of three established protein families (EpsG, WzyC, and O-ag_pol_Wzy), while almost half (n=109) could not be assigned to any family. As sequence homology is not always concomitant with structural similarity, tertiary structures of all Wzy were modelled using Alphafold 2.0, then superimposed using an all-against-all algorithm to generate a TM score matrix for hierarchal clustering. This revealed that all Wzy could be grouped into one of three major structural families, which were shown to correlate with anomeric configurations (α or β) of the 80 known linkages. HMMs were redefined and used to screen large genome collections, resulting in identification of previously undetected wzy. This study forms the basis for developing a novel bioinformatics tool for Wzy identification and linkage prediction.