Molecular characterization of the oral microbiome is a crucial first step in experiments which aim to understand the complex dynamics of the oral microbiome or the interplay with host health and disease. Third-generation Oxford Nanopore Technology (ONT) offers advanced long-read sequencing capabilities, which hold promise for improved molecular characterization by distinguishing closely related microbial species within oral ecosystems in health and disease states. However, the performance of ONT sequencing of oral samples requires validation, and the consistency of this approach across different analytical and sampling conditions is not well understood. This study evaluates various factors that may influence the ONT sequencing outputs of saliva microbiota and compares results with those from Illumina MiSeq's v3v4 amplicon sequencing. Our analysis includes assessments of various stages in the workflow, including different collection and extraction methods, such as robot-extracted saliva DNA used in population-based biobanks, the effects of limited DNA quantities, different bioinformatics pipelines, and different 16S rRNA gene databases. The results demonstrate that ONT provides superior resolution in identifying oral species and subspecies compared to Illumina MiSeq, though the choice of bioinformatics strategy significantly affects the outcomes. Additionally, we confirm the suitability of biobank saliva DNA for large-scale cohort studies, which facilitates the mapping of oral bacterial phylotypes associated with disease states, including less prevalent conditions. Overall, our findings confirm a markedly improved resolution of oral microbiomes by ONT and offer an evidence base to guide the conduct of experiments using this method.