Background: Microbial ecology is reshaping our understanding of the natural world by revealing the large phylogenetic and functional diversity of microbial life. However the vast majority of these microorganisms remain poorly understood, as most cultivated representatives belong to just four phylogenetic groups and more than half of all identified phyla remain uncultivated. Characterization of this microbial ‘dark matter’ will thus greatly benefit from new metagenomic methods for in situ analysis. For example, sensitive high throughput methods for the characterization of community composition and structure from the sequencing of conserved marker genes. Methods: Here we utilize Single Molecule Real-Time (SMRT) sequencing of full-length 16S rRNA amplicons to phylogenetically profile microbial communities to below the genus-level. We test this method on a mock community of known composition, as well as a previously studied microbial community from a lake known to predominantly contain poorly characterized phyla. These results are compared to traditional 16S tag sequencing from short-read technologies and subsets of the full-length data corresponding to the same regions of the 16S gene. Results: We explore the benefits of using full-length amplicons for estimating community structure and diversity. In addition, we investigate the possible effects of context-specific and GC-content biases known to affect short-read sequencing technologies on the predicted community structure. We characterize the potential benefits of profiling metagenomic communities with full-length 16S rRNA genes from SMRT sequencing relative to standard methods.