Next-generation sequencing (NGS) opens up exciting possibilities for improving our knowledge of environmental microbial diversity, allowing rapid and cost-effective identification of both cultivated and uncultivated microorganisms. However, library preparation, sequencing, and analysis of the results can provide inaccurate representations of the studied community compositions. Therefore, all these steps need to be taken into account carefully. Here we evaluated the effects of DNA extraction methods, targeted 16S rRNA hypervariable regions, and sample origins on the diverse microbes detected by 454 pyrosequencing in marine cold seep and hydrothermal vent sediments. To assign the reads with enough taxonomic precision, we built a database with about 2,500 sequences from Archaea and Bacteria from deep-sea marine sediments, affiliated according to reference publications in the field. Thanks to statistical and diversity analyses as well as inference of operational taxonomic unit (OTU) networks, we show that (i) while DNA extraction methods do not seem to affect the results for some samples, they can lead to dramatic changes for others; and (ii) the choice of amplification and sequencing primers also considerably affects the microbial community detected in the samples. Thereby, very different proportions of pyrosequencing reads were obtained for some microbial lineages, such as the archaeal ANME-1, ANME-2c, and MBG-D and deltaproteobacterial subgroups. This work clearly indicates that the results from sequencing-based analyses, such as pyrosequencing, should be interpreted very carefully. Therefore, the combination of NGS with complementary approaches, such as fluorescence in situ hybridization (FISH)/catalyzed reporter deposition (CARD)-FISH or quantitative PCR (Q-PCR), would be desirable to gain a more comprehensive picture of environmental microbial communities.