Culture-independent methods have granted the possibility to study microbial diversity in great detail, but technical issues pose a threat to the accuracy of new findings. Biases introduced during DNA extraction can result in erroneous representations of the microbial community, particularly in samples with low microbial biomass. We evaluated the DNA extraction method, initial sample biomass, and reagent contamination on the assessment of the human gut microbiota. Fecal samples of 200 mg were subjected to 1:10 serial dilutions; total DNA was obtained using two commercial kits and the microbiota assessed by 16S ribosomal RNA (rRNA) gene sequencing. In addition, we sequenced multiple technical controls. The two kits were efficient in extracting DNA from samples with as low as 2 mg of feces. However, in instances of lower biomass, only one kit performed well. The number of reads from negative controls was negligible. Both DNA extraction kits allowed inferring microbial consortia with similar membership but different abundances. Furthermore, we found differences in the taxonomic profile of the microbial community. Unexpectedly, the effect of sample dilution was moderate and did not introduce severe bias into the microbial inference. Indeed, the microbiota inferred from fecal samples was distinguishable from that of negative controls. In most cases, samples as low as 2 mg did not result in a dissimilar representation of the microbial community compared with the undiluted sample. Our results indicate that the gut microbiota inference is not much affected by contamination with laboratory reagents but largely impacted by the protocol to extract DNA.