Ongoing improvements to metagenomic sequencing technologies and analysis approaches have enhanced our ability to study microbial communities in environments with low microbial biomass. However, these advancements have also increased the sensitivity to contaminating DNA that is introduced to the sample from the environment (during sampling, sample handling and processing), contaminated reagents and cross-contamination between samples. Contamination in low-biomass metagenomes is a widely known issue, yet in many areas of research it continues to be inadequately prevented and reported. Here I will share some insights from our work on microbial life in the atmosphere: the most substantial but least explored microbial ecosystem, where advances in knowledge have been hindered by technological limitations and the high risk of sample contamination. I will discuss our approach to minimising contamination during sampling and processing, identifying and removing contamination from metagenomic datasets, and interpreting data in light of possible remaining contaminants. As microbial ecology research extends further into the unknown and the extreme, implementing adequate contamination controls and adopting a sceptical and meticulous approach to data interpretation are critical when working at the limits of detection. Wider awareness and implementation of best practices for low biomass environments will assist in removing barriers to scientific progress in a range of environmentally and medically relevant fields, including atmospheric, glacial, deep subsurface, placental and brain microbiology.