Abstract Introduction and Aims People who inject drugs (PWID) are exposed to associated viral, bacterial and fungal risks. These risks can be reduced by filtration. Large disparities in the quality of filtration exist between the various available filters. This paper compares both performance and user acceptability of three filters for drug injection (cotton filters, Sterifilt ® and wheel filters) by combining epidemiological and bacteriological analyses. Design and Methods A cross‐sectional epidemiological study (ANRS‐Coquelicot) using time–location sampling combined with the generalised weight sampling method was conducted among 985 PWID in France. Two filtration‐based bacteriological studies of 0.20‐ and 0.45‐μm wheel filters, Sterifilt filters and cotton filters were also conducted. Results The bacteriological study highlighted the value of using wheel filters with a porosity of less than 0.5 μm, as they limit the risk of bacterial and fungal infection. The results of this study clearly highlight a distinction between the efficiency of Sterifilt and wheel filters, the latter being more effective. Our epidemiological study highlighted that the use of cotton filters is widespread and routine, but is the subject of much criticism among PWID. Sterifilt is not widely used, and its adoption is slow. Finally, the wheel filter remains a largely untested tool. Discussion and Conclusions Low product retention and ease of use are the two most important factors for filters for PWID. Bacterial and fungal risk filtration is less important. It is essential to educate PWID about the benefits of wheel filters. [Jauffret‐Roustide M, Chollet A, Santos A, Benoit T, Péchiné S, Duplessy C, Bara J‐L, Lévi Y, Karolak S, Néfau T. Theory versus practice, bacteriological efficiency versus personal habits: A bacteriological and user acceptability evaluation of filtering tools for people who inject drugs. Drug Alcohol Rev 2018;37:106–115]