Biorecognition element (BRE)-based carbon nanotube (CNT) chemiresistors have tremendous potential to serve as highly sensitive, selective, and power-efficient volatile organic compound (VOC) sensors. While many research groups have studied BRE-functionalized CNTs in material science and device development, little attention has been paid to optimizing CNT density to improve chemiresistor performance. To probe the effect of CNT density on VOC detection, we present the chemiresistor-based sensing results from two peptide-based CNT devices counting more than 60 different individual measurements. We find that a lower CNT density shows a significantly higher noise level and device-to-device variation while exhibiting mildly better sensitivity. Further investigation with SEM images suggests that moderately high CNT density with a stable connection of the nanotube network is desirable to achieve the best signal-to-noise ratio. Our results show an essential design guideline for tuning the nanotube density to provide sensitive and stable chemiresistors.