An ultralow-level liquid scintillation counting (LSC) technique has been used in measuring radiosulfur (cosmogenic 35S) in natural samples. The ideal half-life of 35S (∼87 days) renders it a new way to examine various biogeochemical problems. A major limitation of the technique is that complex chemical compositions in atmospheric samples may lead to color quenching of LSC cocktails, a serious problem prolonging the pretreatment time (>1 week) and hampering the accurate determination of 35S. For application of the technique where many of the most important atmospheric chemical processes are examined, significant interferences arise and accurate analysis in small samples is not possible. In this study, we optimized the LSC method to minimize/eliminate color quenching in high-sensitivity 35S measurements. The analytical performance of this new method was evaluated using control laboratory experiments and natural aerosol samples. Results show that the new method offers comparable accuracy as the traditional method for normal environmental samples [bias: <±0.03 disintegrations per minute (DPM)] and significantly shortens the pretreatment time to less than 3 days. For samples that were heavily contaminated by color-quenching agents, the accuracy of this new method is notably higher than that of the traditional method (maximum bias: -0.3 vs -1.5 DPM). With the growing use of radiosulfur in the field of Earth and planetary sciences, the accurate determination of 35S would provide a reliable field-based constraint for modeling 35S production in the atmosphere and allow a wide range of atmospheric, hydrological, and biogeochemical applications.