A definitive diagnosis of tuberculosis serositis (TS) is still challenging. Our preliminary practice found that Serous Fluid Drainage Flocky Precipitate (SFDFP) was a useful testing sample to diagnose TS. We designed this study to assess the diagnostic performance of SFDPF for TS compared with conventional bacteriology methods on serous fluid (SF). A cohort study was conducted from July 2014 to April 2016. Patients with suspected TS were consecutively screened. SF and SFDFP were collected and tested by Ziehl-Neelsen stain, MTB culture, and Xpert/RIF assay. We compared the diagnostic performance of SF and SFDFP in several test settings. Through this study, 85 patients were enrolled, of whom 70 (82.4%) were confirmed TS or highly probable TS, 13 (15.3%) were none-TS and 2 (2.4%) indeterminate results were ruled out. The overall sensitivity using both SFDFP and SF was significantly higher than each (60% vs. 48% and 41%, p < 0.05). SFDFP and SF samples had similar diagnostic performance (p < 0.05). No false positive was detected in this study. We concluded that SFDFP is a reliable testing sample for diagnosing tuberculous serositis. SFDFP may significantly improve the diagnostic yield as a supplement to conventional tests.