Bronchoalveolar lavage (BAL) is a major diagnostic tool in lung diseases, including viral respiratory infections. We aimed to better define the situations where viral tests should be performed on BAL fluid (BALF). We retrospectively studied all cases where viral tests [immunofluorescence, immunocytochemistry, viral culture, and/or polymerase chain reaction (PCR)] were performed on BALF during a period of 1 year (2008) in our institution. We compared the characteristics of patients with virus-positive versus virus-negative BALF. Of the 636 BALF samples sent to the microbiology laboratory, 232 underwent viral tests. Of these, 70 (30 %) were positive and identified 85 viruses: herpes simplex virus (HSV)-1 (n = 27), cytomegalovirus (CMV, n = 23), Epstein-Barr virus (EBV, n = 18), human herpesvirus (HHV)-6 (n = 12), respiratory syncytial virus (RSV, n = 3), rhinovirus (n = 1), and adenovirus (n = 1). The variables associated with positive viral tests on univariate analysis were immunosuppression [human immunodeficiency virus (HIV), corticosteroids >10 mg/day for ≥3 weeks, or other immunosuppressive therapy], ground-glass attenuations on computed tomography (CT) scanning, late-onset ventilator-associated pneumonia (VAP), and durations of (i) hospital stay, (ii) intensive care unit (ICU) stay, and (iii) mechanical ventilation before BAL (p < 0.01 for each comparison). On multivariate analysis, only immunosuppression [odds ratio (OR) 6.4, 95 % confidence interval (CI) [2.8-14.3], p < 0.0001] and ground-glass attenuations (OR 3.7, 95 % CI [1.8-7.7], p = 0.0004) remained associated with virus-positive BAL. None of the viral tests performed on BALF for the initial assessment of diffuse infiltrative lung disease (n = 15) was positive. PCR improved the diagnostic yield of viral tests on BALF by 50 %. Testing for viruses on BALF should be mostly restricted to immunocompromised patients with acute respiratory diseases and/or patients with unexplained ground-glass attenuations on CT scanning.