Acute respiratory infections (ARIs) last for less than 30 days and are the most common acute diseases affecting people. Exercise has been shown to improve health generally, but it is uncertain whether exercise may be effective in reducing the occurrence, severity, and duration of ARIs. This is an update of our review published in 2015. To evaluate the effectiveness of exercise for altering the occurrence, severity, or duration of acute respiratory infections. We searched CENTRAL (2020, Issue 2), MEDLINE (1948 to March week 1, 2020), Embase (1974 to 05 March 2020), CINAHL (1981 to 05 March 2020), LILACS (1982 to 05 March 2020), SPORTDiscus (1985 to 05 March 2020), PEDro (searched 05 March 2020), OTseeker (searched 05 March 2020), and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) and ClinicalTrials.gov (searched 05 March 2020). Randomised controlled trials (RCTs) and quasi-RCTs (method of allocation that is not truly random, e.g. based on date of birth, medical record number) of exercise for ARIs in the general population. Two review authors independently extracted data from the included trials using a standard form. One review author entered data, which a second review author checked. We contacted trial authors to request missing data. There were sufficient differences in the populations trialed and in the nature of the interventions to use the random-effects model (which makes fewer assumptions than the fixed-effect model) in the analysis. We included three new trials for this update (473 participants) for a total of 14 trials involving 1377 adults, published between 1990 and 2018. Nine trials were conducted in the USA, and one each in Brazil, Canada, Portugal, Spain, and Turkey. Sample sizes ranged from 16 to 419 participants, aged from 18 to 85 years. The proportion of female participants ranged from 52% to 100%. Follow-up duration ranged from 1 to 36 weeks (median = 12 weeks). Moderate-intensity aerobic exercise (walking, bicycling, treadmill, or a combination) was evaluated in 11 trials, and was most commonly prescribed at least three times a week for 30 to 45 minutes. There was no difference between exercise and no exercise in the number of ARI episodes per person per year (risk ratio (RR) 1.00, 95% confidence interval (CI) 0.77 to 1.30; 4 trials; 514 participants; low-certainty evidence); proportion of participants who experienced at least one ARI over the study period (RR 0.88, 95% CI 0.72 to 1.08; 5 trials; 520 participants; low-certainty evidence); and the number of symptom days per episode of illness (mean difference (MD) -0.44 day, 95% CI -2.33 to 1.46; 6 trials; 557 participants; low-certainty evidence). Exercise reduced the severity of ARI symptoms measured on the Wisconsin Upper Respiratory Symptom Survey (WURSS-24) (MD -103.57, 95% CI -198.28 to -8.87; 2 trials; 373 participants; moderate-certainty evidence) and the number of symptom days during follow-up period (MD -2.24 days, 95% CI -3.50 to -0.98; 4 trials; 483 participants; low-certainty evidence). Excercise did not have a significant effect on laboratory parameters (blood lymphocytes, salivary secretory immunoglobulin, and neutrophils), quality of life outcomes, cost-effectiveness, and exercise-related injuries. There was no difference in participant dropout between the intervention and control groups. Overall, the certainty of the evidence was low, downgraded mainly due to limitations in study design and implementation, imprecision, and inconsistency. Seven trials were funded by public agencies; five trials did not report funding; and two trials were funded by private companies. Exercise did not reduce the number of ARI episodes, proportion of participants experiencing at least one ARI during the study, or the number of symptom days per episode of illness. However, exercise reduced the severity of ARI symptoms (two studies) and the number of symptom days during the study follow-up period (four studies). Small study size, risk of bias, and heterogeneity in the populations studied contributed to the uncertainty of the findings. Larger trials that are designed to avoid risk of bias associated with participant selection, blinding of outcomes assessors, and with adequate reporting of all outcomes proposed for measurement in trials, would help to provide more robust evidence.