Objective: To investigate the impact of exercise therapy on molecular biomarkers related to cartilage and inflammation in individuals at risk of, or with established, knee osteoarthritis by conducting a systematic review of randomized controlled trials (RCTs). Methods: We conducted a literature search up to September 2017 in 5 major databases with no restriction on publication year or language. Data were extracted from the first available follow-up time point, and we performed a narrative synthesis for the effect of exercise therapy on molecular biomarkers related to cartilage and inflammation. A subset of studies reporting sufficient data was combined in a meta-analysis, using an adjusted random-effects model. Results: Twelve RCTs involving 57 study comparisons at 4 to 24 weeks following an exercise-therapy intervention were included. Exercise therapy decreased molecular biomarkers in 17 study comparisons (30%), had no effect in 36 (63%), and increased molecular biomarkers in 4 study comparisons (7%). Meta-analyses of 9 biomarkers showed that exercise therapy was associated with nonsignificant reductions of the C-reactive protein level, C-terminal crosslinking telopeptide of type II collagen, tumor necrosis factor (TNF), soluble TNF receptors 1 and 2, C2C neoepitope of type II collagen, and cartilage oligomeric matrix protein, compared to nonexercising control groups, and exercise therapy had no effect on interleukin-6 and soluble interleukin-6 receptor. Conclusion: Exercise therapy is not harmful, because it does not increase the concentration of molecular biomarkers related to cartilage turnover and inflammation, implicated in osteoarthritis progression. The overall quality of evidence was downgraded to low because of the limited number of RCTs available.