This systematic review and Bayesian meta-analysis examined determinants of exercise effects on subjective sleep quality.

We searched PubMed, Embase, Cochrane Library, Web of Science, Scopus and CINAHL from inception to August 2025 for randomised controlled trials in adults (≥18 years) comparing exercise with non-exercise controls. Two reviewers independently assessed risk of bias (RoB 2). Bayesian multilevel models, MetaForest and structural equation modelling estimated pooled effects and moderators, expressed as standardised mean differences (Hedges’ g).

Two hundred trials (n = 23,523) were included. Compared with active controls, qigong, walking and high-intensity interval training showed the largest benefits. In 20 trials reporting objective sleep efficiency, exercise produced a small but significant improvement, consistent with subjective outcomes. Baseline sleep quality (assessed with Pittsburgh Sleep Quality Index) and population type were the main determinants of response. Populations with poorer baseline sleep quality were much more likely to achieve a clinically important improvement than good sleepers. Intervention duration showed a U-shaped association with benefit, peaking at 25 weeks. Certainty of evidence was very low.

Exercise improves subjective sleep quality in a benefit-on-demand pattern, supporting prioritisation of approximately 25-week programmes for adults with moderate-to-severe sleep problems while acknowledging very low certainty.