AimsSuboptimal device programming is frequent in non‐responders to cardiac resynchronization therapy (CRT). However, the role of device optimization and the most appropriate technique are still unknown. The aim of our study was to analyse the effect of different CRT optimization techniques within a network meta‐analysis.MethodsA systematic search was conducted on MEDLINE, Embase and CENTRAL for studies comparing outcomes with empirical device settings or optimization using echocardiography, static algorithms or dynamic algorithms. Studies investigating the effect of optimization in non‐responders were also analysed.ResultsA total of 17 studies with 4346 patients were included in the quantitative analysis. Of the treatments and outcomes examined, a significant difference was found only between dynamic algorithms and echocardiography, with the former leading to a higher echocardiographic response rate [odds ratio (OR): 2.02, 95% confidence interval (CI) 1.21–3.35], lower heart failure hospitalization rate (OR: 0.75, 95% CI 0.57–0.99) and greater improvement in 6‐minute walk test [mean difference (MD): 45.52 m, 95% credible interval (CrI) 3.91–82.44 m]. We found no significant difference between empirical settings, static algorithms and dynamic algorithms. Seven studies with 228 patients reported response rates after optimization in non‐responders. Altogether, 34.3%–66.7% of initial non‐responders showed improvement after optimization, depending on response criteria.ConclusionsAt the time of CRT implantation, dynamic algorithms may serve as a resource‐friendly alternative to echocardiographic optimization, with similar or better mid‐term outcomes. However, their superiority over empirical device settings needs to be investigated in further trials. For non‐responders, CRT optimization should be considered, as the majority of patients experience improvement.