Background <p>Prolonged sitting has been associated with adverse effects on cognitive function, and interrupting sitting with physical activity (PA) may help mitigate these effects. However, evidence regarding the effects of PA breaks on executive function remains limited. In addition, it remains unclear how intervention characteristics may be associated with variations in effect estimates. This study aims to synthesize evidence from randomized trials on the effects of PA breaks on executive function, and to identify potentially effective intervention characteristics through moderator analysis.</p> Methods <p>A systematic search was conducted in Web of Science, PubMed, Embase, Scopus, Cochrane Library, APA PsycINFO, and SPORTDiscus from database inception to April 25, 2026, for randomized trials investigating the effects of PA breaks during sedentary time on executive function. Although the eligibility criteria allowed randomized controlled trials, randomized crossover trials, and cluster-randomized trials, all studies included in the final synthesis used randomized crossover designs. Risk of bias was assessed using the Cochrane Risk of Bias Tool version 2 for crossover trials, and the quality of evidence was evaluated using the Grading of Recommendations, Assessment, Development, and Evaluations framework. A random-effects three-level meta-analytic model was applied to analyze main effects, exploratory moderator effects, publication bias, and sensitivity analyses.</p> Results <p>Twenty-two randomized crossover trials were included, comprising 123 effect sizes from 418 participants. Low certainty evidence suggests that, compared with uninterrupted sitting, PA breaks may be associated with a small improvement in executive function (Hedges’ <i>g</i> = 0.13, 95% CI [0.03, 0.23], <i>p</i> = 0.01), particularly working memory (Hedges’ <i>g</i> = 0.17, <i>p</i> = 0.02). Regrading performance metrics, reaction time (RT) performance (Hedges’ <i>g</i> = 0.18, <i>p</i> &lt; 0.01) showed a significant positive effect, whereas accuracy showed no significant effect. Exploratory moderator analyses provided preliminary evidence that intervention characteristics may be related to variations in executive function outcomes. Light-intensity activities showed a favorable pooled estimate (Hedges’ <i>g</i> = 0.19, <i>p</i> &lt; 0.01), while moderate-intensity activities did not show a significant pooled effect. For activity type, walking (Hedges’ <i>g</i> = 0.12, <i>p</i> = 0.03) and stair climbing (Hedges’ <i>g</i> = 0.70, <i>p</i> = 0.01) showed significant pooled effects, although the estimate for stair climbing was based on a small number of studies. Regarding frequency, 60-minute interruption intervals showed the largest pooled estimate (Hedges’ <i>g</i> = 0.30, <i>p</i> = 0.01). Break duration was not a statistically significant moderator (<i>p</i> = 0.06), although using 3–5&#xa0;min breaks showed significant pooled effect (Hedges’ g = 0.16, <i>p</i> = 0.02).</p> Conclusions <p>Interrupting prolonged sitting with PA is associated with small acute improvements in executive function, especially in the working memory and RT performance. Moderator analyses suggested that intervention characteristics such as activity intensity, type, and frequency may contribute to heterogeneity in effects. While the certainty of evidence is low and some subgroup estimates are based on a limited number of studies, these findings provide preliminary guidance for optimizing intervention design and highlight the need for further high-quality research.</p> Trial registration <p>CRD420251004155.</p> Graphical Abstract <p></p>