Background: As one of the mainstream forms of programming education, educational robotics (ER) have been a crucial way to develop K-12 students' programming ability. Objectives: The purpose of this study is to clarify the content of programming ability, to verify the effectiveness of ER as a teaching method to improve students' programming abilities and the conditions for more effective instructional design. Methods: The method of meta-analysis was adopted to conduct this study. 4180 K-12 students from 36 studies with a total of 85 effect sizes were captured. Based on the contents of included literature, an ER programming ability model was put forward, which was composed of essential ability and generative ability. The meta-analysis of essential ability (N = 22) and generative ability (N = 64) was carried out respectively. Results and conclusion: The results showed that ER activity has a positive effect on the essential ability (Hedges' g = 0.539 CI [0.327, 0.752], p < 0.001) and generative ability (Hedges' g = 0.535 CI [0.426, 0.643], p < 0.001). Additionally, the results of moderator analysis showed that utilizing ER to solve mathematics problems is able to more effectively develop students' essential ability, while the STEM interdisciplinary ER instruction had the largest effect on generative ability. Simultaneously, the individual programming form is conducive to students' essential ability, while the generative ability is significantly affected by the cooperative programming form. No matter for the evaluation of essential ability or generative ability, ER had the greatest effect on kindergarten children' programming ability. Moreover, the intervention duration ranging from 1 to 5 weeks, the samples size less than 50 and the choice of measurement scale can effectively improve students' programming ability. Implications: This research enriched the theoretical basis of programming ability, and provided reference and guidance for K-12 programming teaching practitioners and researchers in ER teaching design and practise.