<p>Mathematical knowledge can be represented by various external representations, such as algebraic expressions and images. While the potential benefits of multiple representations are worth exploring, the appropriate selection and design of these representations should also be addressed. We conducted a systematic review of 17 studies and a three-level meta-analysis based on 106 effect sizes derived from 15 studies to evaluate the overall effectiveness of multiple representations in mathematics achievement and how their design and effectiveness vary across mathematical domains. Additionally, we systematically coded and quantified the characteristics of the included studies, including their major themes and descriptions of their representation designs. Results showed: (1) Compared with single representations, multiple representations had a significant positive effect on students’ mathematics achievement (Hedges' <i>g</i> = 0.286, <i>p</i> &lt; .05, 95% CI [0.069, 0.502]). The additional types introduced in the experimental group significantly moderated the effects. (2) There was substantial heterogeneity across the included studies. The design and effectiveness of multiple representations differed in magnitude across various mathematical domains. More fine-grained categorization is needed in future research to draw more specific conclusions. (3) More in-depth and detailed research is needed to address the gap in current studies, which pay insufficient attention to integrating specific knowledge content when analyzing representation effects and often provide limited detail on representation design.</p>