Working memory (WM) is a capacity-limited cognitive system that strongly relates to higher-order cognitive abilities including fluid intelligence. It has been suggested that WM training can increase memory capacity, which in turn, improves general intellectual abilities. To evaluate these claims, the present review critically re-assessed nine meta-analysis studies, and revealed that the effect of WM training on fluid intelligence (Gf), executive function, and academic performance is relatively small (averaged Hedges’ adjusted g < .20). Moreover, there were several methodological issues regarding the study design (placebo effect, small sample size), analytical approach (inadequate group comparison, lack of correction for multiple comparisons), and theoretical framework (lack of theoretical account of the training mechanisms) in previous WM training studies. We propose a set of recommendations for future training studies that go beyond training the WM ability per se. This includes theoretically possible methods to enhance intellectual abilities by, for example, learning strategies to effectively encode and recall information into long-term memory.