The prevalence and long-term consequences of dyslexia make it crucial to look for effective and efficient ways of its therapy. Action video games (AVG) were implied as a possible remedy for difficulties in reading in Italian and English-speaking children. However, the studies examining the effectiveness of AVG application in dyslexia suffered from significant methodological weaknesses such as small sample sizes and lack of a control group with no intervention. In our study, we tested how two forms of training: based on AVG and on phonological non-action video games (PNAVG), affect reading in a group of fifty-four Polish children with dyslexia. Both speed and accuracy of reading increased in AVG as much as in PNAVG group. Moreover, both groups improved in phonological awareness, selective attention and rapid naming. Critically, the reading progress in the two groups did not differ from a dyslexic control group which did not participate in any training. Thus, the observed improvement in reading in AVG and PNAVG can be attributed either to the normal reading development related to schooling or to test practice effect. Overall, we failed to replicate previous studies: Neither AVG nor PNAVG remedy difficulties in reading in school children.