Prior meta-analyses demonstrate that people can intentionally distort Big Five personality scores when instructed. As yet, there is no equivalent meta-analysis addressing instructed faking on the dark triad (narcissism, Machiavellianism, and psychopathy). Therefore, we review mean score changes to the dark triad domains and facets under instructed faking. Due to insufficient k for meta-analysis, narcissism and Machiavellianism were systematically reviewed alongside psychopathy. The systematic review revealed inconsistent findings for narcissism and Machiavellianism with several effects in the opposite direction than expected. The psychopathy meta-analysis showed that: (a) scores were significantly lower under fake good compared to answer honestly instructions (d = −0.40); and (b) scores were significantly higher under fake bad compared to answer honestly instructions (d = 1.88). Subgroup analyses revealed significant score decreases under fake good instructions for both primary (d = −0.56) and secondary psychopathy (d = −0.96), and a significant score increase under fake bad instructions for primary (d = 1.69) and secondary psychopathy (d = 1.50). We conclude that dark triad measures are fakeable to a similar extent as the Big Five, and discuss the relevance of our findings for dark triad assessment in several applied contexts.