Language control is needed to regulate the activation of known languages. One can distinguish between two types of language control: reactive and proactive. In contrast to the better understood reactive control, proactive language control is thought to be used to adapt language activation in advance—before conflict has arisen. Proactive language control can be investigated by looking at cognate production in different task settings: Cognates are translation-equivalent words that are similar or even identical across languages in phonological and/or orthographic form (e.g., BABY/BABY [English/German]). Due to cognates’ high form overlap, language co-activation results in low conflict when encountering them. Previous research has shown that pictures of cognates are more easily named (i.e., fewer errors, smaller RTs) by bilinguals than non-cognates. However, such cognate facilitation may depend on the exact task demands and language control state a bilingual is in. Across two experiments, we thus investigated how proactive language control modulates language co-activation and in turn cognate facilitation in situations of low and high conflict. In Experiment 1 (N = 85, preregistered), we manipulated the frequency of cognates versus non-cognates within blocks of a L2 written picture naming task. As hypothesized, we found a larger cognate facilitation effect in blocks with more cognates (majority-cognate; low conflict) than in blocks with more non-cognates (majority-non-cognates; high conflict), suggesting that people proactively adapt language activation to optimize ease of language production. In Experiment 2 (planned N = 80), we plan to investigate the exact mechanism by which co-activation of translation-equivalent words is regulated.