We provide a theoretical framework and empirical evidence for how verbally planning an action creates direct perception-action links and behavioral automaticity. We argue that planning actions in an if (situation)–then (action) format induces sensorimotor simulations (i.e., activity patterns reenacting the event in the sensory and motor brain areas) of the anticipated situation and the intended action. Due to their temporal overlap, these activity patterns become linked. Whenever the previously simulated situation is encountered, the previously simulated action is partially reactivated through spreading activation and thus more likely to be executed. In 4 experiments (N=363), we investigated the relation between specific if–then action plans worded to activate simulations of elbow flexion versus extension movements and actual elbow flexion versus extension movements in a subsequent, ostensibly unrelated categorization task. As expected, linking a critical stimulus to intended actions that implied elbow flexion movements (e.g., grabbing it for consumption) subsequently facilitated elbow flexion movements upon encountering the critical stimulus. However, linking a critical stimulus to actions that implied elbow extension movements (e.g., pointing at it) subsequently facilitated elbow extension movements upon encountering the critical stimulus. Thus, minor differences (i.e., exchanging the words “point at” with “grab”) in verbally formulated action plans (i.e., conscious thought) had systematic consequences on subsequent actions. The question of how conscious thought can induce stimulus-triggered action is illuminated by the provided theoretical framework and the respective empirical evidence, facilitating the understanding of behavioral automaticity and human agency.