Text-based signage is critical in emergency evacuation scenarios, particularly as built environments become increasingly complex with multiple exits and variable layouts, providing situationally specific and detailed guidance. However, limited research has examined how the modality of initial evacuation instructions influences evacuees’ subsequent interaction with text-based signage. This research addresses this gap by examining both the effect of initial instruction modality on pre-evacuation information-seeking and the impact of modality consistency versus switching on wayfinding performance at decision points. A virtual reality experiment was conducted in a simulated metro station with 60 participants. Results show that participants in the text-based condition took longer to initiate information-seeking but found signage more quickly once seeking began, indicating distinct behavioral patterns shaped by information modality. Moreover, modality consistency reduced re-confirmation behaviors and signage fixation durations, supporting more confident and efficient decision-making. These findings highlight the importance of aligning initial emergency instruction modality with on-site signage formats to reduce cognitive switching costs and enhance evacuation efficiency.