Background: Warning apps can provide personalized public warnings, but research on their appraisal and impact on compliance is scarce. This study introduces a virtual city framework to examine affective reactions when receiving an app-based warning, and subsequent behavioral intentions. Methods: In an online experiment, 276 participants (M = 41.07, SD = 16.44, 62.0% female) were randomly allocated to one of eight groups (warning vs. no warning, thunderstorm vs. no thunderstorm, video vs. vignette). Participants were guided through a virtual city by a mock-up touristic app (t1). Then, the app issued a warning about an impending thunderstorm (t2), followed by a virtual thunderstorm (t3). The virtual city tour was presented via vignettes or videos. ANCOVAs were used to investigate trajectories of momentary anxiety, hierarchical regressions analyzed the impact of momentary anxiety on information seeking. Results: Participants who received a warning message and were confronted with a thunderstorm showed the highest increase in momentary anxiety, which predicted information seeking intentions. Conclusions: The findings underscore the importance of affective appraisal in processing warning messages. The virtual city framework is able to differentiate the impact of warning versus event in an online context, and thus promising for future warning research in virtual settings.