Mobile apps are becoming part and parcel of our daily lives. Hence, this study examined the differential effects of app modes on listening comprehension and recognition. From a pool of Egyptian EFL sophomores, 107 students were randomly assigned into 3 groups practicing mobile-assisted listening in 3 modalities: Unimodal (n = 35), Bimodal (n = 35), and Multimodal (n = 37). Following sequential explanatory mixed-method design, scores on pre-post listening tests and responses to closed/open-ended perceptions survey questions were analyzed using Two-way Mixed ANOVA, Linear Regression, and inductive thematic analysis. While all three modalities demonstrated comparable effectiveness in improving listening comprehension, performance variations were observed in listening recognition. Both the uni- and multimodal groups surpassed the bimodal group in listening recognition, while exhibiting similar levels of listening recognition improvement. As such, apps focusing on comprehension exercises demonstrated greater efficacy in triggering lower-level cognitive processes (listening recognition) than those exclusively targeting word recognition. Further, listening recognition was found to be a weak predictor of listening comprehension. Participants' perceptions consolidated these propositions. They offered some app functionality and user-experience enhancement suggestions. They advocated for incorporating more personalized, gamified, and interactive tools (e.g., shadowing, translation, and speed control tools), coupled with increased content and exercise variety (stepping beyond multiple-choice and fill-in-the-blank formats).