This paper explores how qualitative eye-tracking data can reveal usability issues in educational applications. We studied OrChemSTAR, a multimodal iPad app for chemistry learning that combines handwriting recognition, adaptive feedback, and augmented reality (AR). Using a small-scale qualitative study with eye tracking glasses, we analyzed how students interact with three core learning modes: scanning, AR exploration, and guided practice. Rather than relying on quantitative gaze metrics, we used eye tracking as a tool to reveal patterns of confusion, hesitation, and misalignment between user expectations and system responses. Gaze recordings were paired with observational annotations and user feedback to identify friction points, including misinterpreted feedback icons, interface misalignments, and unmet gesture expectations. Our findings demonstrate that even simple eye-tracking studies can expose critical micro-interactions that impact usability, user experience, and cognitive load. We argue that qualitative eye tracking is a valuable addition to UX methodologies in early-stage educational technologies, especially those involving AR, gesture input, or adaptive interfaces. The article concludes with design implications for the development of learning applications.