Background: The COVID pandemic exposed limitations in online learning platforms. This study explores Real-time Shared Gaze Visualizations (SGVs), which use eye-tracking data to restore non-verbal social cues and improve teacher-learner communication. We aim to: (1) identify gaze patterns linked to learning, (2) compare these patterns across visual augmentation conditions, and (3) examine their impact from a teaching perspective. Methods: We analyze eye-tracking data from 75 learners completing microcontroller-based tasks with instructor support. The activity included an instruction phase and a learner-led practice phase. Both quantitative and qualitative analyses were used to examine learning differences across conditions. Findings: Joint visual attention correlated with learning gains during instruction and with instructors' predictions of learner success in practice. Video analysis revealed three key affordances of SGVs: (1) tracking learners' cognitive states, (2) using gaze as deictic gestures, and (3) providing just-in-time support. Contribution: SGVs can enhance remote one-on-one teaching by compensating for missing nonverbal cues. As eye-tracking technology becomes more accessible, SGVs may improve online learning, especially for STEM subjects involving physical objects. Future research should explore their role in learnerled, tangible computing tasks.