Tone correction—the development of pitch accuracy, sound quality, and musical expressiveness—is a core challenge in instrumental music education, particularly for novice learners. This study examines the effectiveness of Pitchreader, an AI-based pitch feedback tool, in enhancing tone correction for first-year music college students learning the dulcimer, a traditional Chinese instrument. Twenty students with no prior dulcimer experience were stratified by music theory entrance test scores and then randomly assigned to either an experimental group (n = 10), receiving AI-assisted instruction, or a control group (n = 10), receiving traditional teaching. The 10-week intervention combined weekly classroom instruction with daily self-practice. Post-test assessments included performance of the G major scale (32 notes) and a fast excerpt from Tractor Coming (169 notes), evaluated via blind expert scoring and pitch accuracy analysis. The experimental group showed significantly higher accuracy (84.0% vs. 59.7% for G major; 81.6% vs. 75.4% for Tractor Coming) and expert scores (mean = 79.3 vs. 69.6), with all t-tests yielding p-values below 0.01. Qualitative interviews indicated that Pitchreader improved students’ motivation, confidence, and self-regulated learning. Participants described the real-time visual feedback as intuitive and helpful, and perceived AI technology as compatible with traditional instruction. These findings align with global research on AI-enhanced music education and extend its application to culturally specific settings. This study offers empirical evidence supporting the integration of AI into traditional instrument pedagogy, contributing to international discourse on equitable and innovative music education. It highlights the potential for culturally adapted AI tools to modernize heritage-based instruction and promote effective learning across diverse educational environments.