Music students welcomed AI features but still wanted a human teacher

The tool moment

A study published 10 October 2025 in Frontiers in Education, titled “How anthropomorphic AI features affect music students' acceptance: a study among Chinese undergraduates,” examined how music students respond to AI tools built with human-like interaction cues, such as voice interaction and conversational feedback, inside their coursework. The authors surveyed 600 Chinese undergraduate music students recruited through stratified random sampling by academic year, sex and prior exposure to AI-generated content, and separately interviewed 33 to 35 students across five universities.

What the documents show

The study reports that students rated voice interaction (mean 4.7) and immediate feedback (mean 4.6) as highly useful, and that “anthropomorphic features were especially helpful for learning music.” It reports the reverse alongside that finding: students expressed discomfort with some human-like qualities and, per the paper, underscored “the irreplaceable role of human expressiveness in music pedagogy,” describing AI as lacking emotional authenticity compared with a human teacher. Students reported combining AI tools with human feedback from teachers, mentors and peers rather than substituting one for the other. A publication record confirms the journal and date.

What stays with the musician

The authors state their own limits directly: the findings rest on “self-reported, cross-sectional data from a single group of Chinese undergraduate music students,” which the paper says restricts causal inference and generalizability. A student, teacher or tool-builder elsewhere cannot assume the same usefulness ratings or comfort with anthropomorphic features would appear in a different education system, age group or musical tradition. What the study does establish is that, within this sample, usefulness and continued reliance on a human teacher were not treated as opposites by the students themselves.

Judge it by listening

This is a survey and interview study, not a listening test, so it offers no method for judging AI-generated audio by ear. What it does offer a reader is a template for asking the right question about any AI music-education tool: does it replace feedback from a teacher, mentor or peer, or does it sit alongside that feedback the way the surveyed students described using it. That framing, not any audio comparison, is what this particular study can responsibly contribute.

• Would this same balance between AI usefulness and reliance on a human teacher appear outside this sample's education system and age group?
• Is an AI tool being used here to replace a teacher's feedback, or to supplement it?
• Which specific anthropomorphic feature, voice interaction or feedback timing, is doing the work the study measured?

A single-country, self-reported sample of undergraduates is real evidence about that sample, not a global verdict on AI in music education. The paper's own caveats about generalizability are part of what it establishes, not a footnote to discard.