Music teachers on AI: mostly lesson prep, rarely in the classroom

Three studies published this year asked music teachers what they actually do with AI. They come from different countries and use different methods, and they land in roughly the same place. AI is doing the paperwork and the planning, and teachers are keeping it away from the parts of the job that make music teaching what it is.

None of the three is large. Read together, they are still the clearest picture so far of how the profession is handling AI. And the pattern holds a lesson for anyone who teaches music for a living, in a school or from a home studio.

What did the three studies find?

Indiana, US: a survey of 54 teachers. Hawon Song and Rachel Louise Geesa surveyed PK-12 music educators in Indiana in April 2026. The survey went to 557 teachers, and 54 took part. Asked directly whether they had used AI in their music teaching, 35% said yes (17 people) and 65% said no (31). Of the 17 users, 82% used it for lesson planning or making materials and 82% for admin. Fewer used it in front of students: 47% for in-class activities, 35% for assessment, 29% for feedback on pitch, rhythm or tone, and 24% for composition or improvisation.

One wrinkle. The same paper also says that only 23% of respondents reported using AI-based tools in their current teaching. The two figures come from differently worded questions, and the paper doesn't reconcile them. So the honest summary is "roughly a quarter to a third", not one tidy number.

The paper appears in the National High School Journal of Science, which describes itself as a peer-reviewed journal aimed at high-school research. The authors flag their own limits: one Midwest state, a sample weighted toward elementary and suburban teachers, and a survey instrument they built themselves.

Australia and Italy: interviews with ten teachers. Michele Biasutti (Padova), Bradley Merrick (Melbourne) and Dawn Joseph (Deakin) interviewed ten experienced music teachers for Music Education Research, published online on 26 May 2026. The publisher's site blocked our checker, so we read the open-access copy in Deakin's repository. Teachers said AI's greatest value was cutting admin and speeding up preparation: reports, emails, rubrics, lesson plans. One said it let them do "in minutes what used to take hours".

Creative use showed up too. Teachers used AI for chord progressions, melodic ideas, sample lyrics and backing tracks, and described it as a "third ear": a supplementary voice that offers options, not the composer. The authors are careful to say these creative uses weren't reported by everyone and were often exploratory.

China: a survey of 352 teachers and 17 interviews. Yijie Zhang and Shuhui Niu studied instrumental teachers in Frontiers in Psychology on 8 July 2026. The strongest predictor of wanting to use AI was simply expecting it to help. Two beliefs pulled the other way: that AI threatens a teacher's artistry, and that hands-on, physical teaching can't be replaced. The second pulled harder among teachers with more than ten years' experience and among teachers of instruments other than piano. In interviews, teachers accepted AI for drilling intonation and rhythm after lessons, while one called that "only a small part of what teaching involves".

What worries teachers?

Authorship comes first. In Indiana, agreement that teachers had ethical concerns about AI, including authorship and originality, averaged 3.98 out of 5. Three in four (75%) worried AI could reduce students' creativity or musicianship. Among non-users, the top reason for staying away was concern about appropriateness or ethics (81%), well ahead of lack of training (55%). In the Australia and Italy interviews, teachers worried about students presenting AI-generated compositions as their own. One asked, "who really made the album?"

Reliability and taste. Teachers in the same study reported hallucinations (one AI system described a teacher as a gymnast), worried about a "flattening of taste", and saw students expecting AI to answer even simple tasks.

Confidence and guidance. Indiana teachers were confident with technology in general (3.92 out of 5) but much less so at evaluating AI-generated content (2.44) and guiding students' ethical use (2.40). The Australia and Italy teachers described vague or absent school policies. The China study summed up its interviews as "ample encouragement but insufficient guidance".

There is also a cost that rarely makes the sales pitch. The Australia and Italy authors describe the hidden work of checking AI output for musical correctness, bias, age-appropriateness and stylistic fit.

What does good practice look like?

What follows is our synthesis of what these teachers did and asked for, not a protocol from any one paper.

1. Start behind the desk. Use AI for drafts of reports, rubrics, lesson outlines and parent emails. Keep the final word yourself. As one Italian teacher put it, AI "can screen the first layer, but the final review must be mine".
2. Check everything before a student sees it. The teachers reported real inaccuracies and hallucinations, and checking output for musical correctness is part of the hidden workload they describe.
3. Use it as a third ear, not the composer. Generate three chord progressions or a rough backing track, then have students change, reject or argue with them. The learning is in the editing.
4. Keep drilling tools for drilling. Pitch and rhythm apps help with practice between lessons. Interpretation, posture and tone stay with you.
5. Move assessment toward process. The teachers shifted to oral and practical tasks, in-class composition, improvisation, and short commentaries in which students note how AI contributed.
6. Make declaring AI use normal. Ask for a one-line note on what tool was used and for what. It's the same habit the industry is building with AI credits on releases.
7. Talk about where ideas come from, including how AI systems are trained, as the Australia and Italy teachers recommend.

A quick checklist before you bring an AI tool into a lesson:

• I know what it's for: prep, practice, or creative prompting.
• I've checked its output for musical and factual errors.
• Students know they must say when and how they used it.
• The assessed work includes a process step AI can't fake: a performance, an oral explanation or a live edit.
• I know my school's policy, or I've asked for one in writing.

What we think

Prep-first adoption isn't timid. It's the rational order. The tasks AI takes on first, such as report boilerplate and rubric formatting, are the ones that make nobody a better musician. Freeing hours from them is a straightforward win, as long as the hours go back to students rather than to more admin.

That's where the risk sits, in our view. The weakest number in the Indiana survey is confidence in evaluating AI output, at 2.44 out of 5, and the teachers asked first for practical examples (74%) and time to explore (67%), ahead of rules (61%). Schools that push AI without giving that time are handing teachers a new verification job and calling it a time-saver.

For musicians who teach, the studies point to where the value is. The interviewees in both Australia and Italy and China drew the same line around embodied, interpretive, relational teaching, and Indiana teachers worried about creativity and musicianship. That is the part no app replicates and the part worth charging for. If you teach privately, say so in how you describe your lessons.

Finally, the authorship anxiety in these classrooms is the same one running through the music industry. Students who learn now to say what the machine did will be better placed when distributors and labels ask the same question.

The caveat is the samples: 54 teachers in one US state, ten interviewees, and one large but single-country survey. Treat this as an early pattern, not a census.