Impacts of Artificial Intelligence in Music Education: A Scoping Review of Instructional Strategies and Student Learning Outcomes

Introduction

The growing presence of artificial intelligence (AI) in music education makes it increasingly important for educators to equip students with tools that are relevant for future professional practice (X. Wang, 2025). Rapid advances in AI are reshaping how information is produced, communicated, and consumed across society, and in music professions AI now contributes to processing, mixing, mastering, and composing (Moffat, 2021). Since the release of AIVA.com in 2016, the quality of generative music-AI has improved markedly, and services such as Suno.com and Udio.com have gained widespread attention (Rahman et al., 2024). These systems can generate complete songs with vocals and lyrics in minutes, enabling large-scale uploads of AI-created music without traditional gatekeeping. This contributed to one of the first major legal cases in the field, where Michael Smith generated thousands of AI-based tracks and used bots to inflate streaming counts, resulting in a 10-million-dollar fraud (U.S. Department of Justice, 2024).

Research shows that AI-generated music can be difficult to distinguish from human compositions (Rahman et al., 2024), with studies finding that “music composed by an algorithm and performed by a human was not distingiusable from that composed by a human” (Castelli & Manzoni, 2022, p.2). Still, music students often rate human compositions higher in emotional depth and complexity (Canyakan, 2024). Creativity remains a multifaceted construct closely tied to human cognition (Farina et al., 2024), and while generative models are trained on vast datasets, they operate by predicting musical sequences rather than engaging in human-like creative evaluation. This distinction was highlighted in Brandt’s (2023) experiment, where an AI-generated continuation of Beethoven’s Ninth Symphony revealed fundamental differences between computational output and human creative cognition. Together, these developments point to a situation in which AI can imitate established musical style with increasing fluency, while wider questions regarding authorship, intentionality, and evaluative judgement remain contested.

The emergence of generative music-AI presents both opportunities and challenges for the music industry. While concerns centre on authorship, originality, and copyright, industry bodies also acknowledge the potential of AI to enhance creative possibilities. As International Federation of the Phonographic Industry (IFPI, 2024) notes, “used responsibly, AI can contribute to amazing creative opportunities and enhance human artistry” (p. 51). At the same time, reliance on copyrighted training data has led to major legal disputes. In 2025, significant shifts occurred in the legal and commercial landscape surrounding generative music-AI, as leading rights holders moved from litigation towards formal licensing agreements with AI companies. Suno entered a “groundbreaking partnership” with Warner Music Group, and Udio established its first strategic licensing agreement with Universal Music Group, signalling an industry-wide transition towards regulated, rights-cleared AI music creation (Universal Music Group, 2025; Warner Music Group, 2025). These developments have prompted platforms to adjust their technical and licensing structures, including clearer attribution practices and mechanisms for authorised use of commercial catalogues. Such changes influence the technological environments in which students now create music, shaping which tools are accessible, under what conditions they may be used, and how issues of ownership and attribution are negotiated in educational settings. For music educators, this introduces both practical and ethical questions about preparing students for professional practice in an increasingly AI-integrated creative ecosystem.

Despite these tensions, emerging evidence suggests that AI can meaningfully enhance learning processes when integrated thoughtfully into music classrooms. Recent initiatives have explored AI for interactive and efficient teaching, such as Wissner’s (2024) music history class where students engaged with an AI chatbot, generated music, and reflected on ethical issues. AI has also been embedded into intelligent instruments, offering new opportunities for experimentation, timbre exploration, and performance (Iversen & Hebert, 2025). These developments align with broader shifts in digital education, where intelligent systems support engagement and reshape instructional practices (Hui & Geng, 2024). AI can personalise feedback, adapt learning materials, and support individual progression (Derakhshan & Ghiasvand, 2024), though teachers report significant training needs; in one study, 94% of secondary music teachers in Madrid desired further support in understanding AI’s pedagogical potential (Caudeli & Vela, 2024). Concerns remain regarding student agency and creativity: Cheng (2025) found that while AI can enhance creative development, overreliance on generative tools may weaken personal creative confidence. These mixed findings underscore the importance of examining how AI affects students’ music learning experiences and the conditions under which AI-supported learning strengthens rather than undermines students’ sense of authorship and control.

This review draws on two established theoretical perspectives. First, the Extended Creativity Framework (Boden, 1998) distinguishes between human creative cognition, characterised by intentionality, evaluation, imagination, and interpretive judgement, and computational generativity, which operates through statistical pattern learning and predictive recombination. This distinction helps explain why AI systems can convincingly imitate stylistic features of human composition while lacking purposive decision-making, reflective agency, and embodied musicianship. Second, Shneiderman’s (2020) Human-Centred AI framework emphasises that AI systems in education should support human autonomy, accountability, and iterative creative control. Framed in this way, AI tools are judged not only by their output quality but also by the extent to which they enable users to set goals, make informed choices, and refine their work over time. Together, these frameworks provide a basis for analysing how students navigate co-creation with AI tools and for interpreting emerging tensions around authorship, originality, and creative agency in contemporary music education.

Aims and Objectives

This scoping review aims to identify and analyse research-based knowledge on generative AI technologies in music education. The focus is on empirical studies documenting AI implementations in educational settings and examining how student learning outcomes and creative processes are affected, as well as studies discussing implications of AI for creativity among music students. Rather than evaluating the technical performance of AI systems in isolation, the review concentrates on how these systems are embedded in pedagogical designs, classroom activities, and assessment practices.

Previous reviews have begun to chart this field. Merchán Sánchez-Jara et al. (2024) synthesised application areas of AI in music education and associated challenges and opportunities, while Y. Zhang, Fen, et al., (2024) examined how AI can facilitate innovative learning in music and related disciplines. Lee and Kwon (2024) conducted a meta-analysis of AI education in South Korean K–12 classrooms. Building on and extending this work, the present review focuses specifically on generative AI technologies in classroom-based music education, emphasising student outcomes, learning experiences, countries of origin, and theoretical and methodological approaches. The intention is to provide a map of how AI is currently being used in empirical studies of music education and to identify patterns that may guide future research and practice.

The objective is to provide an overview of empirical research on generative AI technologies in music education and to examine how students are affected by AI technologies in classroom-based contexts.

This leads to the following research questions:

Method

There are multiple established methodologies for conducting scoping reviews, each offering a structured sequence of steps from study identification to the presentation of results. This review follows Arksey and O’Malley’s (2005) five-step framework: (1) identifying the review questions, (2) identifying relevant studies, (3) selecting the studies, (4) charting the data, and (5) collating, summarising, and reporting the results. This approach was chosen to capture a broad and heterogeneous set of empirical studies, rather than to test narrowly defined hypotheses.

First, the research questions were formulated based on a preliminary review of the literature and refined to address classroom-based music education involving AI. The initial questions were intentionally broad, focusing on AI as a pedagogical tool and on student outcomes, and were then iteratively adjusted to reflect the growing emphasis on generative technologies and creativity-related measures in the field. Second, a search strategy was developed and iteratively tested, resulting in two tailored search strings focused on peer-reviewed articles related to music, creativity, and educational practice. Pilot searches were used to adjust keywords and filters to balance sensitivity and specificity.

Third, inclusion and exclusion criteria were defined a priori and applied systematically to ensure consistency during screening. The full inclusion and exclusion criteria are provided in Table 1 (Supplementary Materials). The criteria were based on empirical focus, relevance to music education and AI, publication type, and publication year (January 2023–March 2025). Studies had to report original empirical data, address AI in relation to music learning or teaching, and be published in peer-reviewed journals. After removing duplicates, titles and abstracts were screened, followed by full-text assessment of potentially relevant studies. Figure 1 presents a PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram documenting the study selection process.

OPEN IN VIEWER

Figure 1.

Prisma 2020 flow diagram.

Fourth, the research questions served as a framework for data extraction and analysis. The author conducted a content analysis using NVivo to organise codes and analytic memos, beginning with open coding and progressively clustering codes into categories related to AI technologies, educational contexts, research designs, theoretical frameworks, and measured outcomes. AI technologies were coded using a single primary category per study, reflecting the main system implemented or examined; studies that integrated multiple systems were coded as hybrid categories. These categories were then compared across studies to identify convergent and divergent patterns. Fifth, the results are presented in narrative and tabular form. For transparency, a full overview of included studies and extracted data is provided in Table 2 (Supplementary Materials). A thematic synthesis highlights how AI tools are used across contexts, what empirical outcomes are associated with their use, and what patterns emerge regarding students’ experiences of creativity and learning. Quantitative elements such as study design frequencies, geographical distribution, and sample sizes are summarised descriptively to provide an overview of the literature landscape.

A scoping approach was chosen because the field of AI in music education is both emerging and conceptually heterogeneous, with diverse tools, contexts, and outcome measures. Scoping reviews are particularly appropriate for mapping evolving areas of research, clarifying key concepts, and identifying gaps before more narrowly focused systematic reviews are undertaken (Mak & Thomas, 2022; Munn et al., 2018). In this case, the aim is not to estimate pooled effects, but to synthesise the range of empirical approaches and outcomes reported so far and to indicate where more targeted or comparative studies might be needed.

Results

This scoping review investigates empirical studies on the pedagogical applications of AI technologies in music education. Participant age groups varied: 13 studies focused on higher education, one on middle school, one on primary school, and two on high school students. Five studies included broad or unspecified age ranges. This distribution indicates that AI in music education has so far been explored most extensively at the tertiary level, often with pre-service music teachers or conservatory students, while systematic work in primary and secondary education remains relatively limited.

Geographically, the literature is highly concentrated (Figure 2). China accounts for 17 of the 22 included studies, followed by South Korea with two; the United States, Thailand, and Germany each contributed one empirical study. Several Spanish-language articles were identified but excluded based on the criteria. Conference proceedings and book chapters from Europe and North America were also found but did not meet inclusion requirements. The findings therefore primarily reflect developments in Chinese and East Asian contexts rather than a globally representative picture of AI in music education. Accordingly, the distributions, classifications, and thematic patterns identified in this review should be interpreted as a descriptive mapping of this specific evidence base rather than as indicative of global patterns in music education research. This imbalance has implications for how the results should be interpreted, especially in light of national policy differences regarding AI, data infrastructure, and curriculum reform.

OPEN IN VIEWER

Figure 2.

Study context by country.

A relatively homogeneous pattern emerged in research design (Figure 3). Seventeen studies used experimental designs (2–8, 10, 12–16, 18, 20, 22–23). Among these, two (4, 12) relied exclusively on test-based data, while the others combined tests or assessments with questionnaires. Three non-experimental studies used questionnaire data only (1, 11, 17), and two relied on interviews (9, 19). Most studies were purely quantitative, with the exception of two qualitative interview-based investigations (9, 19). Performance-based measures and self-report questionnaires were the dominant instruments. Pre- and post-tests, rubric-based assessments of compositions, and expert ratings captured learning outcomes, while self-efficacy and motivation scales were often adapted to music or technology contexts. Teacher-focused studies sometimes drew on Technological Pedagogical Content Knowledge (TPACK)-related frameworks. Many authors designed their own questionnaires to probe attitudes towards AI in music education. Few studies, however, reported systematic validity or reliability information, sample sizes were often modest, and interventions typically spanned a single semester or less. Control or comparison groups were not consistently implemented, which means that reported effects should be interpreted as preliminary rather than strong causal evidence.

OPEN IN VIEWER

Figure 3.

Ai technologies examined in included studies.

The term “AI technologies” covers a broad range of systems in the included studies (Figure 4). Examples include tools for instrumental learning (e.g., Yousician), modern AI music generators such as Suno.com, AI features integrated into Digital Audio Workstations (DAWs) (e.g., the drummer in Logic Pro), and large language model chatbots such as ChatGPT. Across instrumental, theoretical, and chatbot-based tools, systems often provided individualised feedback explaining how students could improve. Three studies specifically investigated chatbots in educational contexts, and four examined AI-based music creativity tools (music generators, plug-ins, or integrated generative features). Seven studies focused exclusively on AI-supported adaptive learning systems. One combined chatbots with adaptive learning, and six integrated adaptive systems with AI-assisted creativity tools. Two survey studies covered broad or non-specific ranges of AI technologies, exploring music students’ and musicians’ perceptions of AI in general.

OPEN IN VIEWER

Figure 4.

Study designs and data collection methods.

A wide range of theoretical frameworks reflects the interdisciplinary nature of AI integration in music education. Cognitive and motivational theories such as Cognitive Load Theory (Sweller, 1988), Self-Determination Theory (Deci & Ryan, 1985), and the Technology Acceptance Model (Davis, 1989) were used to explain cognitive processing, motivation, and attitudes towards AI-enhanced learning. Creativity and learning theories, including the 4C Model of Creativity (Kaufman & Beghetto, 2009), Creative Thinking Theory (Guilford, 1950), and Gardner’s (2011) Theory of Multiple Intelligences framed analyses of how AI tools influence creative output and related dimensions of motivation and cognition. Instructional design and personalisation frameworks such as the Presage–Process–Product (3P) Model (Biggs, 1993), TPACK (Mishra & Koehler, 2006), and Gagné’s (1985) Instructional Model highlighted AI’s potential to support adaptive and co-creative pedagogies. Additional frameworks included human-centred design, flipped classroom pedagogy, AI-enhanced personalised learning, and computational models such as deep learning, long short-term memory (LSTM), and Bayesian networks. Some studies also introduced conceptual lenses on cultural context, emotion-based learning, and ethical AI use, pointing to an emerging interest in aligning AI integration with sociocultural values and responsible practice.

Through thematic coding, four main foci were identified: (1) composition, production, and creativity; (2) instrumental and vocal learning; (3) perceptions of AI tools; and (4) non-musical outcomes such as motivation, intelligence, and self-efficacy.

Discussion

Given the strong concentration of studies within a single national and policy context, the patterns discussed here should be understood as context-sensitive indications of how AI is currently being studied and implemented in music education, rather than as claims about global practice or universally generalisable effects. As shown by the search results, a very high proportion of the included articles were from China. One likely reason is that Chinese universities are prioritising digital transformation to maintain competitiveness in the international education landscape (Xie & Wang, 2024). Qian (2023) notes that “China has shown a significant interest in utilizing the potential of Artificial Intelligence (AI) to revolutionize its education system, particularly music education, in recent years” (p. 2). China was also heavily affected by the COVID-19 pandemic, which accelerated investments in learning technologies and remote teaching (Gai, 2025). A further observation is the limited academic coverage of generative music-AI services such as Udio and Suno, despite their prominence in public discourse. A separate targeted search identified very few empirical studies on these platforms in educational contexts. While some conference papers discussed them, no peer-reviewed empirical classroom studies were found. Given that these services are relatively new, it is likely that research is underway but not yet published. Their potential to reshape mainstream perceptions of AI music generators nevertheless makes them important objects for future study.

Across the reviewed literature, participants navigate a tension between using AI tools to increase efficiency and preserve autonomy. Creators and students use AI to overcome writer’s block (Newman et al., 2023), generate more elaborate compositions (S. Zhang et al., 2024), and streamline creative workload. At the same time, when AI takes over tasks traditionally performed by the creator, the results are often perceived as less authentic and less representative of the musician’s personal imprint. This mirrors earlier debates about composing with pre-fabricated loops, where the legitimacy of the final product depends on demonstrable musicianship and creative manipulation (Bell, 2015).

The lack of transparency in training data, ethical concerns, and corporatist business models also shape perceptions. Participants express fear that AI will displace human composers and erode skill levels, fostering dependency on technology. Many argue that AI cannot be inherently creative (Newman et al., 2023), a view that aligns with broader epistemological debates on whether AI outputs, based on probabilistic recombination, can be considered genuinely creative in the absence of intentional agency (Magni et al., 2024). These concerns are amplified by the opaque nature of commercial AI systems and the difficulty of tracing how and from which sources particular outputs are generated.

A further issue is the early-stage nature of the evidence base. The literature is geographically uneven and heavily dominated by Chinese studies, which limits the generalisability of findings. However, this developmental stage is precisely when a scoping review is appropriate. Scoping reviews are designed to map emerging fields, identify initial empirical signals, and clarify conceptual tendencies before a geographically diverse or methodologically mature evidence base is available (Mak & Thomas, 2022; Munn et al., 2018). The strong representation of Chinese research can be understood in relation to China’s centralised higher education governance model, where national-level directives facilitate rapid institutional adoption and experimentation with AI-supported teaching (J. Wang, 2022). Documenting these early patterns is valuable, as they may foreshadow future pedagogical developments as AI technologies diffuse internationally and as other countries begin to implement similar initiatives.

Measured learning outcomes in the reviewed studies are largely positive, especially concerning motivation, technical skill, and creative output. At the same time, many participants report tensions around authorship and autonomy when using AI music generators (18, 19, 23). These concerns parallel students’ ethical worries about plagiarism and ownership in relation to large language models in academic contexts (Schei et al., 2024). Tigre Moura et al. (2023) found that listeners preferred music created through human–AI collaboration compared with purely AI-generated works, although fully human compositions remained the most highly valued. S. Li (2025) reported that a record label employing AI collaboratively with human producers achieved nearly three times the revenue of a label without AI, suggesting that hybrid approaches may become commercially attractive. Just as literacy in office software and DAWs has become expected in professional settings, it is likely that fluency in AI tools will be increasingly required in creative industries. AI literacy in education may therefore become essential for preparing students for future work and for enabling them to critically assess and shape AI-supported creative practices.

Methodologically, most outcomes synthesised in this review derive from short-term quantitative indicators such as pre- and post-tests, rubric-based composition assessments, exam scores, and self-report scales. These measures capture changes in performance and confidence but only partly address musical creativity as a process that unfolds over time, is negotiated relationally, and is tied to identity and authorship. Future research will need to complement experimental designs with longitudinal and qualitative approaches that examine students’ narratives, classroom interactions, and situated creative practices. Such work could, for example, trace how a cohort of students develops AI literacy across several courses, or how teachers negotiate assessment criteria when AI tools are used in composition and performance.

In the reviewed studies, AI was applied in three main types of classroom activities. First, in music creation and composition, AI tools assisted with idea generation, arranging, and structural experimentation (15, 18, 19, 22, 23). Second, in instrumental and vocal practice, adaptive feedback systems improved accuracy, rhythm, and expressiveness (5, 6, 8, 10, 16, 20). Third, in theoretical subjects such as music history, aural training, and music theory, chatbots and adaptive systems offered explanations and personalised exercises (1, 3, 9, 11). Across these domains, experimental groups generally outperformed control groups in aspects of creativity, technical skill, and theoretical understanding, although traditional instruction sometimes yielded stronger results for posture and physical technique (e.g., 16). Overall, AI appears to function best as a supplement to, rather than a replacement for, teacher guidance.

In response to the research question concerning the impact of generative AI upon student learning outcomes and motivations, the diversity of AI technologies and study designs makes it difficult to isolate precisely how AI enhances learning. Nevertheless, the evidence points towards consistently positive effects on learning outcomes, motivation, and satisfaction with instruction. For educators, this suggests that carefully designed AI integration may be a productive strategy to support engagement and learning, while also recognising that teachers with limited technological expertise may feel challenged by rapid change. At an institutional level, the findings underline the importance of professional development programmes that address both technical and pedagogical dimensions of AI use, including ethical reasoning, critical data literacy, and strategies for maintaining student agency when AI is present in the learning environment.

Although the reviewed studies provide early indications of how AI affects learning, motivation, and creative engagement, the evidence base remains methodologically narrow. Most studies rely on short-term interventions in controlled or semi-controlled settings and do not capture how AI is enacted in everyday classroom practice or how teachers and students negotiate creative agency, authorship, and pedagogical decisions. To develop more robust frameworks for AI in music education, future research should employ qualitative and mixed-methods designs that examine how AI tools mediate classroom interactions, how learners experience AI-supported creative processes, and how power, equity, and AI literacy are configured in different educational contexts. Such studies could, for example, explore how access to AI tools differs across institutions, or how students from different musical backgrounds experience AI-supported composition and performance.

Future studies should also investigate how large language model chatbots can be integrated across music subjects (e.g., composition, analysis, theory, history) and which pedagogical designs yield meaningful learning effects. Given that platforms such as Suno and Udio are widely accessible yet criticised for limited transparency and potential displacement of human producers, further research is needed on how musicians can be trained to work with these systems as collaborative tools rather than substitutes. This includes examining how creative decision-making, authorship, and workflow distribution operate in human–AI co-creation and identifying conditions under which AI expands rather than constrains students’ creative agency.

Conclusion

Across the reviewed studies, many students and practitioners welcome AI as a creative partner, yet simultaneously emphasise the need to maintain human autonomy in the artistic process. AI appears not only as a neutral tool but also as an active agent that reshapes pedagogical relationships, learning experiences, and evolving understandings of creativity. The current evidence suggests that AI-supported tools such as adaptive learning systems and generative music technologies can enhance engagement, motivation, and a sense of accomplishment. These findings are broadly consistent with previous scoping reviews and meta-analyses reporting positive effects of AI on motivation and learning experiences in broader educational contexts (Lee & Kwon, 2024; Merchán Sánchez-Jara et al., 2024; Y. Zhang et al., 2024).

At the same time, several studies reveal emerging tensions when musicians and students encounter AI technologies, particularly around authorship, originality, authenticity, and ethics in human–AI co-creation. The field of AI research in music pedagogy remains methodologically and geographically constrained, dominated by short-term experimental designs and Chinese contexts. Future work will benefit from longitudinal studies, in-depth qualitative and mixed-methods research, and critical perspectives on power relations, AI literacy, and the evolving role of the music educator who actively uses AI. Addressing these issues will be essential for developing pedagogical frameworks that harness AI’s potential while safeguarding human creativity, agency, and core educational values.

Supplemental Material