Empowering Student Learning in Higher Education with Generative AI Art Applications: A Systematic Review

Abstract

Generative Artificial intelligence (AI) art is increasingly integrated into higher education (HE). While its creative potential has been discussed, its actual pedagogical impact and implications for educational equity remain underexplored. This study conducts a systematic review to evaluate how AI art has been applied in HE settings, what teaching and learning outcomes it supports, and what structural barriers exist in its integration. Using the PRISMA framework, 65 peer-reviewed articles published Scopus and Web of Science. The included studies were synthesized thematically and find that generative AI tools are being used to support ideation, multimodal expression, and interdisciplinary projects. However, barriers such as limited faculty training and unclear evaluation standards may hinder equitable access and long-term integration. This review contributes a conceptual framework for understanding the integration of generative AI art, highlighting opportunities and structural limitations. It offers insights for curriculum designers, educators aiming to support responsible, creative, and inclusive uses of AI in arts education.

1. Introduction

With the rapid advancement of generative artificial intelligence (AI) technologies, their application in education, particularly in HE, has grown significantly [1,2,3]. Supported by capabilities such as image generation, text creation, and video synthesis, generative AI art tools such as DALL-E, Midjourney, and Stable Diffusion are not only reshaping the technical logic of artistic production but also introducing new media, methods, and cognitive frameworks for art and design education [4,5,6]. This transformation is driving a fundamental reconfiguration of teaching approaches, curriculum structures, and epistemological foundations across disciplines including visual arts, architectural design, and media communication in HE [7,8].

In the context of arts education, the use of generative AI is exhibiting a shift from functioning solely as a technical aid to acting as an active participant in the learning process [9]. An increasing number of courses are incorporating generative AI tools into instructional practices to support students in concept development, rapid prototyping, multimodal expression, and interdisciplinary collaboration [10]. At the same time, educators are leveraging these tools to explore personalized learning pathways, foster students’ critical thinking, and facilitate dialogs around creativity, authorship, and technological ethics [11,12]. However, while these innovations offer new pedagogical possibilities, they also reveal significant challenges. Disparities in technological access, variations in teachers’ AI literacy and curriculum design capacity, the lack of clear evaluation standards and definitions of originality, and students’ uncertainty about the role of human-AI co-creation all pose substantial barriers to the equitable integration of generative AI art in HE [9,13,14].

Although existing studies have explored the application of AI art in HE, most have focused on operational procedures, course-specific implementations, or student perceptions [15,16,17]. There is a notable lack of systematic and integrative synthesis in this field. In particular, the interrelationship between generative AI art, its intersection with art pedagogy and theory, student learning support, instructional design, and educational equity remains under-theorized, with few studies offering a comprehensive conceptual framework or cross-study dialog to bridge these dimensions.

The purpose of this study is to systematically review the current literature on the application of generative AI art in HE, with a focus on its pedagogical use, its impact on student learning, and the challenges related to educational equity. The study adopts a problem-oriented research framework to address three central questions: (1) Where and how generative AI art has been applied to support learning; (2) What pedagogical practices have demonstrated effectiveness in enhancing creativity, engagement, or outcomes; and (3) What barriers hinder its equitable and sustainable integration into higher education. The goal is to construct a transferable conceptual framework that informs educators, curriculum designers, and policy-makers on how to responsibly and inclusively integrate generative AI art in arts education. The research questions with their potential research value are shown in Table 1.

Table 1. Research questions and values.

Research Questions	Research Values
RQ1: In which domains have generative AI art been applied to support student learning in higher education?	Identify the disciplinary areas and educational settings in which generative AI art has been applied.
RQ2: What pedagogical practices involving generative AI art have been reported to enhance students’ creativity, engagement, or learning outcomes in HE?	Synthesize pedagogical strategies and their reported effectiveness in supporting student learning.
RQ3: What structural barriers and equity-related challenges hinder the effective integration of generative AI art into higher education curricula?	Reveal the key barriers to equitable integration of generative AI art in higher education.

While recognizing the immense potential of AI art in educational environments, this study takes a balanced view on the topic. The researchers acknowledge both the opportunities presented by AI and the significant challenges it brings. The purpose of this review is not to uncritically advocate for AI, but to engage in a thorough and nuanced discussion of its role in the transformation of HE.

2. Literature Review

2.1. The Transformation of Arts Education in the Age of Generative AI

Art education is recognized as an important way to cultivate creative thinking, esthetic literacy and cross-cultural understanding [18]. Its traditional model emphasizes craftsmanship, individual expression, cultural narratives, and systematic knowledge of art history and theory [19]. However, digital technologies are profoundly reshaping its knowledge ecology. Digital creation tools (such as Adobe Creative Suite, Blender, Procreate) make art creation more efficient and innovative [20,21]. Similarly, online courses and distance learning break down geographic barriers to art education [22,23]. Meanwhile, the introduction of generative AI tools are further reshaping creative approaches by providing instant inspiration and innovating teaching strategies, feedback mechanisms, and student engagement [24].

However, a growing paradox is that, despite the widespread penetration of digital tools, curricula and faculty capacity in higher education often lag behind, creating a structural “educational technology divide” [25]. This divide means students’ capacity to engage with AI art technologies is significantly shaped by disparities in institutional resources, access to digital devices, and the technological proficiency of instructors [26,27,28]. For example, some institutions have built digital labs and AI workshops for students to co-create with AI, while others, reliant on traditional resources, leave students on a technological periphery, affecting creative self-expression and equity in learning outcomes.

It is undeniable that, although some researchers [29,30] argue that generative artificial intelligence has revolutionized artistic creation, it is not a tool designed to replace traditional artistic practices. On the contrary, generative AI can complement conventional creative methods, helping students expand their thinking and creativity. In summary, the ongoing shift from traditional to AI-mediated art education represents both an opportunity and a challenge. The educational technology ecology discussed here lays the groundwork for understanding how generative AI is transforming teaching models and influencing the broader goals of equity and creativity in HE. It is worth noting that this study does not deny that traditional tools remain the foundation for developing students’ basic artistic skills, while artificial intelligence can enhance creativity, particularly in tasks such as prototyping, style exploration, and visual thinking, further expanding the boundaries of creative expression.

2.2. Technology Stratification and the Educational Equity Gap in AI-Driven Art Education

Institutional disparities in infrastructure and resources form the foundation of technology stratification in HE. The so-called “education divide”, systematic disparities in resource access based on socioeconomic, geographic, or technological factors, is significantly exacerbated by the uneven integration of AI in art education. Differences between institutions at the level of technological infrastructure, digital resource availability, and faculty reserves directly shape the trajectory of student acquisition of AI art competencies [31]. Typically, well-resourced universities can build cutting-edge learning ecosystems with high-performance computing and professional AI systems, while underfunded institutions remain confined to traditional art programs, hindering the development of practical AI art education [32]. This resource gradient limits students’ exposure to innovative tools and may impact their future competitiveness in future creative industries [24].

This “stratification of educational technology” profoundly impacts the student learning experience and creative development. Students with AI access can leverage these tools as creative partners for real-time feedback, style exploration, and prototyping [33], while others are excluded from such empowerment due to equipment and training gaps. This asymmetrical distribution widens generational breaks in artistic approaches [33], and creates a disconnect between course content and technological developments, forcing students to compensate for learning to adapt to the industry [34].

Equally important is the digital readiness of instructors and the adaptability of curricular frameworks. Teachers’ digital literacy and curriculum design skills are critical, those oriented towards traditional craft and theory may lack the skills to integrate AI, limiting students’ mastery and critical use of the tools [35], as Kohnke et al. [27] point out, teacher readiness is a key variable affecting the equitable diffusion of AI education.

Despite these challenges, emerging technologies also offer pathways to bridge these divides. Online education platforms and AI-powered pedagogical tools can democratize arts education by lowering technological barriers [23]. Examples include MOOC platforms like Coursera, which break geographical monopolies on top-tier courses, and generative tools like ChatGPT and DALL-E, which reshape participation in art creation [36]. To truly achieve the goal of “supporting students’ learning with AI”, the future education reform needs to be centered on the following key paths: strengthening the training of teachers in AI artistic literacy, optimizing the embedding of AI elements in the curriculum structure, and promoting the construction of a sharing mechanism for educational technology resources, so as to build a more inclusive and creative arts education ecosystem [28]. Thus, the dual role of generative AI in exacerbating and potentially bridging the educational divide is a key research area, particularly regarding equitable learning outcomes and inclusive creative development.

2.3. Generative AI Art in Education: Concepts, Functions, and Controversies

Generative AI art involves creating artworks with AI systems that use algorithms, training data, and machine learning. While its conceptual origins date to the 1970s [37], advances in deep learning since the 2010s have propelled it to the forefront of cultural and educational discourse [36]. In HE, these technologies are evolving from mere tools into co-creative agents that shape learning trajectories and expressive modalities. They manifest in three primary forms: (1) AI-assisted art, where human creators retain control but use AI for tasks such as prototyping, style transfer, or composition [38]; (2) AI-generated art, where the algorithm independently produces outputs with minimal or no human intervention [39]; and (3) AI in art analysis, an application considered to be in its nascent stages, where machine learning is used to assess style, predict market trends, or support art history research [35].

In arts education, generative AI presents opportunities to enhance creativity, encourage cross-disciplinary thinking, and facilitate multimodal expression [40,41]. These tools allow students to ideate rapidly, engage in visual thinking, and explore communication forms beyond traditional media. For some, they lower skill barriers, democratizing artistic expression. Educators are thus exploring how to embed AI into curricula to support reflective practice, adaptive learning, and real-time feedback, thereby fostering experimentation and creative confidence.

However, the pedagogical potential of generative AI is tempered by significant concerns. Some scholars warn that overreliance may impair the development of technical skills, media literacy, and esthetic judgment [36]. As Garcia [24] have pointed out, AI has brought transformation to art creation, but its core role is to expand the boundaries of creativity, rather than replace traditional forms of artistic expression. Others point to students’ limited understanding of AI systems, leading to superficial engagement or ethical blind spots [33]. Further complicating educational use are unresolved issues of authorship, copyright, and data provenance [42,43]. Consequently, educators face a dual responsibility: to harness the benefits of AI without undermining core artistic values, while encouraging students to first build a foundation through traditional artistic techniques, and then use generative AI tools to stimulate further creativity. As Epstein and Hertzmann [44] argue, AI art represents a cultural turning point that challenges the very definition of creativity. These tensions highlight the need for a structured, evidence-based review to clarify the pedagogical value and structural limitations of generative AI in HE. This review builds on these definitional, pedagogical, and ethical dimensions to critically assess how generative AI art is being implemented and what implications it holds for student learning and educational equity.

In the following sections, the method of this study is first introduced. This is followed by a detailed description of the literature search, which targeted the Scopus and Web of Science databases, and outlines the keywords and inclusion/exclusion criteria, as well as the analytical framework used to review the included papers. This is followed by an assessment of the contributions and limitations of the existing literature. Finally, the limitations of this study will be explored and recommendations for future research made.

3. Methods

3.1. Review Design and Rationale

This study adopts the Systematic Literature Review (SLR) approach to provide a panoramic view of the application of AI art in HE in recent years. As a structured research paradigm, this method can effectively identify research hotspots in the field, track the trajectory of technological evolution, and reveal existing challenges through a standardized process of literature screening and analysis [45].The academic value of SLR is manifested in three dimensions: First, the method can systematically integrate fragmented research results and construct a multidimensional theoretical cognitive framework, thus breaking through the perspective limitations of a single study [46]. Second, the risk of selection bias can be minimized and the academic rigor of the included literature can be ensured through the development of a verifiable search strategy, a double-blind screening mechanism, and a standardized quality assessment tool [47]. Finally, systematic analysis based on the chain of evidence can distill the core application scenarios of AI art in the educational arena, which can diagnose the weaknesses of the current research system and provide strategic guidelines for the integration of educational technology [48].

In accordance with the PRISMA 2020 statement [49], the review follows a staged and standardized process covering data retrieval, inclusion/exclusion screening, coding, and thematic analysis. To ensure clarity and transparency, the research process was structured into three iterative stages, namely preparing, conducting, and reporting as guided by the digital humanities research framework proposed by Xiao and Watson [47]. The operational model is visualized in Figure 1.

Figure 1. Three key stages of literature data for this study (Source: Author illustration).

3.2. Preparing Stage

This study aims to systematically review the application domains, pedagogical implications, and structural equity barriers associated with the use of generative AI art in HE. Acknowledging the interdisciplinary nature of this emerging field, the study formulates three research questions (RQs), guided by a problem-oriented analytical framework, to explore how AI art technologies intersect with pedagogy, student learning, and educational equity in curriculum design and instructional practices.

Methodologically, this review adheres to the PRISMA 2020 guidelines for systematic reviews [50], which promote transparency and replicability through a standardized procedural flow. The use of a structured protocol comprising predefined search strategies, inclusion and exclusion criteria, and quality validation mechanisms helps to minimize selection bias and ensure the alignment between literature synthesis and the research objectives. This preparatory stage lays the methodological foundation for the subsequent phases of data collection and analysis.

3.3. Conducting Stage

The search covered relevant literature from the core databases of Scopus, Web of Science (WoS), which guarantees the quality of the results retrieved, and WoS and Scopus, which are the two bibliographic databases that are often considered the most comprehensive sources of data for a variety of purposes [51]. It is worth mentioning that the WoS database covers all indexing categories in its citation index, without distinguishing between Science Citation Index Expanded (SCI-EXPANDED), Social Sciences Citation Index (SSCI) or Arts & Humanities Citation Index (A&HCI), etc., because WoS has a high reputation and visibility in the field of literature databases [52]. In addition, Scopus, as a latecomer produced in 2004, is characterized by a sufficiently large data sample and the quality of the articles it includes is relatively high [53]. As shown in Table 2, the data sources for this systematic review are from these two databases, and the scope of the search is clarified.

Table 2. Retrieve databases and search scope.

Database	Search Scope
Scopus	Article title, Abstract, Keywords (TITLE-ABS-KEY)
Web of Science	TS (Topic)

3.3.1. Retrieve Keywords

After determining the search database, it is the further thinking and deliberation on the search terms. Since the concept of AI involves knowledge of big data, human neural networks, machine learning, and cloud computing to simulate human intelligence, it covers a wide range of topics [54]. To focus more on the research topic and to obtain information that can help explain the research topic, the researcher categorized the keywords, one is “AI art” and the other is “higher education”, as shown in Table 3. The inclusion of Boolean operations in the search enriches the results and avoids missing literature, which to a large extent ensures the scientific validity and rigor of systematic reviews [55]. In addition, synonyms of keywords were conceptualized to try to cover different expressions of the same meaning to ensure the comprehensiveness of the data. Specific searches for each database are presented. According to the advanced document search requirements of Scopus and WoS, there are certain requirements for the search strings characters. In this study, based on the selection of TITLE-ABS-KEY and TS, they are grouped by asterisks (*), AND and OR.

Table 3. Retrieve keywords and synonyms search.

Keywords	Synonyms Search
AI art	AI Art *, Artificial Intelligence Art *, AI in Art, Generative Art *, AI-Generated Art *, AI-Enhanced Art *, AI-assisted Art, Creative AI, Artificial Intelligence in Creativity, AI Design
Higher education	Higher Education, University, Design Education, Art Education, Creative Arts Education, University Students, Undergraduate, College

Note: To include more terms, use the asterisk (*) as a wildcard for truncation searches when retrieving data.

3.3.2. Screening and Cleaning Procedures

After identifying the topic, the search results showed 1290 Scopus searches and 823 WoS, for a total of 2113 documents. Obviously, not all of this amount of data would be sufficient for this study, so the process of screening was necessary. In addition, the language type was not qualified as a way to ensure that high quality literature written in non-English languages was not missed. This process eliminated some documents that did not meet the needs (n = 157). Document types were selected as articles and conference proceedings, while document types such as books, book chapters, editorial materials, and comments were excluded (n = 486). This was followed by comparison machine cleaning of articles from both Scopus and WoS dual searches, retaining one or the other, with a final screened result of 1368 documents.

As shown in Figure 2, the authors carefully read all titles, keywords, and abstracts to identify irrelevant literature and exclude it (n = 1305) for the sake of scientific rigor in academic research. The rationale for this was based on the research of Kitchenham et al. [45] who argued that extensive automated searches can yield relevant data quickly, but manual culling and observation is more likely to ensure the accuracy of the data and facilitate the analysis of research trends. Finally, after manually adding 2 searches related to the topic, the literature data for the systematic review was identified (n = 65).

Figure 2. PRISMA-compliant data retrieval, screening, and cleaning processes (Source: Authors).

3.3.3. Qualifying Criteria for Data Selection

The procedure of screening and cleaning of the data is necessary to provide greater clarity on how to answer the research question, as well as certain requirements on inclusion and exclusion criteria. As Swift and Wampold [56] argued, this decision can better define the scope of the review, have a significant impact on the breadth of the literature, and provide relevant guidance on answering the research question. Therefore, Table 4 presents this content. The manipulation of this series of steps strengthens the conducting stage and creates a solid foundation for the following reporting stage.

Table 4. Inclusion and exclusion criteria.

Criteria	Inclusion	Exclusion
Keywords are reflected in the title, abstract and keywords of the literature.	✓
Peer-reviewed literature.	✓
Articles related to AI art and higher education.	✓
Technical articles not related to higher education.		✓
Early childhood, elementary, middle and high school education levels.		✓
Inaccessible, non-full-text, non-peer-reviewed research.		✓
Studies related to medicine and therapy.		✓
Duplicate articles in searches.		✓
Journals and Conference proceedings.	✓
Book chapter, Book, Editorial materials, Comments, etc.		✓

Note: A checkmark (✓) indicates compliance with the corresponding criteria.

3.4. Reporting Stage

This stage focused on synthesizing the findings of the 65 selected studies in alignment with the three research questions: application domains, pedagogical practices and impacts, and equity-related barriers. The method combined narrative synthesis and thematic classification [57]. First, key themes and patterns were extracted from the literature, and based on these, the literature was categorized. These classifications were based on core concepts and recurring patterns in the research, with the aim of clearly presenting the trends and status of each research area. Using a structured data matrix (see Appendix A), key attributes such as study origin, methodological design, AI tools used, and target educational contexts were extracted to support comparative analysis. Each study was classified based on methodological type (qualitative, quantitative, or mixed-methods) and then grouped thematically around core educational functions and challenges.

The analytical framework integrated descriptive statistics with qualitative coding [58,59], enabling thematic mapping across three core dimensions: application domains, pedagogical practices and learning impacts, and equity-related barriers. This composite approach supported both in-depth case-level interpretation and cross-study comparison, generating generalizable insights relevant to HE curriculum design and inclusive policy-making.

4. Results

This section presents the results of the systematic review and provides a thematic synthesis of studies, aligned with the three research questions outlined earlier. The findings are organized into four subsections. First, a descriptive overview summarizes the general trends in publication years, research methods, and geographic distribution. This is followed by three thematic sections that correspond to the core dimensions of the review: (1) application domains of generative AI art in HE (RQ1), (2) pedagogical practices and their reported impacts on student learning (RQ2), and (3) equity-related barriers (RQ3).

4.1. Overview of Included Studies

A total of 65 documents were screened in this study, focusing on the application of AI art in HE. The data were mainly obtained from two core databases, Scopus and WoS, to ensure the breadth and academic quality of the research.

In terms of temporal trends, the literature analysis shows that research on AI art in HE has shown a significant growth trend in the last decade, especially after 2020 when the number of studies has surged. In addition, the number of studies peaks in 2023–2024. Although the number of studies in 2025 is relatively small, this trend may be limited by the fact that only formally published articles were included in this study, and some of the 2025 literature is still pending publication.

In terms of research methodology, AI art in HE encompasses four broad categories: qualitative, quantitative, mixed-methods, and theoretical and review studies (see Figure 3). Qualitative research (38 articles) was dominant, with extensive use of case studies, interviews, and focus groups. They mainly explored the ways of applying AI art in teaching practice, teachers’ and students’ experiences, and curriculum integration strategies, showing a high degree of academic interest in the actual operation of generative AI art at the classroom level. Especially in 2024, this type of research reaches its peak, indicating that with the popularization of AI-generated art technologies, researchers are increasingly concerned about the practical application scenarios of AI art in education and how educators can effectively integrate these emerging technologies. In contrast, quantitative studies are relatively few in number, with a total of 13 articles, assessing the impact of the AI tool on students’ creativity, motivation and engagement, which provided initial data to support the effectiveness of the technology’s learning support. For example, Ting et al. [60] explored the acceptance and perceptions of AI paintings and literature to understand the progress and capabilities of AI in art. Such studies, while less quantitative than qualitative studies, provide more data-supported conclusions in measuring the educational effects of AI in the arts.

Figure 3. Trends in research methods (Source: Authors illustration).

In addition, mixed-methods studies (9 articles) combine qualitative and quantitative analyses to provide a more comprehensive perspective in assessing student learning outcomes, the creative process, and instructor feedback. For instance, Vartiainen and Tedre [61] investigated the feasibility of using text-to-image generative AI for creative production in craft education, analyzing its potential impact on teaching practices, student creativity development, and artistic expression. This type of research can more effectively reveal the actual role of AI art in HE and provide strong support for curriculum design and optimization of teaching methods. Finally, theoretical and review papers are the least in number with only 5. They are centered on sorting out the status of research on AI art in HE, its core issues, and future directions. As an example, Park [62] explores the potential of AI for new inquiry and creativity in the art curriculum, but also critically examines the ethical issues that may exist in its implementation in the art classroom. This type of research, though small in proportion, is valuable for theory building and guiding future research paths in the field.

Overall, the literature analysis in this study reveals a diverse range of research methodologies and rapidly evolving trends in AI art in HE. The growth in the number of studies reflects the importance of AI in the field of arts education, and the application of different research methods provides multidimensional insights into the field. Considering that the current application of AI art in HE is still in the exploratory stage, many studies still focus on case studies and qualitative methods to analyze its practice and impact in the classroom. This section serves as a general overview of the research sample and sets the stage for subsequent thematic analyses around the three research questions.

4.2. Application Domains of Generative AI Art in Higher Education

In response to Research Question 1 (RQ1), this section provides a systematic overview of the practice of generative AI in HE based on the included literature. The application, in this context, not only includes the introduction of technology, but also emphasizes its actual pedagogical function and educational significance in supporting student learning. This dimension also constitutes the primary objective of this study, which is to identify the key areas where AI art plays a role in teaching and learning activities, and to provide a contextual basis for the subsequent analyses of pedagogical effectiveness (RQ2) and structural challenges (RQ3). In HE, the adoption of AI art is not evenly distributed, and differences in resource allocation, technology access, and pedagogical adaptability across institutions have led to significant differences in the degree of adoption [63]. Top colleges and universities have leveraged AI technologies to innovate in various aspects such as art conservation, personalized learning, and interdisciplinary research, while colleges and universities with limited resources are still lagging behind in terms of technological equipment and curriculum design [64]. Comprehensive analyses show that the current use of AI arts focuses on six typical areas, each of which serves the student learning process in different ways: (1) art preservation and exhibition, (2) application in different art and architecture courses, (3) AI-enhanced creative practices, (4) personalized learning, (5) interdisciplinary applications, and (6) social and educational impacts of AI art.

As shown in Table 5, the application of AI art in HE is polymorphic. In art conservation and exhibitions, AI is used for tasks such as digital restoration, computer visual recognition, and virtual curation, which is used for tasks such as digital restoration, computer visual recognition, and virtual curation in educational contexts involving heritage and exhibition design [65].

Table 5. Areas of application of AI art in HE.

Field of Application	Specific Examples	Number of Publications
Art Preservation and Exhibition	AI for art conservation and exhibition spaces. AI technology can create detailed digital archives by scanning and digitizing artworks, helping to preserve their condition. Adoption of generative AI techniques to aid art restoration and digital display	2
Application in Different Art and Architecture Courses	AI for concept generation and prototyping in design education Visual arts teaching AI for teaching gamification courses In interior design education Branding and advertising design education	19
Personalized Learning	AI-powered personalized learning STEAM education	7
AI-Enhanced Creative Practices	As a creative aid for students For enhancing the creative process Used in character design courses Artistic creations for emotional expression, e.g., eye-tracking studies Promoting greater diversity in art creation	13
Expansion of Teaching Resources and Critical Discussion	AI literacy training [66]. Teachers’ use of AI tools in arts instruction Ethical and creative discussion of AI in arts education	6
Interdisciplinary Applications	Integration with human–computer interaction (HCI) Combining the Internet of Things (IoT) Application in architecture education Application in fashion design education	9
The Social and Educational Impact of AI Art	Students’ attitudes towards AI art Teachers’ acceptance of AI teaching Impact of AI art on the traditional art industry	9

Application in different art and architecture courses. Mainly in architecture, visual communication, advertising and interaction design courses, tools such as DALL-E and Midjourney are widely used for sketch generation, style migration and conceptual exploration, which is used in creativity-related coursework or assignments [6,67,68]. Meanwhile, the rise of personalized learning has enabled AI technologies to provide customized instruction based on students’ learning trajectories, such as STEAM education that combines AI technologies to provide interdisciplinary art and design training for students from different disciplinary backgrounds [4]. In the field of AI-enhanced creative practices, AI is used to optimize the design process, assist in character design, and generate emotionally expressive artworks, further expanding the possibilities of artistic creation [69,70]. New studies, such as Liu et al. [71] and Chen et al. [72], show that AI tools like Stable Diffusion and Midjourney enhance creativity and esthetics, respectively. Additionally, in the area of teaching resources expansion and critical discussion leadership. Teachers use AI art tools to design interactive teaching tasks that lead students to think critically about the ethical boundaries of originality, copyright, and AI, broadening the social dimension of teaching goals [26,66]. Finally, several studies highlight the use of generative AI in cross-disciplinary collaborations. Applications include the integration of AI with human–computer interaction (HCI), the Internet of Things (IoT), and fashion and product design. These contexts emphasize student exposure to emerging media and digital composition [73,74].

The reviewed literature indicates that generative AI art is being applied across a diverse set of instructional contexts. These applications span technical, creative, and interdisciplinary functions, establishing a broad foundation for subsequent analysis of pedagogical practices (Section 4.3) and equity considerations (Section 4.4).

4.3. Pedagogical Practices and Reported Learning Impacts of Generative AI Art in Higher Education

In response to Research Question 2 (RQ2), this section focuses on generative AI art practices that have been empirically demonstrated to be pedagogically effective in the literature. These pedagogical activities are widely used in design, visual arts, media and interdisciplinary courses, and their effectiveness in enhancing learning engagement, creative expression and cognitive outcomes is typically assessed in research through student feedback, assignment analyses, questionnaires and interviews.

The identified teaching practices can be broadly grouped into three overlapping categories (Figure 4): (1) student-facing instructional tasks such as creative triggers, co-creation, and reflective analysis; (2) teacher-facing strategies including curriculum design and pedagogical integration, especially how to apply the TPACK model [75] to curriculum design to help teachers effectively integrate technology and teaching content; and (3) technology-enhanced instructional support involving eye-tracking, emotional analytics, and immersive environments. As shown in Figure 4, the overlaps between these categories represent their dynamic interactions: the synergy between Technology-Enhanced Support and Teacher-Facing Strategies enables AI-augmented design; the intersection of Technology-Enhanced Support and Student-Facing Tasks creates interactive feedback loops; and the confluence of Teacher Strategies and Student-Facing Tasks provides essential pedagogical scaffolding.

Figure 4. Conceptual categorization of pedagogical practices involving generative AI art (Source: Authors illustration).

Table 6 provides detailed examples of each practice category, including their instructional objectives, AI tools, evaluation methods, and reported learning impacts.

Table 6. Reported learning impact of generative AI art pedagogical practices in higher education.

Types of Teaching Practices	Use of AI Tools	Objectives/ Tasks Common	Assessment Methods	Reported Learning Impact
Creative Trigger Tasks	DALL-E, Midjourney, RunwayML	Stimulating Students’ Visual Imagination and Creative Thinking, Rapid Ideation and Prototyping	Student work analysis, instructor interviews, self-reflection	Enhanced expressive diversity, stimulated creative enthusiasm, and improved visual storytelling ability
Collaborative Co-creation	AIGC tools (e.g., ChatGPT, image generators)	Completion of an inter-professional AI arts co-creation project to improve communication and collaboration	Group project presentations, observational notes, learning journals	Improved collaboration, multimodal expression, and cross-disciplinary integration skills
Critical Reflection	AI artwork, generating image examples	Guiding students to explore AI ethics, originality and expression	Class discussions, reflective journals, questionnaires	Strengthened students’ critical understanding of AI technologies and ethical reasoning
Data-based Assessment	Eye-tracking, sentiment analysis systems	Analyzing Student Creative Behavior, Visual Attention and Emotional Response	Eye-tracking data, affective computing, behavioral tracking	Support personalized feedback and adaptive instruction
AI-driven pedagogical support	Generative AI tool + TPACK framework + curriculum platform	Supporting teachers in curriculum design to aid student understanding and engagement	Teacher interviews, TPACK assessment scales, learner feedback surveys	Facilitated the transformation of teachers’ roles toward learning facilitation and increased student engagement

Notes: The table is based on a comprehensive summary of the included studies, as detailed in Appendix A. For ease of reading and presentation, literature sources are not itemized in the table.

The widespread use of AI art has introduced new approaches to teaching and learning in HE. These practices have been reported to support innovation in curriculum design, idea generation, and student engagement. It has been shown that AI-assisted tools not only optimize content, but also show positive potential to stimulate student creativity and enhance personalized learning experiences [67,76]. In courses such as Visual Arts and Architectural Design, AI image generation tools such as Midjourney and DALL-E are widely used for conceptualization and prototype development, helping students to get into a creative state quickly and improve their efficiency of expression and stylistic sensitivity [77]. Hwang and Wu [78] found that AI positively impacts design students’ creative cognition, encouraging innovative thinking. Meanwhile, Hiçyilmaz [79] revealed that AI supports artistic creativity and learning, with students reporting a strong sense of empowerment and engagement in using AI as a creative tool. For example, students use AI to generate visual cues as triggers for narrative or project design, which is associated with increased creative fluency and diversity of expression. In addition, Some of the teaching practices integrate AI and gamified learning, such as embedding Low-Rank Adaptation of Large Language Models training into the curriculum, so that students not only understand the technical principles in the process of manipulating AI tools, but also enhance their initiative and innovation at the creative level [68]. Other studies have highlighted that critical reflection tasks, in which teachers guide students to analyze the expressive, technological, and ethical dimensions of AI-generated works, have been used to support the development of technical literacy and to enhance students’ artistic judgment [28,66].

AI art is being used as a tool for cognitive dialog in teaching and learning, rather than just a creative medium, thus reconfiguring the learning relationship between the student and the tool. From the teacher’s perspective, AI art tools are changing the distribution of roles in the classroom. Teachers are beginning to shift from being traditional “knowledge lecturers” to “facilitators” and “learning partners,” encouraging students to develop a more critical and collaborative approach by exploring the biases, limitations, and creative mechanisms of AI systems. a more critical and collaborative learning posture [26].

In addition, the introduction of AI technology has stimulated the exploration of “intelligent assessment”. For example, eye-tracking and sentiment analysis technologies have been used to capture students’ mood swings, visual attention and cognitive engagement levels during the creative process, providing data to support instructional feedback and learning assessment [69]. This data analysis based on the learning process creates new possibilities for pedagogical feedback and personalized instruction. For example, the integration of VR with an intelligent feedback system provides students with an immersive art experience that broadens the boundaries of their esthetic knowledge, especially in terms of understanding stylistic evolution and historical genres [73]. Intelligent creation tools provide technical support and can also give advice in real time by analyzing students’ creative behavior, effectively helping students break through the creative bottleneck and stimulating continuous creative motivation.

In summary, as shown in Table 6, current generative AI art teaching practices exhibit significant diversity and effectiveness. These practices not only play a role in promoting student learning outcomes and enhancing creativity and motivation but are also reconfiguring the role of teachers and the teaching ecology, laying the foundation for a more intelligent, personalized and equitable art education model in the future.

4.4. Barriers and Future Opportunities for Equitable Integration

In response to Research Question 3 (RQ3), this section identifies and synthesizes key structural barriers that hinder the equitable integration of generative AI art in HE. Although the reviewed literature highlights the pedagogical potential of AI art, its adoption remains uneven across institutions and learner populations, often reinforcing existing inequities. Based on thematic analysis, three major categories of equity-related challenges emerge: (1) technological access disparities, (2) instructor capacity and institutional resource gaps, and (3) ambiguities in assessment and ethical accountability. For each barrier, the reviewed studies also suggest potential strategies for future improvement, particularly in areas such as curriculum innovation, technical infrastructure, and teacher training.

Technological access limitations remain a significant barrier to the equitable implementation of generative AI art. Empirical studies indicate that institutions with fewer resources often struggle to provide up-to-date platforms and design tools, which in turn limits both students’ and instructors’ capacity to engage meaningfully with AI-based creative environments [80,81]. Moreover, the absence of systematic training in tool use reinforces a dependency on default system functionalities rather than critical and creative software manipulation [63]. This indicates a lack of systematic training in tool use and raises concerns about passive reliance on default system functions, which may hinder the development of creative agency [70].

In addition to student challenges, teacher preparedness also varies significantly across institutions. Research shows that a large number of instructors, as evidenced by studies [77], have not received formal training in the pedagogical use of generative AI, which limits their ability to integrate these tools meaningfully into course design or classroom activities [66]. Heaton et al., [26] also reported anxiety or skepticism about the role of AI in creative disciplines, particularly where the use of algorithmic tools is perceived to conflict with traditional notions of originality and artistic judgment. These instructor-level barriers are compounded by institutional issues, including the absence of laboratories, image platforms, or reliable digital infrastructure needed to support AI-driven learning environments [9].

Assessment and ethical uncertainties present a third major barrier to the equitable adoption of generative AI art. One central concern involves the lack of clear guidelines for evaluating AI-assisted student work. Current literature reveals a lack of consensus on how to distinguish between human contributions and machine-generated outputs, particularly in assessing individual creativity and originality [82]. Without consistent assessment criteria, students may become uncertain about the legitimacy of their AI-enhanced submissions and hesitant to fully explore the creative potential of these tools. However, institutions with established AI support systems and clear assessment standards have seen students confidently integrate AI into their creative processes [83]. Related ethical concerns include questions surrounding authorship, data provenance, and stylistic appropriation. The legitimacy of training datasets, the cultural sensitivity of stylistic simulation, and the boundaries of intellectual property remain unresolved in many educational contexts. In several studies, students have been reported to experience cognitive and emotional dissonance when using generative AI tools in environments where these ethical dimensions are unaddressed. This ambiguity is often intensified by the limited capacity of instructors to facilitate informed discussions about the social and moral implications of AI-assisted creation [28,67]. In such cases, the absence of institutional frameworks for ethical literacy and reflective pedagogy may undermine both student confidence and learning equity.

5. Discussions

This study systematically reviews the application practices, teaching effectiveness and integration challenges of generative AI art in HE, and comprehensively responds to the established research questions around the three dimensions of “application areas”, “effective teaching practices” and “equal barriers” [74,84]. The increase in research on generative AI in art education reflects the growing interest in its potential. It is foreseeable that future research is likely to continue to focus on case studies, especially to test the effectiveness of AI art tools in different educational settings and to delve into their long-term impact on students’ creativity, critical thinking, and approach to arts education. In addition, as AI-generating technologies mature, more evidence-based, mixed-methods studies may be conducted in the future to more systematically assess the role and challenges of AI art in HE.

The review confirms that generative AI art has been adopted across diverse domains in HE, including design education, visual arts, interdisciplinary co-creation, and critical reflection activities. For example, Fleischmann [67] discovered that generative AI tools are accelerating product design and prototyping, making design education more interactive and experimental. In terms of pedagogical practice, it has demonstrated a positive impact in stimulating student creativity, promoting collaboration, and enhancing critical thinking. Wang [85] suggested that AI is not only an assistive tool, but also part of the creative process, and that its creativity and autonomy remain important topics of discussion in education. Furthermore, a study by Weng et al. [86] identified that AI can improve students’ design efficiency and skills, but its over-reliance may lead to a decrease in students’ creativity. Therefore, there is a need to balance the relationship between AI assistance and human creativity in educational settings [87]. The findings of this study confirm the great potential of generative AI as a “cognitive partner” in educational scenarios but also emphasize the practical limitations of not being able to advance in isolation from institutional and equity mechanisms.

Despite the great potential AI art show in HE, uneven access to technology exacerbates the educational divide, as Kohnke et al. [27] noted, resource-rich institutions may provide state-of-the-art AI labs and professional training, whereas institutions with limited resources may lack the appropriate infrastructure and faculty support. This is consistent with the findings of this study, for which Cho [66], for example, observed that some art faculty members have limited competence in AI technologies and struggle to adequately integrate AI tools in their teaching, which impacts the quality of teaching and further exacerbates the technological divide. To bridge this gap, policymakers should ensure equitable resource distribution, particularly by providing technical support and teacher training to under-resourced institutions. Only through such measures can AI applications be more widely implemented, ensuring equal access to its potential for all students.

In addition, the ethical and social implications of AI art education need to be looked at.AI-generated artworks involve issues such as intellectual property, data bias, and academic integrity [88]. Thus, for instance, Sáez-Velasco et al. [28] identified a general concern among students and teachers about whether AI would replace human creativity or even affect the job market in the arts industry, despite recognizing the value of AI in creative assistance. This concern is also reflected in the study of Heaton et al. [26], who showed that some art educators are wait-and-see and resistant when confronted with AI teaching tools. Therefore, future AI art education needs to pay more attention to fairness, ethical issues, and strengthen the exploration of technology ethics, art copyright, and data security in the curriculum to ensure the rational use of AI [89].

These findings suggest that while generative AI art holds potential as a transformative force in HE, its equitable integration requires deliberate attention to structural, instructional, and ethical design factors. Addressing these barriers should be a central concern in future research, policy development, and pedagogical innovation. Based on the thematic synthesis of 65 studies, this review proposes a conceptual framework illustrating the integration process of generative AI art in HE (see Figure 5). The model consists of five interrelated layers: (1) Application Domains, where generative AI art is used across diverse educational contexts, art preservation, design courses, etc.; (2) Pedagogical Practices, which translate these applications into concrete teaching strategies and learning tasks, this includes group co-creation and TPACK, among others; (3) Reported Learning Impacts, including enhanced creativity, engagement, and reflective capacity among students; (4) Equity Barriers, which highlight structural limitations that hinder inclusive implementation; and (5) Strategic Pathways, which offer actionable solutions to address these challenges, include infrastructure support and platform availability, AI literacy training and co-design with faculty, and new evaluation rubrics and reflective tasks. The framework connects specific application domains with pedagogical impacts, identifies equity-related barriers, and suggests future pathways for inclusive and responsible implementation. This model serves as a transferable structure for educators and researchers seeking to design, assess, and scale AI art-based interventions in HE contexts.

Figure 5. A Conceptual Framework for the Equitable Integration of Generative AI Art in Higher Education (Source: Author illustration).

6. Conclusions

Through a systematic literature review, this study provides a selective overview of the application practices, pedagogical effectiveness and integration challenges of generative AI art in HE. The findings show that AI art technologies have gradually penetrated multiple educational scenarios such as art and design courses, personalized learning paths, interdisciplinary projects and teachers’ pedagogical support, which not only stimulate students’ creative expression, but also promote changes in curriculum design and teaching paradigms. At the level of educational equity, AI art tools to a certain extent lower the threshold of creativity, broaden the path of student participation, and have the potential to promote digital literacy and democratize teaching. However, their effective integration is still limited by structural factors such as uneven distribution of resources, high technological thresholds, insufficient teacher training and lack of assessment mechanisms.

As a key contribution, this study proposes a conceptual framework that maps the integration process of AI art in HE. The model links specific application domains to pedagogical practices and observed learning outcomes, while identifying equity-related barriers and outlining strategic pathways for inclusive implementation.

This study has the following limitations: at first, the data sources are concentrated in the Scopus and Web of Science databases, which may neglect the important results of other academic platforms or conference papers; second, although the systematic literature review method can effectively integrate the existing research findings, it is limited by the secondary data analysis mode and lacks the empirical support of the primary data; furthermore, the rapid iterative nature of AI technology The long-term effects of AI technology’s educational impact have not yet fully emerged, especially in the areas of the evolution of the nature of artistic creativity and the reconstruction of educational equity mechanisms, which still need to be continuously tracked. Future research could unearth more context-specific literature by also incorporating specialized databases such as the Art Index, ERIC, or dissertation repositories, or adopt a longitudinal tracking design, combining quantitative experiments and mixed research methods, to systematically examine the specific paths of AI art tools on the cognitive process of learning, innovation of teaching methods and creativity cultivation, and at the same time focus on key issues such as ethical norms of technology, definition of originality of creations, and design of educational policies, in order to promote the sustainable educational application of intelligent technology.