Emerging Drivers of Adoption of Generative AI Technology in Education: A Review

Featured Application

The Three-Tier Framework supports researchers, educators, and policymakers in developing targeted, evidence-based strategies for GenAI integration. It enables context-sensitive adoption across diverse educational settings and informs professional development, curriculum design, and policy to promote the responsible and effective use of GenAI in both formal and informal learning environments.

Abstract

This concept-centric review identifies and synthesizes emerging drivers of Generative AI (GenAI) adoption in education, addressing a critical gap by offering the first structured integration of empirically supported predictors. Based on 27 peer-reviewed studies featuring validated research models, the review distils 11 predictors into a Three-Tier Framework. Core predictors—Performance Expectancy and Trust—consistently influence adoption across contexts. Moderate predictors—Effort Expectancy, Facilitating Conditions, Social Influence, Perceived Behavioral Control, and Perceived Compatibility—show variable relevance depending on technological and institutional factors. Emerging predictors—Habit, AI Literacy, Anxiety, and Playfulness—capture evolving socio-technical and individual dynamics, reflecting the rapid development of GenAI technologies. While the current literature offers valuable insights, gaps remain in addressing ethical concerns, barriers to adoption, teacher professional development, student engagement, and the influence of cultural and contextual diversity. The findings emphasize the need to iteratively refine the Three-Tier Framework by incorporating these dimensions and adapting to technological advancements. By consolidating empirical evidence and distinguishing between mature and emerging predictors, this review advances theoretical understanding of technology acceptance in education. It provides a structured foundation for guiding future research, informing policy and practice, and supporting responsible, context-sensitive GenAI integration across diverse educational settings.

1. Introduction

The integration of generative artificial intelligence (GenAI) into education represents a transformative shift in pedagogical practices and learning experiences. As educational institutions increasingly explore its potential to support personalization, enhance learner engagement, and optimize instructional strategies, understanding the drivers of its acceptance and adoption becomes a critical academic and practical pursuit. However, despite growing interest, the current discourse remains fragmented, lacking a systematic framework that comprehensively consolidates the factors influencing GenAI uptake in educational settings.

This review addresses this gap by synthesizing the existing literature to uncover key empirically supported predictors—such as performance expectancy, trust, habit, and AI literacy—that shape the adoption process. By leveraging a concept-centric approach [1], the study avoids simple narrative listings and instead builds its analysis around validated theoretical models and core constructs, offering a structured and theoretically grounded understanding. Importantly, the review does not limit itself to specific publication venues or temporal scopes, ensuring broad and unbiased coverage of relevant empirical and theoretical contributions.

Although several domains have begun to explore GenAI’s application, including healthcare, mental health, and human activity recognition, education and training contexts remain among the most investigated fields. Studies such as systematic meta-review by Kiuchi et al. [2] and survey by Ferrara [3] exemplify this interdisciplinary reach, examining conversational agents and large language models (LLMs) in contexts ranging from elderly care and mental health to behavioral modeling using wearable data. Across these fields, ethical concerns—including privacy, transparency, psychological impacts, and interpretability—frequently surface as barriers to adoption.

Focusing on healthcare, reviews by Hirata et al. [4] and Nerella et al. [5] illustrate both the promise and limitations of LLMs in diagnostic imaging, clinical NLP, and decision support, noting the need for technological refinement and ethical safeguards. Similarly, review studies by Yim et al. [6] and Goktas and Grzybowski [7] reveal that GenAI currently plays a supportive role, enhancing diagnostic decision-making and patient engagement but rarely replacing human judgment. These insights echo broader concerns raised by Hobensack et al. [8] and Ng [9] regarding the reliability and transparency of GenAI tools.

In education, recent systematic reviews and scoping studies have begun to trace both the opportunities and risks associated with GenAI. Ogunleye et al. [10] and Yusuf et al. [11] identify key adoption trends, ethical challenges, and themes such as academic integrity, assessment redesign, and institutional readiness. Sekli et al. [12] and Drobnjak et al. [13] highlight GenAI’s potential in supporting personalized feedback, adaptive content generation, and conversational tutoring while also noting constraints such as the hallucination effect and the need for human oversight.

At a pedagogical level, in their review study Sengul et al. [14] emphasize the gap between educational and industrial GenAI adoption, particularly in computing education, and call for alignment of curricular design with real-world AI practices. These concerns are further echoed in the scoping review of Ali et al. [15] who offer a categorization of user, operational, technological, and ethical challenges in ChatGPT’s educational use. Meanwhile, Amarathunga’s [16] systematic review contributes bibliometric insight, mapping global trends and flagging the technology’s potential impact on learner creativity and academic dishonesty.

The ethical dimension remains central to the discourse. Scoping reviews by Hagendorff [17] and Yan et al. [18] explore the normative landscape of GenAI, with detailed taxonomies of risks related to fairness, privacy, hallucinations, and societal impacts. They call for open-source, human-centered approaches to ensure trustworthy and transparent applications, particularly when sensitive student data and evaluative processes are involved.

These studies together illuminate the cross-domain interest in GenAI, the educational sector’s unique adoption challenges, and the urgency for a unified framework to understand the underlying psychological, technological, and contextual factors driving user acceptance.

Research Question and Review Objective

Despite a rapidly growing body of research, to the best of the author’s knowledge, no comprehensive synthesis currently exists that consolidates empirically supported factors predicting the acceptance and adoption of GenAI in education. While existing reviews often highlight trends, challenges, or ethical implications, they seldom examine validated theoretical models or focus systematically on adoption predictors. To fill this gap, the present review formulates the following research question:

RQ: What are the key factors that influence the acceptance and adoption of generative AI (GenAI) technologies in educational settings?

The goal is to identify emerging drivers and to analyze how core predictors are conceptualized, operationalized, and empirically evaluated in education-related research. Particular attention is paid to theoretical underpinnings, methodological rigor, participant demographics, and contextual diversity. In doing so, the study offers practical and theoretical insights for researchers, educators, and policymakers aiming to facilitate responsible and effective GenAI integration in learning environments.

The subsequent sections detail the review methodology, present a conceptually organized synthesis of findings, and discuss implications for future educational practices and research directions.

2. Research Approach

This review aims to uncover the key factors influencing the acceptance and adoption of generative AI (GenAI) technologies in educational settings. This research focus stems from a gap-oriented analysis of the literature, which identified a lack of consolidated, empirically grounded insights into the predictors of GenAI adoption. While general technology acceptance research is well-established, studies specifically targeting GenAI remain fragmented, particularly regarding emerging socio-technical and individual-level factors.

2.1. Database and Search Strategy

The review adopts a concept-centric approach, structuring the analysis around key theoretical models and predictors relevant to educational technology adoption rather than being limited to specific journals or timeframes (cf. [1]). This strategy supports a deeper exploration of key constructs while minimizing biases associated with dominant publication venues or temporal trends.

To ensure coverage of high-quality, peer-reviewed research, the Web of Science (WoS) Core Collection database was exclusively used, with no restrictions on category or publication date. Both journal articles and conference proceedings were included. This approach enabled the identification of relevant studies that offer insights into the factors, impacts, and practical considerations shaping the integration of GenAI in education.

The search strategy was designed to capture studies involving generative AI technologies—such as large language models (LLMs), GPT, DALL·E, and other tools capable of producing human-like content—and to focus on theoretical and empirical discussions of their adoption. The search string used was as follows:

(“generative AI” OR “generative artificial intelligence” OR “large language model*” OR “LLM*”) AND ((“theor*” OR “model”) AND (“adoption” OR “acceptance”)) AND (“education*” OR “learn*”)

The search was applied using the “TOPIC” filter (title, abstract, author keywords, and Keywords Plus) in the WoS Core Collection. Truncation (e.g., “LLM*”) was used to include multiple relevant term variants.

2.2. Inclusion and Exclusion Criteria

This review focused on empirical studies that apply theoretical models or acceptance frameworks to investigate the adoption of generative AI (GenAI) technologies in educational contexts, particularly those supporting teaching and learning. GenAI technologies, as defined here, include advanced tools such as large language models (LLMs) like GPT and image generators like DALL·E that produce human-like outputs across modalities.

To ensure methodological rigor, the following inclusion criteria were applied:

−

Empirical, peer-reviewed studies published in English (journal articles and conference proceedings),

−

Application of theoretical models (e.g., TAM, UTAUT) to examine GenAI adoption or acceptance in education,

−

Clearly formulated hypotheses and empirically tested relationships between constructs.

These criteria ensured the selection of studies providing validated, model-based evidence on predictors such as perceived usefulness, ease of use, and behavioral intention—critical for understanding adoption dynamics in educational settings.

The exclusion criteria were as follows:

−

Purely theoretical, conceptual, or review papers,

−

Meta-analyses or studies without primary data analysis,

−

Studies using surveys alone, even when employing validated instruments, without deeper model-driven examination of interrelationships between constructs,

−

Non-peer-reviewed works, such as book chapters.

In particular, studies that lacked explicit modeling or did not operationalize constructs within a theoretical framework were excluded due to concerns about the reliability and interpretability of their findings. While survey-based studies offer preliminary insights, their omission ensures consistency in analytical depth and alignment with the review’s concept-centric approach. Among the excluded works—despite their use of validated scales—are studies by Huang et al. [19], Wang et al. [20], Wu et al. [21], Herani and Angela [22], Smith et al. [23], Petrič [24], Chan and Zhou [25], Tangadulrat et al. [26], and Sallam et al. [27].

2.3. Screening and Selection Process

The literature search was conducted in mid-December 2024 using the WoS Core Collection, with no restrictions on category or publication year. This initial search yielded 161 records containing the specified terms in the title, abstract, keywords, or author keywords fields. After excluding one book chapter and non-relevant entries, titles, abstracts, and full texts were rigorously screened for relevance and alignment with the inclusion criteria.

From this process, 63 publications were retained for in-depth review. Ultimately, 27 studies met the final inclusion criteria: empirical investigations grounded in theoretical models, presenting hypotheses and model-based analysis specific to GenAI adoption in education.

By narrowing the selection to studies with clearly operationalized constructs and validated relationships, this review provides a transparent, methodologically robust foundation for identifying emerging drivers of acceptance and adoption—in line with the goals and standards of a concept-centric review.

3. Results and Discussion

This section presents the main findings of the review, focusing on empirical studies that examine the factors influencing the acceptance and adoption of Generative AI (GenAI) technologies in educational settings. To ensure clarity and coherence, the results are organized into four thematic subsections:

−

Regional and Cultural Dimensions of GenAI Adoption in Education, which examines the geographic distribution of the selected studies and explores how regional and cultural contexts shape the conceptualization and adoption of GenAI technologies;

−

Participant Demographics and Sampling Approaches, which provides an overview of participant types and sample sizes, highlighting the diversity of educational levels, roles, and research designs;

−

Adoption Models and Theoretical Frameworks Used in GenAI Research, which analyzes the conceptual foundations and theoretical models employed across the studies; and

−

What Influences GenAI Adoption in Education? A Closer Look, which explores the key predictors of adoption across individual, institutional, and socio-technical domains.

3.1. Regional and Cultural Dimensions of GenAI Adoption in Education

A review of the publication timeline shows that 26 of the 27 included studies were published in 2024, with only 1 study—Duong et al. [28]—appearing earlier. This concentration underscores both the recency of GenAI integration in education and the rapid academic response to its emerging significance. It also reflects the novelty and dynamism of the field, still in the early stages of theoretical consolidation and empirical investigation. Since the review was conducted in mid-December 2024, it is likely that additional relevant publications have since emerged, indicating a continually evolving research landscape.

Geographically, the studies reflect broad international engagement. The majority originated from China (11 studies), reflecting the country’s strong emphasis on AI research and its integration in education. Other active contributors include India, Thailand, the United Arab Emirates (UAE), and Vietnam, each with two studies. Additional countries represented by single studies include Malaysia, Norway, Poland, Singapore, South Korea, and Tanzania. One study involved a cross-national comparison between Egypt and Poland, and another encompassed multiple countries. This geographical diversity underscores the global relevance of GenAI in education and signals growing international interest in its adoption across varied educational systems.

These trends are visualized in Figure 1, which maps the geographic distribution of the reviewed studies. The figure illustrates China’s leading position, followed by a broader representation across Asia, Europe, and Africa, confirming the widespread attention GenAI is receiving worldwide.

Figure 1. Geographic distribution of empirical studies on GenAI adoption in education.

Beyond geographic distribution, the reviewed studies reflect diverse educational, cultural, and socio-technical contexts that shape how GenAI is perceived and adopted in practice. A substantial proportion of studies were conducted in East and Southeast Asia, particularly China (11 studies), followed by India (2), Thailand (2), and Vietnam (2). These studies frequently emphasize performance-oriented educational goals, AI literacy, and institutional support, aligning with national policies focused on digital transformation. For example, studies from China and Malaysia (e.g., [29,30]) incorporate predictors such as trust in AI, personal innovativeness, and perceived learning performance, suggesting an adoption discourse oriented toward technological efficacy and learner outcomes.

In contrast, European studies, such as those from Norway [31] and Poland [32,33], often integrate predictors like habit, hedonic motivation, and facilitating conditions, and in some cases challenge established assumptions (e.g., non-significance of effort expectancy or social influence). These patterns may reflect greater emphasis on individual autonomy, user experience, and long-term engagement within European educational settings.

Studies from Africa (e.g., Tanzania [34]) emphasize factors such as information quality and social influence, which are particularly salient in contexts where access to reliable digital resources and peer validation play a key role in shaping technology use. These findings highlight the importance of adapting GenAI tools to local infrastructure and educational needs, though direct discussion of systemic infrastructural constraints is limited in the reviewed literature.

In the Middle East, research from the United Arab Emirates (e.g., [35]) reflects a hybrid integration of global acceptance frameworks (e.g., TAM, UTAUT) with regionally relevant constructs such as user satisfaction, perceived efficiency, and learning performance. However, another UAE-based study [36] focuses more on perceived anthropomorphism and social sustainability, demonstrating how local values and social perceptions of AI can shape the design of adoption models.

These emerging patterns suggest that while core predictors of GenAI adoption—such as performance expectancy, perceived usefulness, and social influence—remain broadly applicable, their relative influence, interpretation, and model integration differ across cultural and institutional settings. These insights affirm the need for culturally responsive research that accounts for local pedagogical norms, technological readiness, and educational policy when studying GenAI adoption.

3.2. Participant Demographics and Sampling Approaches

Another important aspect concerns the types of participants and users examined in the analyzed studies. A significant majority of the research (n = 21) focused on higher education students, including bachelors, masters, and doctoral students (e.g., [36,37,38,39,40,41]), with data primarily collected through web-based surveys conducted in university settings (e.g., [31]). Among these, several studies specifically targeted business students (e.g., [34,42,43]) and English as a Foreign Language (EFL) students (e.g., [29,44,45]). Additionally, individual studies included healthcare students [46] and art and design college students [47], while others did not specify the students’ study programs. Only one study focused on elementary education students [48].

Beyond students, a smaller number of studies examined educators, particularly higher education teachers [32,49], pre-service teachers [50], or a mix of educators and higher education students [35,51]. Figure 2 illustrates the distribution of participant types in studies on GenAI adoption in education.

Figure 2. Participant types in studies on GenAI adoption in education.

Sample sizes varied significantly, ranging from the smallest group of 16 healthcare students [46] to the largest group of 1389 higher education students in a study by Duong et al. [28]. Notably, smaller sample sizes—typically up to 400 participants—were more common, highlighting a tendency toward moderate-sized participant groups in these studies.

3.3. Adoption Models and Theoretical Frameworks Used in GenAI Research

The review highlights the application of various technology acceptance and adoption models in the selected studies, reflecting a diverse approach to understanding the factors influencing the adoption of Generative AI (GenAI) in education. Among the 27 studies analyzed, the most frequently applied models were the Technology Acceptance Model (TAM) [52,53], the Unified Theory of Acceptance and Use of Technology (UTAUT) [54], and the extended UTAUT (UTAUT2) [55].

Other frameworks, such as the Innovation Diffusion Theory (IDT) [56,57] and the Expectation–Confirmation Model (ECM) [58], were also applied in individual studies. Specifically, TAM was employed in seven studies, UTAUT/UTAUT2 in eight studies, and the Theory of Planned Behavior (TPB) [59,60] in three studies. Additionally, seven studies combined multiple theoretical frameworks to enhance the explanatory power of individual models. The breakdown of model utilization is as follows (in alphabetical order):

−

Expectation–Confirmation Model (ECM) [58], a model based on Oliver’s [61] leading cognitive theory in the field of consumer satisfaction that explains post-adoption satisfaction and behavior through Expectation–Confirmation Theory (ECT), was applied in one study [38].

−

Innovation Diffusion Theory (IDT), proposed by Rogers [56,57] as one of the most popular models for studying the adoption and use of innovations, was adopted in a single study [39].

−

Technology Acceptance Model (TAM), introduced by Davis [52,53], a leading established theory and a parsimonious yet powerful model for explaining, predicting, and enhancing user acceptance in diverse technological deployments, was utilized in seven studies [28,29,37,40,44,46,62].

−

Integration of TAM and Self-Determination Theory (SDT) [63], a theoretical approach widely used in studies of motivational behavior that examines relationships among the social environment, psychological needs, motivation, outcomes, and well-being, was explored in two studies [41,47].

−

Combination of TAM and Theory of Planned Behavior (TPB) [59,60], a theory stating that behavior is a direct function of behavioral intention and perceived behavioral control, was investigated in one study [49].

−

Integration of TAM and UTAUT, a rich theoretical approach to explaining behavioral intentions and technology use, was analyzed in one study [35].

−

TPB was employed in three studies [36,51,64].

−

Combination of TPB and Behavioral Reasoning Theory (BRT), a theory that explores the cognitive processes and reasoning that bridge the gap between intentions and actual behavior, was examined in one study [30].

−

UTAUT was utilized in three studies [33,48,50].

−

UTAUT2, an extension of the original UTAUT that emphasizes the context of consumer use, was applied in five studies [31,32,34,42,43].

−

Combination of UTAUT and SDT was investigated in two studies [45,65].

It is also important to note that the studies using TAM, UTAUT, UTAUT2, and TPB did not employ the core models in their original forms but instead extended or modified them to address specific aspects of GenAI adoption. These adjustments were made to improve the models’ relevance and accuracy in the context of education and technology integration:

−

Extended Models: These models expand the original framework by adding new constructs or pathways while maintaining the core structure to address gaps or new challenges specific to the context.

−

Modified Models: These models fundamentally alter or redefine the original framework by removing constructs, changing relationships, or integrating elements from other models to better fit the research focus.

These adjustments in the use of models reflect the dynamic nature of GenAI adoption research, where traditional technology acceptance models are often reshaped to capture the unique factors at play in the context of education. Figure 3 illustrates the utilization patterns of models and theoretical frameworks in exploring the adoption of GenAI technology in education.

Figure 3. Overview of models and theories employed in GenAI technology adoption studies in education.

3.4. What Influences GenAI Adoption in Education? A Closer Look

The adoption of Generative AI (GenAI) technologies in education has attracted significant scholarly interest due to their potential to enhance teaching practices and learning experiences. However, the successful integration of these technologies depends on a range of interrelated factors that reflect both opportunities and challenges. Drawing from the 27 empirically grounded studies included in the review, eleven key predictors were identified, illustrating the complex interplay of individual, institutional, and socio-technical dimensions.

These predictors include individual-level factors, including individual attitudes and readiness to adopt GenAI, perceptions of its ease of use and usefulness, and emotional factors such as anxiety and playfulness. Technological and skill-based enablers—namely, AI literacy, technological infrastructure, and facilitating conditions—further influence adoption, as do institutional support mechanisms. Additionally, social and cultural influences, such as social influence, trust in AI, and perceived compatibility with existing practices, play a pivotal role in determining behavioral intention and actual use.

Together, these predictors offer a comprehensive lens through which the adoption of GenAI in education can be understood. While this subsection provides a high-level synthesis, a more detailed categorization and interpretation is presented in the following section: “The Three-Tier Framework for GenAI Adoption in Education”.

By offering this closer look, the subsection emphasizes the context-sensitive and multidimensional nature of GenAI adoption. Educators and learners do not operate in a vacuum; their decisions are shaped by internal dispositions, external support, and broader socio-technical environments. Understanding these nuances is essential for fostering effective adoption strategies tailored to diverse educational contexts.

−

Performance Expectancy (Perceived Usefulness): Users adopt GenAI when they perceive it as beneficial for achieving educational goals. Grassini et al. [31], Jang [42], and Zheng et al. [45] consistently highlight its role in shaping Behavioral Intention (BI). Additionally, Duong et al. [28] and Liu et al. [44] emphasize that Performance Expectancy influences not only BI but also actual use, making it essential for both initial adoption and sustained engagement.

−

Trust: Trust plays a central role in GenAI adoption by fostering user confidence and mitigating concerns about privacy, ethics, and transparency. Shahzad et al. [40] reveal that trust plays a crucial role in GenAI awareness, acceptance, and adoption. Similarly, Ma [29] and Yap et al. [46] highlight that perceived trust directly influences BI to use GenAI, highlighting its central role in fostering adoption

−

Effort Expectancy (Perceived Ease of Use): While simplicity enhances adoption in some contexts, its influence is often secondary to functionality. Du and Lv [48] and Lu et al. [49] show that Effort Expectancy indirectly boosts perceptions of usefulness, whereas Grassini et al. [31] and Wang et al. [50] suggest that ease of use has limited impact in educational settings where performance and trust are prioritized.

−

Facilitating Conditions: External resources and institutional support are critical for bridging the gap between BI and actual use. Zheng et al. [45] and Du and Lv [48] highlight the importance of accessible infrastructure, while Grassini et al. [31] report that Facilitating Conditions may be less impactful in resource-rich environments, emphasizing their context-dependent nature.

−

Social Influence: Social norms and external encouragement can shape adoption decisions. Changalima et al. [34], Jang [42], and Zheng et al. [45] find that Social Influence significantly impacts BI, particularly in environments where peer or societal expectations drive behavior. Wang and Reynolds [65] find SI to significantly influence both BI and usage intentions. However, studies like Sudan et al. [43] and Strzelecki et al. [32] report minimal effects, suggesting that intrinsic motivators often outweigh external pressures.

−

Perceived Behavioral Control: Confidence in one’s ability to use GenAI effectively is an important determinant of BI. Lu et al. [49], Ivanov et al. [51], and Al-Qaysi et al. [30] underscore its role in empowering users, while Al-Emran et al. [36] observe that its significance may diminish when external enablers, such as of facilitating conditions or social pressures, are strong.

−

Perceived Compatibility: Compatibility between GenAI and users’ needs, workflows, and values significantly foster adoption. Raman et al. [39] demonstrate that seamless integration into educational practices enhances BI, while Yap et al. [46] validate its role in making GenAI intuitive and relevant for users.

−

Habit: Habit strongly predicts both BI and continued use. Zheng et al. [45], Grassini et al. [31], and Strzelecki et al. [32] emphasize that routine engagement with GenAI fosters sustained adoption. Sudan et al. [43] further highlight the role of prior usage experiences in enhancing the chance of continued engagement and long-term use.

−

AI Literacy: Greater AI literacy equips users with the knowledge needed to engage confidently with GenAI. Jang [42] and Wang et al. [64] find that understanding AI principles enhances user confidence and perceived benefits, leading to positive adoption intentions.

−

Anxiety: Concerns about job displacement, ethical implications, or technological complexity can act as barriers to adoption. Shen et al. [47], Li et al. [62], and Wang et al. [50] show that anxiety negatively impacts BI by reducing perceptions of usefulness and ease of use. Addressing these fears is essential for fostering acceptance.

−

Playfulness (Hedonic Motivation): Enjoyment and intrinsic satisfaction derived from GenAI usage strongly influence BI. Zheng et al. [45] and Yap et al. [46] highlight that enjoyment and playfulness foster user engagement and enhance the likelihood of continued use, while Strzelecki et al. [32] emphasize the enjoyment and pleasure of GenAI use as strong predictors of BI.

These findings collectively underscore the multifaceted nature of GenAI adoption in education, shaped by a dynamic interplay of technological, individual, and social factors. The relative weight of these predictors often varies by context, depending on institutional type, learner demographics, or resource availability. This variation points to the importance of context-aware approaches when planning and evaluating GenAI adoption initiatives.

To support both research and practice, a systematic categorization of these predictors may prove valuable. Such a framework can guide future investigations, enabling researchers to extend existing models, accommodate emerging evidence, and identify new variables. Ultimately, this structured approach will help develop a more comprehensive and adaptable understanding of GenAI adoption across diverse educational settings. To address the complexity and contextual variability of the identified predictors, a more structured representation is essential. Accordingly, the following section proposes a conceptual framework that organizes these predictors into coherent categories. This framework aims not only to clarify the relationships among key adoption factors but also to support researchers and practitioners in systematically analyzing and applying them within diverse educational environments.

4. The Three-Tier Framework for GenAI Adoption in Education

Building on the synthesized findings of this review, this section introduces the Three-Tier Framework for GenAI Technology Adoption in Education. Designed to offer conceptual clarity and practical utility, the framework organizes the 11 key predictors into three interconnected tiers—Core Predictors, Moderate Predictors, and Emerging Predictors. This tiered structure enables a nuanced understanding of the diverse and interdependent factors that influence educators’ and learners’ acceptance and integration of generative AI tools in educational settings.

The framework provides a holistic perspective by clustering predictors based on their conceptual centrality, empirical support, and frequency of appearance across the reviewed studies. Each tier reflects a different level of influence and maturity in the research landscape, thereby offering both theoretical grounding and practical insight into the dynamics of GenAI adoption.

• Core Predictors (consistently significant across studies):
  These factors consistently exert a strong influence on the adoption of GenAI technologies, serving as essential drivers rooted in individual perceptions and socio-technical dynamics. They are the most widely recognized and stable determinants of how individuals perceive and engage with AI technologies:

  ⚬

  Performance Expectancy (Perceived Usefulness)

  ⚬

  Trust

• Moderate Predictors (context-dependent):
  These predictors vary in their influence depending on specific educational or technological contexts, emphasizing the importance of task and environmental factors in shaping GenAI adoption. While significant, their impact is less consistent and often contingent on situational factors:

  ⚬

  Effort Expectancy (Perceived Ease of Use)

  ⚬

  Facilitating Conditions

  ⚬

  Social Influence

  ⚬

  Perceived Behavioral Control

  ⚬

  Perceived Compatibility

• Emerging Predictors (increasing relevance):
  Reflecting evolving individual and socio-technical dimensions, these factors are gaining prominence as GenAI technologies mature and expand their applications in education. Although less established than core and moderate predictors, they are increasingly supported by empirical research and are expected to grow in relevance:

  ⚬

  Habit

  ⚬

  AI Literacy

  ⚬

  Anxiety

  ⚬

  Playfulness (Hedonic Motivation)

Each predictor is analyzed in terms of its definition, underlying rationale, and empirical evidence, as synthesized from the 27 studies included in this review (see Table 1). This structured approach ensures the framework’s relevance across contexts and its usefulness as a foundation for future research, policy formulation, and institutional implementation.

Table 1. The Three-Tier Framework for GenAI Technology Adoption in Education.

The Three-Tier Framework for GenAI Technology Adoption in Education
Performance Expectancy/Perceived Usefulness Definition Performance Expectancy, or Perceived Usefulness, refers to the degree to which users believe that using GenAI technologies will enhance their performance and effectiveness in educational tasks. It encompasses users’ perceptions of how GenAI can improve productivity, learning efficiency, task quality, and overall educational outcomes, making it a consistently significant predictor of technology adoption. Empirical Evidence Studies highlight the strong role of Performance Expectancy, or Perceived Usefulness, in shaping Behavioral Intention (BI), with users more likely to adopt GenAI when they perceive it as beneficial for their educational goals [31,33,42,45,50]. Numerous studies further validate the significance of Performance Expectancy in predicting BI and usage intentions, including Strzelecki et al. [32], Wang and Reynolds [65], Du and Lv [48], and Soliman et al. [41]. Additionally, Duong et al. [28] and Liu et al. [44] demonstrate its influence not only on BI but also on actual use, highlighting its pivotal role in both initial adoption and sustained engagement. These findings position Performance Expectancy as a foundational element due to its consistent significance in driving GenAI adoption in educational contexts. Rationale Performance Expectancy, or Perceived Usefulness, is a universally recognized core predictor of GenAI adoption in education. Both students and educators prioritize technologies that deliver tangible benefits, such as enhancing academic performance, improving productivity, or increasing instructional effectiveness. Empirical evidence consistently highlights that users who perceive GenAI as capable of improving learning outcomes or task quality are more likely to develop positive attitudes and behavioral intentions toward its adoption. This factor not only drives initial acceptance but also supports sustained usage, as users are intrinsically motivated by tools that align with their goals for academic and instructional improvement.
Trust Definition Trust refers to users’ confidence in the reliability, security, accuracy, integrity, and fairness of GenAI technologies. It encompasses the belief that these technologies will perform as expected, provide unbiased results, and handle sensitive data responsibly. Trust also includes transparency regarding the technology’s functioning and adherence to ethical guidelines, particularly in terms of privacy, security, and avoiding misuse. Empirical Evidence Shahzad et al. [40] underscore the crucial role of perceived trust in GenAI awareness, acceptance, and adoption in higher education. Their study reveals that perceived trust significantly moderates the relationship between GenAI awareness and its perceived ease of use, usefulness, and intelligence. Furthermore, perceived trust directly influences Behavioral Intention (BI) to use GenAI, highlighting its central role in fostering adoption [29,46]. In addition, trust also emerges as a critical factor in other studies, albeit implicitly. Shen et al. [47] link trust to user confidence by addressing concerns related to privacy, ethics, and originality. Wang et al. [64] implicitly tie trust to transparency and accountability through their focus on AI ethics. Similarly, Li et al. [62] identify AI confidence, interactivity, and content design as factors fostering positive attitudes, further underscoring trust’s implicit role in GenAI acceptance. These findings collectively position trust as a foundational element due to its consistent significance in fostering adoption across various contexts. Rationale Trust is consistently recognized as a core predictor of GenAI adoption, particularly in educational settings where the technology is relied upon for decision-making, content generation, and sensitive processes. Users’ confidence in the reliability, security, transparency, and ethical integrity of GenAI directly influences their willingness to adopt and integrate it into their workflows. In education, where AI is increasingly used for personalized learning, assessments, and content creation, it is essential for users to trust that the technology will deliver accurate, unbiased results and safeguard privacy. Establishing trust mitigates skepticism and concerns about misuse or misinformation, thereby lowering adoption barriers and supporting the long-term success and integration of GenAI tools in educational environments.
Effort Expectancy/Perceived Ease of Use Definition Effort Expectancy, also referred to as Perceived Ease of Use, represents the degree to which users believe that using GenAI technologies will be easy and effortless. This factor reflects the perceived simplicity and learnability of the technology, influencing users’ willingness to adopt and use it. Empirical Evidence Studies demonstrate that when users perceive GenAI as easy to use, they are more likely to adopt it, as evidenced by its significant influence on Behavioral Intention (BI) [33,45,48,65]. Effort Expectancy also indirectly shapes perceptions of usefulness and control; for instance, Lu et al. [49] found that ease of use enhances users’ views of GenAI’s utility, while Saihi et al. [35] emphasized the adoption benefits of simpler interfaces. Additionally, Duong et al. [28] linked Effort Expectancy to both BI and actual use, highlighting its role in sustaining long-term engagement. Conversely, several studies suggest that Effort Expectancy has limited significance in driving BI for GenAI technologies. Grassini et al. [31], Jang [42], and Wang et al. [50] observed no significant impact of Effort Expectancy on BI in educational contexts, where users prioritize functionality and performance over ease of use. Similarly, Shen et al. [47] and Sudan et al. [43] reported weak or negligible correlations, suggesting that simplicity plays a minor role when trust or perceived benefits dominate decision-making. These findings suggest that Effort Expectancy’s influence on GenAI adoption is moderate and context-dependent, particularly in settings where users possess baseline digital skills or prioritize performance over usability. Rationale Effort Expectancy is a moderate but important predictor of GenAI adoption, particularly in contexts where ease of use reduces cognitive load and supports broader adoption. For novice users or those with lower digital literacy, intuitive designs and simplified interfaces are critical for ensuring a smooth onboarding process and effective engagement with the technology. In educational settings, ease of use enhances users’ ability to achieve learning outcomes efficiently, making it a valuable attribute. However, for experienced users or those focused on performance and functionality, the importance of Effort Expectancy diminishes, as they prioritize utility over usability. Despite this variability, ensuring the technology is user-friendly remains vital for fostering adoption among diverse user groups with varying levels of proficiency.
Facilitating Conditions Definition Facilitating Conditions refers to the availability of resources, infrastructure, and support necessary for users to effectively adopt and utilize GenAI technologies. This includes access to devices, platforms, training, and institutional support. Empirical Evidence Zheng et al. [45] emphasize the critical role of Facilitating Conditions (FCs) in both shaping Behavioral Intention (BI) and sustaining actual use, highlighting their prominence across the adoption lifecycle. Similarly, Strzelecki and ElArabawy [33] and Du and Lv [48] find that FCs significantly influence actual usage behavior, demonstrating their impact in translating intention into action. Saihi et al. [35] and Ma [29] further validate FCs as a determinant of both initial adoption and continued use, stressing the worth of accessible resources, infrastructure, and institutional support in ensuring successful GenAI integration in educational contexts. Conversely, other studies suggest that Facilitating Conditions may have a limited role in predicting BI for GenAI technologies. Wang and Reynolds [65] report that while autonomy, competence, and relatedness strongly influence BI and usage, FCs lack significant impact on usage intentions. Grassini et al. [31] and Wang et al. [50] similarly find no statistical association between FCs and BI, emphasizing that users may prioritize intrinsic motivators like performance expectancy and habit. Strzelecki et al. [32] further note that while BI strongly predicts usage, FCs do not directly influence BI. These findings indicate that the role of FCs in GenAI adoption may depend on contextual factors, such as resource availability and the prominence of other adoption drivers. While FCs may be less influential in resource-rich environments, they remain essential in under-resourced settings where external support is critical for bridging the gap between intention and use. Rationale Facilitating Conditions serve as a moderate yet fundamental predictor of GenAI adoption by enabling users to translate intentions into actual use. In educational settings, access to resources such as platforms, training, and institutional support is crucial, especially in under-resourced environments, where these conditions can bridge gaps and remove barriers to adoption. While their impact may be less pronounced in resource-rich settings where intrinsic motivators dominate, Facilitating Conditions remain critical for ensuring equitable adoption. By providing the necessary tools and support, FCs empower users—regardless of motivation levels—to fully engage with GenAI technologies, making them a vital component of the adoption process.
Social Influence Definition Social Influence refers to the degree to which individuals perceive that important others (such as peers, instructors, or societal norms) believe they should use GenAI technologies. It reflects the impact of social pressure or encouragement on an individual’s decision to adopt and use technology. Empirical Evidence Numerous studies highlight the significant role of Social Influence (SI) in shaping Behavioral Intention (BI) to adopt GenAI technologies. Changalima et al. [34], Jang [42], Zheng et al. [45], and Wang et al. [50] report that SI positively and significantly impacts BI, emphasizing the value of external encouragement and societal norms in adoption decisions. Strzelecki and ElArabawy [33], Saihi et al. [35], and Du and Lv [48] further corroborate these findings, illustrating that SI, often reflected through peer or societal expectations, plays a crucial role in adoption decisions. Additionally, Wang and Reynolds [65] find SI to significantly influence both BI and usage intentions, particularly in environments where social validation drives behavior. Conversely, several studies challenge the influence of Social Influence in predicting GenAI adoption. Grassini et al. [31] report no significant impact of SI on BI, with factors like performance expectancy and habit being more influential. Similarly, Sudan et al. [43] and Strzelecki et al. [32] found minimal or negligible effects of SI, highlighting that intrinsic motivators, such as perceived benefits and personal habits, often outweigh external social pressures. These findings suggest that while Social Influence can play a pivotal role in specific cultural or institutional contexts, its overall impact on GenAI adoption is inconsistent and highly context-dependent. Rationale Social Influence emerges as a moderate predictor of GenAI adoption, reflecting its context-dependent nature. In educational settings, peers, instructors, and societal expectations can significantly shape individuals’ decisions to adopt and use GenAI technologies. This influence is particularly pronounced in collectivist cultures, where group norms and social validation strongly motivate behavior. In these contexts, SI enhances both initial adoption and continued use, as individuals are driven by the expectations and encouragement of their social circles. Conversely, in individualistic cultures or performance-oriented environments, the impact of SI may diminish, as personal autonomy and intrinsic motivators, such as perceived utility and habit, take precedence. While social support or pressure can facilitate or hinder adoption based on cultural and institutional attitudes, its variability across contexts positions Social Influence as a moderate yet necessary predictor of GenAI adoption.
Perceived Behavioral Control Definition Perceived Behavioral Control (PBC) refers to the perceived ease or difficulty of performing a behavior, which is influenced by both internal factors (such as personal control and self-efficacy) and external factors (such as resources, obstacles, and support available for technology adoption). It reflects the users’ belief in their ability to control their interaction with GenAI technologies. Empirical Evidence Research highlights the dual nature of Perceived Behavioral Control (PBC) in influencing GenAI adoption. Lu et al. [49] and Ivanov et al. [51] demonstrate that PBC significantly impacts Behavioral Intention (BI), emphasizing that users’ confidence in their ability to manage GenAI usage plays a critical role in adoption decisions. Similarly, Al-Qaysi et al. [30] identify PBC, alongside attitudes and subjective norms, as a strong predictor of technology use, underscoring its importance in shaping both intention and behavior. Conversely, other studies question the consistent significance of PBC in predicting adoption. Wang et al. [64] report that while attitudes strongly influenced usage intentions, PBC did not significantly impact BI. Likewise, Al-Emran et al. [36] observed that attitudes and subjective norms were more dominant predictors of GenAI adoption, with PBC playing a minimal role. These findings suggest that the influence of PBC may depend on external factors such as facilitating conditions or social pressures, which can overshadow users’ perceptions of control. Overall, while PBC is relevant in certain contexts, its role in GenAI adoption is highly variable, shaped by the interplay of individual and environmental factors. Rationale Perceived Behavioral Control (PBC) is a moderate predictor of GenAI adoption, shaped by its context-dependent nature. In settings where users feel confident in their ability to manage the adoption process—such as those with robust self-efficacy and access to necessary resources—PBC significantly influences Behavioral Intention (BI) and actual usage. It empowers users to overcome barriers and navigate challenges, particularly in environments where personal agency plays a central role. However, in contexts dominated by external factors like institutional support, social norms, or organizational mandates, the influence of PBC diminishes, as external determinants take precedence. This variability highlights the interplay between PBC and other factors, such as facilitating conditions and social influence. While not universally strong, PBC remains influential in specific contexts where individual control and resource availability are critical. This makes it a moderate but fundamental factor in the framework, addressing diverse user needs and supporting adoption in scenarios where autonomy and self-efficacy drive engagement.
Perceived Compatibility Definition Perceived Compatibility refers to the degree to which GenAI aligns with users’ needs, values, and educational practices. Basically, it reflects how well GenAI fits into users’ existing systems, workflows, and expectations within an educational context. Empirical Evidence Studies consistently show that Perceived Compatibility significantly influences the adoption of GenAI, with alignment between the technology and users’ needs, values, and educational practices playing a key role. Raman et al. [39] found that GenAI is more likely to be adopted when it fits well with users’ educational workflows and needs, emphasizing the technology’s relevance and usability. Similarly, Yap et al. [46] highlighted that Perceived Compatibility facilitates Behavioral Intention (BI) by making GenAI more intuitive and useful, particularly when it integrates seamlessly into existing systems and practices. The absence of opposing findings suggests that, within the contexts studied, compatibility with existing practices consistently supports adoption. However, this could reflect a limited range of studies in the field, and future research may reveal more complex dynamics or variations depending on specific educational environments. Although Perceived Compatibility has not been challenged significantly in the reviewed studies, its influence remains context-dependent, and further exploration of its role across diverse educational settings is warranted. Rationale Perceived Compatibility is a moderate predictor of GenAI adoption, reflecting its influence in aligning technology with users’ existing practices, systems, and values. In educational contexts, when GenAI integrates seamlessly into established workflows and addresses specific needs, it enhances BI and usage. As demonstrated by [39,46], this alignment fosters adoption by making GenAI intuitive and contextually valuable. While the absence of opposing findings supports its consistent positive influence, the limited body of research suggests the need for further exploration across diverse settings. The context-sensitive nature of Perceived Compatibility, coupled with its critical role in reducing resistance and increasing acceptance, establishes it as a moderate yet essential factor within the framework for GenAI adoption.
Habit Definition Habit refers to the extent to which users perform behaviors automatically based on prior experiences, with actions becoming ingrained through repeated use. In the context of GenAI, this involves users becoming accustomed to using the technology regularly, such that future use is influenced by prior interactions. Empirical Evidence Several studies [31,32,45] consistently highlight that Habit significantly influences Behavioral Intention (BI) and the sustained use of GenAI, with repeated interactions and integration into users’ routines being key drivers of continued adoption. The findings from Sudan et al. [43] also reinforce this by showing that habit influences not only BI but also adoption decisions, emphasizing the importance of prior usage experiences in enhancing the likelihood of continued engagement. The absence of opposing findings further strengthens the argument that habit plays a crucial role in shaping GenAI adoption, particularly regarding sustained use after the initial adoption phase. Rationale Habit is considered an emerging predictor of GenAI adoption, gaining significance over time as users interact with the technology regularly. Unlike traditional predictors such as perceived usefulness and ease of use, which primarily influence initial adoption, habit becomes more influential as users incorporate GenAI into their daily routines. Studies show that users who develop habits around GenAI usage tend to demonstrate higher levels of sustained adoption, driven by the reinforcement of positive outcomes through consistent engagement. In educational settings, regular use of GenAI for learning, research, or problem-solving fosters automaticity, ensuring the technology’s long-term integration. As such, habit is a fundamental emerging factor in the Three-Tier Framework, contributing significantly to long-term GenAI adoption in education.
AI Literacy Definition AI Literacy refers to the knowledge and understanding of Generative AI (GenAI) concepts, applications, and limitations, as well as the ability to evaluate and effectively use GenAI tools. It encompasses both the theoretical knowledge of how these systems work and the practical skills needed to interact with them in a meaningful and informed way. Empirical Evidence AI literacy has emerged as a meaningful predictor of GenAI adoption, with studies demonstrating its positive influence on Behavioral Intention (BI). Jang [42] found that higher AI literacy significantly shapes BI, suggesting that users with a better understanding of AI are more inclined to adopt GenAI technologies. Similarly, Wang et al. [64] identified AI literacy as a key factor in enhancing user confidence, with greater AI knowledge leading to increased trust in the technology and its perceived benefits. These findings highlight AI literacy as a central factor in fostering positive intentions and confidence toward GenAI adoption, underscoring its importance as an emerging predictor. Rationale AI Literacy is an emerging predictor of GenAI adoption, as it provides users with both the theoretical understanding and practical skills necessary for effective interaction with the technology. Informed users are more likely to trust and engage with GenAI tools in meaningful ways. By reducing misconceptions, alleviating concerns about misuse, and fostering confidence, AI literacy increases BI and facilitates adoption. In educational contexts, where critical thinking and informed decision-making are vital, AI literacy helps users navigate the complexities of GenAI technologies, supporting their long-term integration into academic workflows. As AI tools become more pervasive, enhancing AI literacy will likely become an even more critical factor in ensuring successful and sustained adoption, positioning it as a crucial emerging predictor within the Three-Tier Framework.
Anxiety Definition Anxiety refers to concerns or fears regarding potential negative outcomes of using Generative AI (GenAI) technologies, such as job displacement, ethical dilemmas, privacy breaches, or misuse of data. This factor encompasses users’ apprehensions about the risks associated with adopting and integrating GenAI into their personal, educational, or professional environments. Empirical Evidence Shen et al. [47] found that job replacement anxiety notably reduces Behavioral Intention (BI) to adopt GenAI, as users fear that the technology may replace their current or future job roles. Wang et al. [50] reported that GenAI anxiety significantly diminishes users’ intentions to adopt the technology, particularly by negatively affecting performance expectancy and effort expectancy. Similarly, Li et al. [62] demonstrated that AI anxiety leads to a lower perception of the technology’s usefulness and a reduced likelihood of adoption. These findings underscore that anxiety can create substantial barriers to GenAI adoption, highlighting the imperative of addressing these concerns for successful implementation and engagement with the technology. Rationale Anxiety is an emerging predictor of GenAI adoption, as concerns about job displacement, ethical dilemmas, privacy risks, and the misuse of data can significantly deter users from adopting the technology. These anxieties often undermine trust, reduce perceived usefulness, and heighten resistance, especially in contexts where users are unfamiliar with or skeptical of AI technologies. Fears about GenAI can create substantial barriers to adoption, hindering its integration into workflows and educational practices. As GenAI continues to evolve and become more widespread, these concerns are likely to intensify, making it an increasingly critical factor in shaping adoption decisions. Addressing these anxieties through transparent communication, clear ethical guidelines, and trust-building measures will be necessary for mitigating fears and fostering a more positive outlook toward GenAI adoption, particularly in educational settings where such concerns are often amplified.
Playfulness/Hedonic Motivation Definition Playfulness or Hedonic Motivation refers to the degree to which users perceive using GenAI as enjoyable, fun, or intrinsically satisfying. This factor highlights the enjoyment and creative fulfilment that users derive from interacting with GenAI technologies, which can enhance the overall user experience beyond practical or utilitarian outcomes. Empirical Evidence Zheng et al. [45] found that Hedonic Motivation significantly predicts Behavioral Intention (BI), emphasizing that the enjoyment users experience while engaging with GenAI plays a crucial role in their decision to adopt the technology. Similarly, Yap et al. [46] identified perceived playfulness as a top predictor of BI, further reinforcing the idea that the intrinsic satisfaction users derive from interacting with GenAI enhances their likelihood of continued use. Strzelecki et al. [32] also highlighted that hedonic motivation strongly influences BI. These studies consistently emphasize that the fun and creative aspects of interacting with GenAI are major factors in fostering user engagement, which is fundamental for adoption and long-term integration. Rationale Playfulness or hedonic motivation is an emerging predictor of GenAI adoption, as the enjoyment, creativity, and intrinsic satisfaction users derive from interacting with the technology significantly enhance their BI and likelihood of continued use. Studies suggest that when GenAI is perceived as fun and engaging, users are more inclined to incorporate it into their routines, leading to deeper and more sustained engagement. In educational contexts, hedonic motivation plays a particularly important role by stimulating creativity and exploration, which are vital for dynamic and innovative learning experiences. While playfulness alone may not be sufficient for widespread adoption, it supports ongoing user engagement and fosters a more positive attitude toward GenAI. This makes hedonic motivation an essential, though emerging, factor that strengthens adoption by making interactions both enjoyable and rewarding.

The Three-Tier Framework for GenAI Adoption in Education is both adequate and credible, grounded in established models such as TAM, UTAUT, and UTAUT2, while integrating emerging predictors specific to educational contexts. Core Predictors such as Performance Expectancy and Trust consistently demonstrate strong influence on adoption. Moderate Predictors emphasize the context-dependent nature of factors like infrastructure, social dynamics, and compatibility with existing practices. Meanwhile, Emerging Predictors—including Habit, AI Literacy, and Anxiety—are gaining relevance as GenAI tools become increasingly embedded in educational settings.

Although further exploration is needed to understand the interplay and relative influence of these predictors, the framework offers a sound foundation for investigating GenAI adoption in education. As research progresses and technologies evolve, Emerging Predictors are expected to gain greater prominence, calling for ongoing refinement of the framework to accommodate new insights and address emerging challenges.

Importantly, the Three-Tier Framework complements the Triad of Predictors proposed by Granić [66,67], originally introduced in Granić [68], which categorizes determinants of behavioral intention into three thematic dimensions:

-

User Aspects: Personal traits or characteristics that influence technology acceptance.

-

Task and Technological Aspects: Features of the technology and the specific tasks it supports.

-

Social and Environmental Aspects: Broader contextual and social influences on adoption behavior.

Building on this foundation, Davis and Granić [69] further developed the Triad by integrating constructs from various extensions and modifications of the Technology Acceptance Model (TAM).

Together, the Three-Tier Framework and the Triad of Predictors provide complementary perspectives on GenAI adoption. The Triad offers a macro-level conceptual structure emphasizing the interplay between individual, technological, and social dimensions. In contrast, the Three-Tier Framework delivers a more granular categorization—Core, Moderate, and Emerging—designed to support targeted analysis and intervention.

By integrating both approaches, researchers and practitioners gain a comprehensive understanding of adoption dynamics. The Triad of Predictors offers strategic, system-level insight, while the Three-Tier Framework identifies actionable predictors for specific educational scenarios. Combined, they form a robust analytical foundation to advance research, guide implementation, and support the effective integration of GenAI technologies in education.

5. Conclusions

This review advances the understanding of Generative AI (GenAI) adoption in education by proposing a Three-Tier Framework comprising 11 key predictors, synthesized from 27 empirical studies. The framework provides a valuable conceptual foundation for interpreting the multifaceted nature of GenAI adoption. However, to ensure its continued relevance, it is necessary to acknowledge several limitations, identify opportunities for refinement, and outline promising directions for future research.

5.1. Limitations of the Review

While this review offers important insights into GenAI adoption, certain methodological constraints and literature gaps must be acknowledged.

First, several relevant dimensions remain underexplored in the reviewed studies. These include ethical concerns (e.g., data privacy, bias), the role of teacher professional development, the impact of student autonomy, and contextual factors such as institutional policies and socio-cultural dynamics. These elements deserve further attention to understand how they interact with adoption predictors.

Second, the review exclusively used the Web of Science Core Collection to ensure access to high-quality, peer-reviewed literature. While this increased methodological rigor, it may have excluded relevant studies from other databases or grey literature—particularly those representing emerging markets or non-English-speaking regions. As such, the findings may not fully capture the global diversity of GenAI adoption.

Third, although the Boolean search strategy supported precision, it may have omitted studies using alternative terminologies or conceptual framings. Moreover, the review focused primarily on studies reporting significant adoption predictors, placing less emphasis on conflicting or null findings, which may limit insights into barriers or resistance to adoption.

Future research should address these limitations by incorporating a broader range of sources, including non-English publications and studies from underrepresented regions. Greater attention to barriers, ethical considerations, and diverse contextual variables will offer a more balanced and comprehensive understanding of GenAI integration in education.

5.2. Potential Extension of the Framework

The Three-Tier Framework provides a structured and theory-driven approach to understanding GenAI adoption. However, as technologies and educational contexts evolve, the framework must remain dynamic, inclusive, and adaptable. Future research should focus on refining and updating the framework by incorporating new predictors—such as ethics and emotional factors—and by accounting for teacher development, student engagement, and contextual and cultural diversity. These extensions are especially relevant as GenAI tools increasingly support personalized learning, automate instructional tasks, and reshape classroom dynamics.

Currently, the framework is based on studies from a relatively limited number of countries and education systems. Expanding the empirical base to include diverse cultural, regional, and socio-economic contexts will enhance its generalizability and global relevance. Understanding how local values, policy environments, and educational norms affect adoption is essential to making the framework broadly applicable.

Additionally, as new GenAI applications—such as multimodal AI, adaptive systems, and emotion-aware interfaces—are introduced, the relevance of existing predictors may shift. Regular empirical validation and theoretical refinement will be necessary to ensure that the framework reflects these ongoing technological changes.

Finally, future studies should actively include research reporting null or conflicting results to provide a more balanced view of the adoption landscape. Incorporating such findings will strengthen the framework’s robustness and prevent overly optimistic or one-sided interpretations.

In summary, extending the Three-Tier Framework involves both addressing current limitations and preparing it for future technological and contextual developments. By doing so, it can continue to serve as a reliable and adaptable tool for researchers, educators, and policymakers working toward responsible and effective GenAI integration across diverse educational environments.

5.3. Directions for Future Research

Building on the findings and limitations identified in this review, several key areas warrant further exploration to refine theoretical models, inform policy, and guide practical implementation. These research directions focus on underexplored predictors, cultural and contextual variation, evolving technologies, and interdisciplinary perspectives. Collectively, they aim to ensure the continued relevance and robustness of the proposed Three-Tier Framework for GenAI adoption in education.

−

Refining the Three-Tier Framework and longitudinal studies: Future research should refine and expand the Three-Tier Framework by incorporating emerging predictors and examining their stability over time. As GenAI technologies mature, predictors such as trust, performance expectancy, and social influence may evolve in significance. Longitudinal studies are particularly valuable for capturing temporal shifts in adoption dynamics and understanding how sustained exposure to GenAI tools reshapes user attitudes and behaviors.

−

Cultural, contextual, and demographic variations in GenAI adoption: The review highlights that GenAI adoption is shaped by regional, cultural, and institutional contexts. Studies across regions such as East Asia, Europe, and Africa suggest differences in the conceptualization of adoption drivers, reflecting variations in educational values, technological infrastructure, and policy environments. Future research should investigate how cultural dimensions—such as collectivism vs. individualism, or institutional centralization—affect perceptions and adoption. In parallel, studies should examine how contextual factors (e.g., educational level) and demographic characteristics (e.g., age, gender, socio-economic status) influence adoption behavior. Special attention should be given to teacher-specific factors, such as AI literacy, pedagogical readiness, and access to professional development, particularly in under-resourced settings.

−

Ethical considerations and adoption barriers: Ethical concerns, including data privacy, algorithmic bias, and fairness, remain insufficiently addressed in the existing literature. These issues can significantly influence trust, perceived risk, and ultimately, adoption intentions. Future studies should explore how ethical considerations interact with predictors such as perceived usefulness, trust, and social influence. Additionally, research should identify barriers such as GenAI-related anxiety, fear of obsolescence, and lack of infrastructure to develop interventions that promote equitable and responsible adoption.

−

Student engagement and learner-centered adoption: Student-related factors—such as engagement, autonomy, and motivation—are underrepresented in current GenAI adoption studies. As GenAI tools increasingly support personalized and student-centered learning, research should focus on how these technologies influence learner engagement, self-regulation, and educational outcomes. Understanding these dynamics will be essential for maximizing the pedagogical potential of GenAI while mitigating risks such as over-reliance or reduced critical thinking.

−

Technological advancements and predictor evolution: Rapid advancements in GenAI technologies, such as multimodal interfaces and adaptive AI systems, may alter the relevance and impact of existing adoption predictors. Research should investigate how these new technological affordances reshape user experiences and expectations, and whether they give rise to new predictors not yet captured by current models. This continuous adaptation is essential to ensure that theoretical frameworks remain relevant in a fast-evolving landscape.

−

Interdisciplinary collaboration and model development: The complexity of GenAI adoption necessitates interdisciplinary collaboration across fields such as education, computer science, psychology, and sociology. Future research should draw on diverse disciplinary insights to build more comprehensive models that reflect the cognitive, emotional, social, and technological dimensions of adoption. Such collaboration will help address multifaceted challenges and contribute to the development of holistic and sustainable GenAI integration strategies.

By pursuing these research directions, future studies can enhance the explanatory and predictive power of GenAI adoption models; improve the adaptability of the Three-Tier Framework; and inform inclusive, effective, and ethically grounded integration of GenAI technologies across diverse educational settings.