Usage Intention of AI Among Academic Librarians in China: Extension of UTAUT Model

Abstract

This study explores how academic librarians adopt artificial intelligence (AI) technologies, using the Unified Theory of Acceptance and Use of Technology (UTAUT) as its main framework, expanded with elements from Personal Innovativeness in IT (PIIT) and the Technology Readiness Index (TRI). A quantitative approach was applied, gathering data from 340 academic librarians and analyzing them using PLS-SEM. The results indicate that facilitating conditions (β = 0.345, p < 0.001) and effort expectancy (β = 0.123, p = 0.034) significantly influence behavioral intention, while performance expectancy (β = 0.091, p = 0.085) and top management support (β = 0.000, p = 0.997) show limited direct effects. These findings challenge some traditional assumptions of the UTAUT model. Additionally, attitudes were found to mediate the relationship between effort expectancy and social influence on behavioral intentions, while individual readiness and personal innovativeness moderate these relationships (β = −0.069, p = 0.003), highlighting the importance of individual traits. The model demonstrated strong predictive power, with R² values of 0.677 for behavioral intention and 0.574 for actual behavior, along with Q² predict values exceeding 0.56. By incorporating PIIT and TRI, this study broadens existing models of technology adoption, offering deeper insights into how organizational factors, personal traits, and readiness interact to influence AI adoption. Practical recommendations include introducing adaptive training programs, personalized support systems, and AI-driven infrastructure enhancements to encourage effective AI integration. Future research should consider longitudinal studies to examine how readiness and innovativeness evolve over time, explore cross-cultural differences, and refine strategies to ensure sustainable AI adoption in diverse academic settings.

1. Introduction

Contemporary artificial intelligence (AI), particularly generative AI, employs advanced technologies such as machine learning, large language models, and extensive datasets to generate human-like content across various domains, including writing, imagery, coding, and complex problem-solving in disciplines such as medicine and mathematics [1]. For those seeking to deepen their understanding, resources like Coursera’s AI Terms provide essential definitions commonly encountered in the discourse on these technologies. Among professionals, many librarians have become acquainted with generative AI chatbots such as ChatGPT, DALL-E, Gemini, and Copilot, in addition to numerous other products cataloged in the AI aggregator “There’s an AI for That” [2]. Generative AI is often viewed as a disruptive force [3], compelling librarians and other professionals to both grasp and adapt to these emerging tools while addressing the challenges they pose [4]. Analogous to the concept of information literacy, AI literacy is increasingly indispensable [5], enabling users to discern the extensive benefits and inherent limitations of AI. This necessitates a rigorous educational approach, requiring educators and librarians to thoroughly learn, assess, and apply AI within their specific disciplinary contexts.

The rapid expansion of AI technologies has significantly impacted the operational dynamics of academic libraries, particularly in how librarians manage services and how users interact with resources [6]. Previously, access to most library services was restricted to in-person interactions, but the advent of AI-driven tools has revolutionized this paradigm, enabling remote access to library services at any time from any location [7]. AI has notably enhanced search and discovery processes within academic libraries, where algorithms swiftly analyze vast datasets to facilitate more precise and efficient searches [8]. Furthermore, the adoption of AI not only improves efficiency [9] but also yields cost savings [10], enhances service quality, and boosts user engagement, driven by the need to meet the evolving demands of library users [11].

Despite these advancements, the adoption of AI by academic librarians varies significantly across different population groups, social settings, and cultural contexts [12], particularly in China where the internet user base has expanded dramatically, offering libraries opportunities to broaden their reach [13]. However, the uptake of AI in libraries remains modest, indicating that AI is still perceived as an innovative yet underutilized technology [14]. This low adoption rate poses substantial challenges for library institutions and underscores a persistent preference among many users for traditional methods of accessing library services [15].

To enhance the understanding and adoption of AI in academic libraries, this study utilizes the Unified Theory of Acceptance and Use of Technology (UTAUT), a foundational model that elucidates technological adoption behaviors. Although the UTAUT has been extensively applied, the need for a systematic investigation and theorization of the context-specific factors influencing consumer technology use remains [13]. Moreover, there is considerable debate among scholars regarding the sufficiency of the UTAUT’s constructs to fully explain user acceptance of new technology in voluntary contexts, which may limit its explanatory power.

This study proposes an extension of the UTAUT to include additional factors such as top management support and technological innovativeness, which are anticipated to provide a more comprehensive theoretical perspective on user technology acceptance in the context of AI [16,17]. Indeed, top management support is critical as it influences the resources allocated for AI adoption and its integration within organizational practices. Furthermore, technological innovativeness might significantly shape user adoption behaviors, with individuals exhibiting high levels of innovativeness more likely to embrace and utilize new technologies [18].

Additionally, this study introduces attitude as a mediating factor, hypothesizing that it mediates the relationships of performance expectancy [19], effort expectancy [20], social influence, and top management support [21] with behavioral intention. This inclusion aims to deepen the understanding of how various factors interplay to influence the acceptance and adoption of AI technologies among academic librarians.

Generally, this study will be helpful for library managers and policymakers to explain the currently relatively low penetration rate of AI and formulate strategies to encourage the adoption and acceptance of AI by academic librarians in China, where AI is still considered an innovation. This paper also contributes to the literature related to theories and models of technology adoption and acceptance, which have been recommended to be expanded to new contexts by many [13] and specifically to the generalizability and applicability of the UTAUT in a new context (AI in academic libraries), new user group (librarians), and new cultural setting (China), which is a critical step to advance a theory. Taking into consideration the fact that China is a country with a population that is diverse in terms of culture, education, income, and language, these features will add an interesting dimension to the work and provide a unique insight into the nature of factors that are important to library institutions in such an environment.

2. Literature Review

2.1. AI Applications in Libraries

Artificial intelligence (AI) has become a game-changer in library operations, driving improvements in service delivery, internal processes, and user experiences. AI technologies like virtual reference services, automation tools, and advanced data analytics have allowed libraries to reimagine traditional ways of functioning. According to Choukimath et al. [22] and Oyelude [23], AI’s introduction of virtual reference services has expanded accessibility and addressed key challenges in service delivery. These innovations not only boost operational efficiency but also improve the quality of information offered, enabling libraries to better adapt to the evolving needs of their users.

AI has also streamlined internal operations by automating routine tasks, thereby reducing the workload of library staff. For example, Badgujar and Badgujar [15] describe AI-driven library automation systems that simplify resource navigation, while Mupaikwa [24] emphasizes AI’s role in cataloging and classifying activities. Advanced AI applications, such as pattern recognition, knowledge mapping, and human–machine interaction, have further enhanced user experiences. Technologies like semantic ontologies and machine-generated classifications enable more personalized and accurate information retrieval, as noted by Mojca [25] and J. Li and Wang [26]. These capabilities align with the fundamental mission of libraries: to enhance knowledge accessibility while maintaining operational sustainability [27].

In addition to these operational benefits, AI has supported personalized services and environmentally sustainable practices by reducing reliance on physical resources [28]. For instance, AI-driven bibliographic services, as highlighted by Luca et al. [29], accelerate information retrieval, increasing efficiency and user satisfaction. However, the integration of AI also comes with challenges. The reliance on large datasets and complex algorithms raises concerns about data privacy and ethical AI usage [30]. Additionally, the automation of routine tasks has sparked fears about the de-skilling of library staff and potential job losses, as noted by J. Li and Wang [26].

A significant challenge in adopting AI is the gap in AI literacy among librarians. Choukimath et al. [22] and Oyelude [23] emphasize the importance of understanding AI functionalities and implications for effective and ethical deployment. There is a pressing need for targeted education and training, as the role of librarians increasingly involves managing and critically evaluating AI applications. Equipping librarians with AI-related skills is essential to ensure their active participation in the digital transformation of libraries.

Despite AI’s transformative potential, its adoption in libraries remains inconsistent. While some institutions have successfully integrated AI into their operations, others face obstacles like limited resources, resistance to change, and a lack of alignment between AI initiatives and institutional goals. Moreover, most studies on AI in libraries have focused on its technological capabilities or operational impacts, with limited attention paid to behavioral and psychological factors. Specifically, the roles of individual readiness, personal innovativeness, and attitudes in influencing behavioral intentions and actual AI usage have been underexplored.

This gap highlights the need for a deeper understanding of how academic librarians perceive and adopt AI technologies. While frameworks like the Unified Theory of Acceptance and Use of Technology (UTAUT) have been extensively applied in technology adoption studies, they often overlook individual-level factors like personal innovativeness and readiness. By incorporating constructs from Personal Innovativeness in IT (PIIT) and the Technology Readiness Index (TRI), this study aims to provide a more nuanced perspective on these influences. Addressing these gaps, this study examines the interplay between organizational factors, individual traits, and readiness dimensions, offering actionable insights to improve AI adoption in academic libraries.

2.2. Theoretical Underpinning

This study is theoretically anchored in the Unified Theory of Acceptance and Use of Technology (UTAUT), developed by Venkatesh et al. [31] and expanded by Venkatesh et al. [13]. The UTAUT explains how users’ intentions and behaviors toward adopting new technologies are influenced by four key constructs, performance expectancy, effort expectancy, social influence, and facilitating conditions, with moderating effects from factors such as age, gender, and experience. While the UTAUT has received widespread empirical validation, it has been critiqued for its limited consideration of individual differences, which are especially crucial when adopting innovative and complex technologies like AI. To address this limitation, scholars such as Andrews et al. [16] and Chen et al. [32] suggest enhancing the UTAUT model by integrating psychological and contextual dimensions to better capture technology acceptance across diverse environments.

This study builds on this refinement by incorporating constructs from Personal Innovativeness in IT (PIIT) and the Technology Readiness Index (TRI). PIIT, as defined by Agarwal and Prasad [33], highlights an individual’s willingness to explore and experiment with new technologies as a key determinant of technology adoption. Individuals with high PIIT are more likely to embrace emerging technologies, even when faced with significant effort expectancy or minimal institutional support [34]. Furthermore, PIIT can moderate the relationships within the UTAUT, amplifying the effects of core constructs like performance expectancy [35], particularly in settings where users engage with novel or rapidly evolving technologies. By integrating PIIT, this study directly addresses criticisms of the UTAUT by accounting for the influence of inherent personality traits on the adoption process.

The inclusion of the TRI, introduced by Parasuraman [36], further enriches the theoretical framework by capturing broader dimensions of readiness that shape technology acceptance. The TRI evaluates factors such as optimism, innovativeness, discomfort, and insecurity, offering insights into how psychological dispositions impact individuals’ evaluations of new technologies [37]. For example, individuals with high optimism are more likely to perceive AI as useful and user-friendly, while those with high discomfort or insecurity may hesitate to adopt technologies perceived as complex or unfamiliar. Incorporating the TRI allows this study to address gaps in understanding how readiness and self-efficacy bridge the intention–behavior gap, particularly in academic contexts [38].

By combining these theoretical elements, this study proposes an extended model where UTAUT constructs (Figure 1) serve as a foundation for explaining behavioral intention and actual technology use, while PIIT and the TRI enhance the model’s capacity to capture individual variability in AI adoption. This integration is especially pertinent given AI’s rapidly evolving and complex nature, which heightens the need for individual-level readiness and innovativeness in professional environments. In such contexts, the risks associated with misusing or misunderstanding AI can be significant [39], emphasizing the importance of personal and institutional readiness.

2.3. Hypothesis Development

2.3.1. Performance Expectancy and Behavioral Intention

Kim and Pae [40] established that satisfaction is strongly tied to performance and disconfirmation, while behavioral intention is influenced by expectations, performance, and disconfirmation. Similarly, Alblooshi and Abdul Hamid [41] and Afan Suyanto et al. [20] identified a direct impact of performance expectancy on behavioral intention. However, the link between performance expectancy and actual performance remains inconclusive. While R. A. Gordon [42] observed a significant relationship, Hashiguchi [43] found no such connection. X. Xu et al. [44] further noted that attitude, subjective norms, perceived behavioral control, and perceived benefits play key roles in determining the intentions of Chinese academic librarians to collaborate on research. These findings align with Yakubu et al. [45], who emphasized the relevance of the Theory of Planned Behavior in shaping librarians’ intentions to adopt AI in Nigerian academic libraries. In higher education, Alzahrani [46] highlighted the influence of performance expectancy on students’ attitudes and behaviors toward AI.

H1: 

There is a significant relationship between performance expectancy and behavioral intention.

2.3.2. Effort Expectancy and Behavioral Intention

Moya et al. [47] highlighted that behavioral intention mediates the link between effort expectancy and actual system usage, showing a strong positive effect. Bagozzi et al. [48] suggested that the level of effort required can act as a moderator between intention and behavior, an idea further reinforced by Sheeran et al. [49], who observed that intention accounts for significant behavioral variance. Additional research by Monica and Ramanaiah [50] explored how goal conflict and expected reward influence the intention–behavior relationship. Neuroscientific evidence from Khachouf et al. [51] indicates that voluntary modulation of mental effort can impact this relationship, further supporting the significant role of effort expectancy in shaping behavioral intentions.

H2: 

There is a significant relationship between effort expectancy and behavioral intention.

2.3.3. Social Influence and Behavioral Intention

Studies by Q. Ahmed and Paracha [52] demonstrated that social influence positively affects behavioral intention, with age moderating this relationship. This aligns with findings from Shih and Farn [53], which revealed that subjective norms and attitudes are critical predictors of behavioral intention. Vitória et al. [54] also emphasized social influence as a significant factor in predicting behavior, with descriptive norms playing a crucial role. Furthermore, Vera and Espinosa [55] highlighted how involvement and social intentions shape the relationship between social influence and behavioral intention.

H3: 

There is a significant relationship between social influence and behavioral intention.

2.3.4. Top Management Support and Behavioral Intention

The influence of top management support on employees’ attitudes and commitment, which subsequently shapes behavioral intentions, has been extensively documented. J. Y. Lee et al. [21] demonstrated that strong management support positively impacts employee attitudes and organizational commitment, both of which are key predictors of behavioral intention. Lynn McFarlane et al. [56] corroborated this, showing that job satisfaction and organizational commitment significantly influence behavioral intentions. Bajwa [57] emphasized how top management support amplifies the positive effects of HR practices on employee outcomes. Similarly, Ahn Shin [58] noted that top managers’ cognitive frames influence decision-making, risk-taking, and performance, underscoring the critical role of managerial support in shaping employee behaviors and intentions.

H4: 

There is a significant relationship between top management support and behavioral intention.

2.3.5. Facilitating Conditions and Behavioral Intention

Gardner et al. [59] argue that habitual behavior can weaken the link between intention and behavior, particularly under specific facilitating conditions. Ajzen [60] and Conner et al. [61] emphasize that intentions, especially those based on anticipated emotional reactions, are critical predictors of behavior. Experimental evidence from Webb and Sheeran [49] further supports that changes in behavioral intentions lead to behavioral modifications. However, Miniard et al. [62] question whether context-specific intention measures are more effective than direct ones. Additionally, Bagozzi and Yi [63] highlight how factors such as intention formation and question format moderate the attitude–behavior link. While facilitating conditions influence the strength and direction of intention–behavior relationships, the mechanisms underlying these interactions remain unclear.

H5: 

There is a significant relationship between facilitating conditions and behavioral intention.

2.3.6. Facilitating Conditions and Actual Behavior

Facilitating conditions, such as social context [64], electronic communication platforms [65], and organizational and technical infrastructure, significantly impact actual behavior. For instance, facilitating conditions have been shown to increase rental rates in virtual video rental environments [65] and boost employee motivation in organizational contexts like the postal sector. These effects are influenced by mood, norms, and self-monitoring [66], as well as cultural factors, such as the Indian mindset [67]. Gardner et al. [59] further note that habitual behaviors can diminish the influence of intentions, particularly when self-control is compromised. Eskenazi and Neumaier [68] observe a biological basis for this, with increased 5-HT6 receptor expression reducing habitual behaviors. Additionally, Khatab et al. [69] highlight the importance of inhibitory control in cooperative settings, such as among children with autism. These studies underline the complex dynamics between facilitating conditions and actual behavior.

H6: 

There is a significant relationship between facilitating conditions and actual behavior.

2.3.7. Technological Innovativeness as a Moderator

Several studies examine the moderating role of technological innovativeness in the intention–behavior relationship. Simarmata and Hia [70] found that personal innovativeness in IT predicts behavioral intention, while Fu and Elliott [71] and Nayak et al. [72] identified moderating effects of perceived product innovativeness and entrepreneurial motivation, respectively. Jackson et al. [73] emphasize that strong attitudes and personal innovativeness influence mediators that shape behavioral intentions. Scott and Bruce [74] further explore how organizational factors moderate the effect of personal change values and innovative intentions on outcomes. Moghavvemi et al. [75] add that perceived desirability, feasibility, and performance expectancy are key precursors to IT adoption, with an individual’s propensity to act serving as a moderator.

H7: 

Technological innovativeness moderates the relationship between behavioral intention and actual behavior.

2.3.8. Attitude as a Mediator

Attitude plays a significant role in mediating the relationship between performance expectancy and behavioral intention. King [76] first highlighted this, and subsequent research has validated it across various contexts, including QRIS usage in MSMEs, numerical progression tasks [77], and effort-based behaviors [48]. Reinhard and Dickhäuser [78] examine how specific and general self-concepts, task difficulty, and mood influence performance expectancy and subsequent intentions. Putwain et al. [79] further explore how expectancy of success, attainment value, and behavioral engagement predict achievement. Locke and Bryan [80] discuss how goals and intentions impact performance levels, task choices, and preferences. Collectively, these studies underscore the mediating role of attitude in shaping the link between performance expectancy and behavioral intention.

H8: 

Attitude mediates the relationship between performance expectancy and behavioral intention.

2.3.9. Effort Expectancy, Attitude, and Behavioral Intention

Numerous studies have explored the relationship between effort expectancy, attitude, and behavioral intention, emphasizing the mediating role of attitude. Moya et al. [47] found that behavioral intention mediates the link between effort expectancy and actual system usage, aligning with Schultz and Oskamp [81], who highlighted that attitudes have greater predictive power for behavior when effort levels are high. Similarly, Maddux and Galinsky [82] emphasized the role of outcome expectancy, a component of attitude, in shaping behavioral intentions. These findings underscore the strong mediating effect of attitude in the relationship between effort expectancy and behavioral intention, influenced by the effort required and the degree of intention formation.

H9: 

Attitude mediates the relationship between effort expectancy and behavioral intention.

2.3.10. Social Influence, Attitude, and Behavioral Intention

Attitude also plays a critical mediating role in the relationship between social influence and behavioral intention. Research by Bagozzi and Yi [63] and Ashraf et al. [83] consistently demonstrates that attitudes, shaped by factors like intention formation, effort levels, and cognitive structures, mediate this relationship. Normative beliefs and the motivation to comply further influence attitudes, creating a complex interplay between attitudes, social influence, and behavioral intentions. These findings suggest that social influence impacts behavioral intention indirectly through attitude formation.

H10: 

Attitude mediates the relationship between social influence and behavioral intention.

2.3.11. Top Management Support, Attitude, and Behavioral Intention

In organizational settings, top management support has been shown to influence employees’ job satisfaction and organizational commitment, which in turn shape their behavioral intentions. Hu et al. [84] emphasized the importance of top management participation in influencing employee behaviors. Dewettinck and van Ameijde [85] further highlighted the mediating roles of perceived behavioral integrity and psychological empowerment. Additional studies by Lynn McFarlane et al. [56] confirm the impact of job satisfaction and top management actions on behavioral intentions, with van Lill et al. [86] pointing to the importance of goal-setting styles and distinguishing between attitudes and behaviors.

H11: 

Attitude mediates the relationship between top management support and behavioral intention.

2.3.12. Facilitating Conditions, Behavioral Intention, and Actual Behavior

Behavioral intention serves as a critical mediator in the relationship between facilitating conditions and actual behavior. Gardner et al. [59] found that this mediation depends on the strength and direction of intentions, as well as the level of intention formation and specific goals. Conner et al. [61] further identified that the extent to which intentions are based on anticipated affective reactions moderates this relationship. Additionally, Sniehotta et al. [87] highlighted the roles of action planning, self-efficacy, and action control as mediators that link intentions to behavior. These findings collectively underline the significance of behavioral intention as a bridge between facilitating conditions and actual behavior.

H12: 

Behavioral intention mediates the relationship between facilitating conditions and actual behavior.

By blending the UTAUT’s core predictors with PIIT and the TRI, this study provides a robust and context-sensitive framework that addresses limitations in the original UTAUT model. This comprehensive approach offers a nuanced understanding of the dynamic interplay between institutional conditions, social influences, and individual predispositions in shaping the acceptance and use of AI technologies. It thus offers valuable insights into the factors driving AI adoption among academic librarians (

Table 1. Summary of hypotheses and expected relationships.

Hypothesis	Relationship	Supporting Literature	Expected Direction
H1	Performance Expectancy → Behavioral Intention	[40,41,42,44]	Positive
H2	Effort Expectancy → Behavioral Intention	[47,48,49]	Positive
H3	Social Influence → Behavioral Intention	[52,53,88]	Positive
H4	Top Management Support → Behavioral Intention	[21,56,89]	Positive
H5	Facilitating Conditions → Behavioral Intention	[59,60,61]	Positive
H6	Facilitating Conditions → Actual Behavior	[64,65]	Positive
H7	Technological Innovativeness × Behavioral Intention → Actual Behavior	[70,72,73]	Moderating Effect
H8	Attitude mediates Performance Expectancy → Behavioral Intention	[76,77]	Mediating Effect
H9	Attitude mediates Effort Expectancy → Behavioral Intention	[47,60,81,82]	Mediating Effect
H10	Attitude mediates Social Influence → Behavioral Intention	[63,82]	Mediating Effect
H11	Attitude mediates Top Management Support → Behavioral Intention	[84,85]	Mediating Effect
H12	Behavioral Intention mediates Facilitating Conditions → Actual Behavior	[59,61,87]	Mediating Effect

), filling a critical gap in the existing literature.

3. Research Methodology

3.1. Research Design

Building upon frameworks from previous research on internet banking (IB) acceptance in Lebanon [90] and AI chatbot adoption among research scholars using the Unified Theory of Acceptance and Use of Technology (UTAUT) [32], this study employs a quantitative approach to explore the adoption of AI technologies among academic librarians in China.

A cross-sectional survey design was implemented to collect data from academic librarians across various universities in China. The study targeted librarians who actively engage with AI technologies in their professional roles. Given the increasing integration of AI tools in academic institutions, it was essential to include participants with direct exposure to AI applications in library environments.

A purposive sampling method was adopted to reach librarians working in institutions where AI adoption is emerging or actively being explored. The study primarily focused on librarians from major Chinese universities, ensuring that participants had the necessary knowledge and experience with AI-based tools in their professional activities.

Data collection was conducted over two months using a structured questionnaire consisting of 32 validated items, designed to capture demographic attributes and theoretical constructs related to AI adoption. To enhance response rates and diversity, the survey was distributed through multiple channels:

• In-person distribution at major university libraries, allowing for direct engagement with participants.
• Institutional email invitations sent to professional librarian networks to ensure wider participation.
• AI and Library Technology Discussion Forums, where academic librarians actively discuss AI-related advancements in library management.

To address potential non-response bias, multiple follow-up reminders were sent via email, and library administrators were encouraged to facilitate participation. Additionally, responses were carefully reviewed to ensure data integrity, with incomplete or disengaged submissions systematically filtered out.

3.2. Participants

The survey gathered responses from 340 academic librarians (

Table 2. Demographic analysis.

Demographic Factor	Categories	Frequency	Percentage
Gender	Male	180	52.9
	Female	160	47.1
Age Group	18–25	45	13.2
	26–35	95	27.9
	36–45	110	32.4
	46–55	65	19.1
	56–65	25	7.4
Educational Level	Bachelor’s Degree	110	32.4
	Master’s Degree	180	52.9
	Doctoral Degree	50	14.7
Professional Title	Librarian	140	41.2
	Senior Librarian	90	26.5
	Library Manager/Director	70	20.6
	Research Librarian	30	8.8
	Technical Librarian	10	2.9
Years of Experience	Less than 1 year	20	5.9
	1–5 years	75	22.1
	6–10 years	95	27.9
	11–20 years	100	29.4
	More than 20 years	50	14.7
Type of Institution	Public University	240	70.6
	Private University	70	20.6
	Government Research Library	30	8.8
Familiarity with AI	Not Familiar	34	10.0
	Somewhat Familiar	68	20.0
	Moderately Familiar	102	30.0
	Very Familiar	95	27.9
	Extremely Familiar	41	12.1
Region	Eastern China	120	35.3
	Southern China	80	23.5
	Western China	60	17.6
	Northern China	50	14.7
	Central China	30	8.9

) across various universities in China, achieving a response rate of approximately 70% after data screening. The respondents represented a balanced gender composition, with a slight male majority (approximately 53% male, 47% female), suggesting minimal gender-related bias in technology adoption attitudes. Age-wise, most respondents fell within the 36–45-year bracket, comprising experienced mid-career professionals likely balancing traditional library practices with newer AI demands, followed closely by younger early-career (26–35) and older late-career (46–55) librarians.

Educationally, the sample primarily consisted of individuals holding advanced degrees, predominantly master’s degrees (about 53%), followed by those with bachelor’s (32%) and doctoral degrees (around 15%). Such educational diversity highlights varying analytical capacities and potential differences in receptivity to AI technologies across librarian roles. Professionally, participants ranged from frontline librarians (approximately 41%), senior librarians (27%), library managers, and directors (21%) to specialized roles such as research and technical librarians (combined approximately 12%). This mix provided a comprehensive range of operational and strategic perspectives on AI adoption.

Respondents also varied significantly in professional experience, from early-career individuals with less than five years (28%) to highly experienced librarians with over 20 years of service (approximately 15%). Most notably, librarians with substantial experience of 11–20 years formed the largest group (29%), indicating strong institutional familiarity but also potential challenges regarding resistance to technological changes.

Institutional affiliation was largely skewed toward public universities (approximately 71%), with fewer respondents from private universities (21%) and government research libraries (9%), reflecting institutional variations in resources and organizational culture influencing AI integration. Familiarity with AI varied substantially; around 30% considered themselves moderately familiar, while approximately 28% identified as very familiar, indicating substantial yet uneven AI knowledge. About 12% identified as extremely familiar and likely serve as institutional AI champions, whereas a noticeable minority (10%) expressed minimal familiarity, indicating clear training opportunities.

Geographically, responses reflected regional variations in technological infrastructure and economic development, with most participants from the economically advanced Eastern (35%) and Southern China (24%). Western (18%), Northern (15%), and Central China (9%) represented smaller shares, underlining potential regional disparities affecting AI adoption readiness and resource availability.

Overall, these demographic factors, including gender, age, education, experience level, institutional type, familiarity with AI, and geographic location, provide valuable insights into the multifaceted landscape influencing AI adoption among academic librarians in China. While the sample size (340) is sufficient for robust Structural Equation Modeling (SEM) analysis—exceeding standard thresholds (minimum 200 respondents)—it nonetheless represents a limitation in terms of broader generalizability, particularly given uneven regional representation and the under-representation of librarians from private and governmental institutions.

3.3. Measurement of Items

In preparing the measurement items (Appendix A) for this study on the adoption of artificial intelligence (AI) by academic librarians in China, each construct was clearly defined and operationalized based on established research to ensure the survey accurately captured the aspects of AI adoption pertinent to academic libraries. Performance expectancy, representing the belief that AI improves job performance, was measured using items adapted from Venkatesh et al. [13] and Kijsanayotin et al. [91]. These items assessed the perceived benefits of AI in enhancing efficiency and productivity in library operations. Similarly, effort expectancy, concerning the ease of use of AI technologies, drew from Venkatesh et al. [13], focused on how user-friendly and accessible these technologies are for librarians.

Social influence, which reflects the impact of others’ opinions on an individual’s intention to use AI, utilized items from Venkatesh et al. [31] and Kijsanayotin et al. [91]. These items evaluated the influence of colleagues and other key figures in the academic community on librarians’ decisions to adopt AI. Facilitating conditions, defined as the organizational and technical infrastructure that supports AI usage, included items adapted from Ifinedo [92]. This construct measured the availability of resources and institutional support necessary for effective AI integration in library settings.

Technological innovativeness, capturing an individual’s readiness to adopt new technologies, was operationalized using definitions from R. Agarwal and Prasad [33] and Yi et al. [38]. These items examined librarians’ willingness to explore and adopt emerging technologies ahead of their peers, emphasizing the personal initiative in technology adoption. Behavioral intention and actual use of AI, key indicators of technology adoption, were measured through items sourced from Venkatesh et al. [13]. These constructs assessed librarians’ future plans to use AI and their current engagement with AI technologies.

Finally, attitude toward AI was defined based on Chatterjee et al. [93], focusing on librarians’ positive or negative perceptions of AI in their professional roles. This included their overall sentiments regarding AI’s impact on their responsibilities and professional activities.

The study employed a seven-point Likert scale (1 = strongly disagree to 7 = strongly agree) to evaluate each construct in the research model. Actual AI behavior was captured through self-reported data, as direct system-use logging was not feasible. Respondents provided either oral consent (for in-person surveys) or email consent (for online responses). The questionnaire, originally developed in English, was translated into Chinese by a professional translator and then back-translated into English to ensure accuracy.

Before the main study, a pilot test with 30 participants was conducted to refine survey items, establish content validity, and test reliability. Feedback from the pilot study informed necessary adjustments to the questionnaire, ensuring its suitability for the main study.

4. Analysis and Findings

4.1. Descriptive Analysis

The data presented in Figure 2 across various tables provide detailed statistical validation for this study investigating the adoption of AI technologies by academic librarians in China. These statistics show that the measurement model (

Table 3. Measurement model statistics.

Construct	Code	OL	VIF	CA	CR	AVE
AB	AB1	0.826	2.001	0.858	0.904	0.702
	AB2	0.874	2.391
	AB3	0.826	1.881
	AB4	0.823	1.836
ATT	ATT1	0.871	1.942	0.828	0.897	0.743
	ATT2	0.838	1.815
	ATT3	0.877	1.910
BI	BI1	0.711	1.425	0.862	0.907	0.711
	BI2	0.879	2.496
	BI3	0.890	2.837
	BI4	0.880	2.632
EE	EE1	0.868	2.201	0.883	0.919	0.740
	EE2	0.828	1.998
	EE3	0.883	2.611
	EE4	0.860	2.422
FC	FC1	0.826	1.941	0.847	0.898	0.687
	FC2	0.869	2.407
	FC3	0.779	1.645
	FC4	0.837	2.017
PE	PE1	0.774	1.546	0.812	0.888	0.726
	PE2	0.890	2.029
	PE3	0.887	2.024
SI	SI1	0.908	3.158	0.916	0.941	0.799
	SI2	0.893	2.976
	SI3	0.880	2.643
	SI4	0.895	2.927
TI	TI1	0.896	2.072	0.796	0.880	0.711
	TI2	0.811	1.642
	TI3	0.820	1.641
TMS	TMS1	0.880	2.358	0.872	0.912	0.721
	TMS2	0.855	2.227
	TMS3	0.798	1.934
	TMS4	0.862	2.239

OL—Outer Loading; VIF—Variance Inflation Factor; CA—Cronbach’s Alpha; CR—Composite Reliability; AVE—Average Variance Extracted; AB—actual behavior; ATT—attitude; BI—behavioral intention; EE—effort expectancy; FC—facilitating condition; PE—performance expectancy; SI—social influence; TI—technological innovativeness; TMS—top management support.

) is robust, with each construct demonstrating strong item reliability and distinctiveness.

Our results confirm that all survey questions strongly relate to their intended variables, ensuring that the model measures AI adoption correctly. Additionally, we checked for potential issues where different predictors might be too closely related (multicollinearity) and found that all values were within acceptable limits, confirming that each factor contributes uniquely to the model [94]. Both Cronbach’s Alpha (CA) and Composite Reliability (CR) values exceed the 0.7 threshold, suggesting reliable internal consistency. Furthermore, Average Variance Extracted (AVE) values are above 0.5 for all constructs, confirming that the majority of variance in the indicators is accounted for by the constructs they aim to measure [95].

Discriminant validity, assessed through the Heterotrait–Monotrait Ratio (HTMT) and Fornell–Larcker Criterion (FLC), demonstrates that each construct is unique and captures distinct aspects of AI technology adoption among librarians (

Table 4. Discriminant validity (HTMT).

Construct	AB	ATT	BI	EE	FC	PE	SI	TI	TMS	TI × BI
AB
ATT	0.808
BI	0.784	0.828
EE	0.765	0.811	0.740
FC	0.687	0.634	0.826	0.589
PE	0.747	0.848	0.775	0.728	0.692
SI	0.824	0.776	0.783	0.807	0.695	0.724
TI	0.841	0.837	0.838	0.806	0.777	0.798	0.770
TMS	0.778	0.684	0.730	0.656	0.795	0.728	0.818	0.836
TI × BI	0.442	0.371	0.384	0.430	0.239	0.289	0.412	0.486	0.295

and

Table 5. Discriminant validity (FLC).

Construct	AB	ATT	BI	EE	FC	PE	SI	TI	TMS
AB	0.838
ATT	0.687	0.862
BI	0.678	0.708	0.843
EE	0.670	0.698	0.656	0.860
FC	0.587	0.534	0.698	0.517	0.829
PE	0.635	0.719	0.661	0.623	0.587	0.852
SI	0.731	0.683	0.695	0.730	0.614	0.644	0.894
TI	0.698	0.716	0.697	0.765	0.636	0.660	0.832	0.843
TMS	0.682	0.594	0.639	0.585	0.689	0.632	0.739	0.701	0.849

). HTMT values are below the 0.90 threshold, and FLC values show that the square root of AVE for each construct is greater than the correlations among constructs. This pattern underscores that constructs share more variance with their own measures than with others, ensuring that they are distinctly measuring different facets of the phenomenon under study [96].

Overall, these tables underscore a statistically robust framework, confirming that the constructs of performance expectancy, effort expectancy, social influence, facilitating conditions, technological innovativeness, and top management support are well defined. They effectively measure various facets influencing librarians’ attitudes, intentions, and behaviors toward AI adoption. This ensures that the study’s findings are reliable and valid, providing a solid basis for understanding how different factors influence AI adoption in academic libraries in China. These metrics and their interpretations follow established guidelines in psychometric evaluations as recommended in academic research, ensuring the methodological rigor of the study [94].

4.2. Hypothesis Testing and Discussion

Our model successfully explains a large portion of AI adoption behavior among academic librarians. Specifically, the model predicts 67.7% of variations in behavioral intention (BI), 63.7% in attitude (ATT), and 57.4% in actual AI use (AB). This indicates strong explanatory power. Additionally, we tested whether our model can make accurate predictions for new data using Q² Predict values, which all exceed 0.56, further confirming its reliability. These values are 0.564 for AB, 0.623 for ATT, and 0.633 for BI (

Table 6. Model fit statistics.

Construct	R2	R2 Adjusted	Q2 Predict	RMSE	MAE
AB	0.574	0.571	0.564	0.664	0.493
ATT	0.637	0.634	0.623	0.618	0.436
BI	0.677	0.673	0.633	0.610	0.393

). Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) are used to assess the model’s prediction accuracy, with lower values indicating better performance. For AB, the RMSE is 0.664 and the MAE is 0.493; for ATT, the RMSE is 0.618 and the MAE is 0.436; and for BI, the RMSE is 0.610 and the MAE is 0.393. These statistics collectively illustrate that the model is well fitted, with accurate predictions and reliable adjustments for variations within the data, providing a solid statistical foundation for understanding how academic librarians perceive and engage with AI technologies.

The structural model (Figure 3) results present a comprehensive understanding of the adoption of artificial intelligence (AI) by academic librarians, shedding light on key drivers, inhibitors, and contextual nuances that define their engagement with emerging technologies. The results show that (

Table 7. Structural model statistics.

Hypothesis Path		Original Sample	Sample Mean	Standard Deviation	T Statistics	p Values	f2	Support
H1	PE → BI	0.091	0.095	0.053	1.724	0.085	0.010	No
H2	EE → BI	0.123	0.122	0.058	2.125	0.034	0.018	Yes
H3	SI → BI	0.151	0.152	0.067	2.238	0.025	0.021	Yes
H4	TMS → BI	0.000	0.001	0.065	0.004	0.997	0.000	No
H5	FC → BI	0.345	0.343	0.051	6.735	0.000	0.176	Yes
H6	FC → AB	0.124	0.122	0.050	2.468	0.014	0.017	Yes
H7	TI × BI → AB	−0.069	−0.068	0.023	2.996	0.003	0.026	Yes
H8	PE → ATT → BI	0.105	0.103	0.026	3.988	0.000		Yes
H9	EE → ATT → BI	0.079	0.079	0.029	2.764	0.006		Yes
H10	SI → ATT → BI	0.052	0.052	0.022	2.393	0.017		Yes
H11	TMS → ATT → BI	0.009	0.009	0.014	0.609	0.542		No
H12	FC → BI → AB	0.102	0.103	0.023	4.513	0.000		Yes

) a substantial proportion of the variance in behavioral intention (BI), attitude (ATT), and actual behavior (AB) is well accounted for by the predictors.

One surprising finding was that performance expectancy (the belief that AI would improve job performance) did not significantly influence librarians’ decisions to adopt AI (β = 0.091, p = 0.085). This challenges previous research, which often shows that users adopt new technologies if they expect them to enhance their efficiency. However, our results suggest that academic librarians may value other factors—such as ease of use and workplace support—over performance benefits when deciding to use AI. Librarians often engage in specialized tasks, such as curating archival materials, providing expert research support, or cataloging rare collections, where the direct relevance of AI may not be readily apparent. This finding resonates with studies such as Yakubu et al. [45], who argue that in high-autonomy, knowledge-intensive settings, individuals are often influenced by factors other than purely performance-related benefits. Moreover, Alzahrani [46] highlight how perceived misalignment between a technology’s capabilities and users’ job roles can dilute the motivational effects of PE. For academic librarians, skepticism about AI’s ability to enhance their core professional functions may diminish its perceived value, particularly if its applications seem superficial or geared toward general rather than specialized tasks.

Top management support (TMS) similarly demonstrates a non-significant direct influence on BI (p = 0.997), which stands in stark contrast to findings from corporate studies, where managerial endorsements often play a pivotal role in shaping employee attitudes and technology adoption behaviors (Lee et al. [21]). However, this outcome aligns with studies suggesting that high-autonomy professionals, such as librarians, researchers, and healthcare providers, place greater emphasis on peer networks and contextual enablers than on hierarchical directives (Lynn McFarlane et al. [56]). In academic libraries, professional norms, independent decision-making, and the influence of trusted colleagues may overshadow the impact of managerial support. This suggests that while top management support can be an enabler, its effectiveness is contingent on its alignment with librarians’ intrinsic motivations and day-to-day operational needs. Visible but misaligned managerial efforts—such as advocating AI without offering adequate training or practical support—are unlikely to resonate with this professional group.

On the other hand, facilitating conditions (FCs) emerged as a critical factor influencing both BI (p < 0.001) and AB (p = 0.014). These findings align with a well-established body of research highlighting the importance of resource-rich environments in bridging the intention–behavior gap [64]. When librarians have access to technical resources, training programs, and robust IT infrastructure, their intentions to use AI are more likely to translate into sustained usage. This underscores the critical role of institutional investment in creating environments that not only encourage but actively enable technology adoption. Furthermore, the importance of habitual behaviors, as emphasized by Gardner et al. [59] and Eskenazi and Neumaier [68], suggests that sustained access to enabling conditions can help librarians overcome inertia or resistance, allowing them to integrate AI into their workflows over time.

Effort expectancy (EE) and social influence (SI) also demonstrate significant positive effects on BI (p = 0.034 and 0.025, respectively), emphasizing the importance of ease of use and peer validation. These findings reinforce the conclusions of Moya et al. [47], who argue that technologies perceived as user-friendly and validated by respected peers are more likely to be adopted. Sheeran et al. [49] further extended this perspective by suggesting that social norms and perceived simplicity can outweigh performance considerations, particularly in contexts where users are unfamiliar with or hesitant about new technologies. For librarians, the endorsement of trusted colleagues and the intuitive design of AI interfaces appear to be decisive in shaping adoption intentions.

An especially nuanced finding concerns the moderating role of technological innovativeness (TI) in the relationship between BI and AB (p = 0.003). While high-TI individuals are often early adopters and champions of innovation [33], the findings suggest that excessive innovativeness can inadvertently impede consistent and meaningful technology usage. As Fu and Elliott [71] argue, overly innovative users may focus on exploring advanced features or multiple platforms without integrating these tools into cohesive workflows, thereby hindering broader organizational adoption. This finding underscores the importance of balancing innovation with usability, ensuring that early adopters are guided to maximize the practical impact of their experimentation. Structured pilot programs, phased rollouts, and targeted mentoring from technologically innovative librarians to their peers could mitigate the risks of over-innovation while leveraging their enthusiasm for broader organizational benefit.

The mediating role of attitude (ATT) further highlights its centrality in bridging contextual, individual, and social factors with BI. Significant mediating effects are observed for the relationships of PE, EE, and SI with BI, consistent with King [76], who underscores the importance of attitude formation in driving sustained technology adoption. This finding suggests that even when direct effects—such as those of PE—are muted, positive attitudes toward AI can compensate by reinforcing intentions. Librarians with favorable attitudes, shaped by peer influence, ease of use, and exposure to AI’s potential benefits, are more likely to develop strong behavioral intentions. Conversely, the lack of significant mediation for TMS emphasizes that managerial support alone is insufficient to shape attitudes unless coupled with contextual enablers that align with librarians’ professional values.

The findings also challenge traditional assumptions about the linear pathways to the adoption of technology, particularly in high-autonomy professional environments. In corporate settings, where hierarchical structures and managerial influence often dominate, top management support and performance expectancy are reliable predictors of adoption. However, in academic libraries, a more nuanced interplay of facilitating conditions, peer influence, and individual attitudes appears to take precedence. This divergence underscores the importance of tailoring technology adoption frameworks to the specific socio-professional contexts of the target population.

In summary, our findings suggest that providing AI training, improving workplace resources, and ensuring ease of use are the most effective strategies for encouraging AI adoption among academic librarians. Interestingly, top management support and expected performance improvements were less important than ease of use and peer influence. These insights can help universities and library managers design better AI implementation strategies.

5. The Implications of This Study

5.1. Theoretical Implications

This study extends the Unified Theory of Acceptance and Use of Technology (UTAUT) by incorporating constructs from Personal Innovativeness in IT (PIIT) and the Technology Readiness Index (TRI), offering innovative theoretical insights into technology adoption. By addressing the limitations of the UTAUT in accounting for individual-level variability, this study demonstrates how personal traits such as technological innovativeness and readiness influence behavioral intentions and actual usage. The findings challenge traditional assumptions of the UTAUT, particularly regarding the limited impact of performance expectancy and top management support on behavioral intentions, suggesting that individual readiness and innovativeness may attenuate these effects.

PIIT adds depth to the framework by capturing the proactive behaviors of early adopters who can accelerate the diffusion of AI technologies by influencing their peers. However, this study reveals that excessive innovativeness can hinder actual usage due to overcomplexity or lack of integration support, highlighting a nuanced role for technological innovativeness. The TRI further enriches the model by illustrating how readiness dimensions—optimism, discomfort, and insecurity—interact with facilitating conditions and effort expectancy to shape technology adoption. For instance, individuals with high optimism are more likely to perceive AI as beneficial and user-friendly, while those with high discomfort may require targeted interventions to overcome perceived barriers.

This study’s theoretical contribution lies in demonstrating the interplay between systemic predictors and individual readiness factors. It provides a foundation for future research to explore dynamic, context-sensitive interactions, such as how readiness and innovativeness evolve over time or how they interact with cultural and organizational factors in technology adoption. Additionally, the insights into the counterproductive effects of over-innovation open avenues for refining the application of the TRI and PIIT to other emerging technologies.

5.2. Practical Implementation

The findings offer innovative and specific strategies to enhance AI adoption among academic librarians. First, rather than generic training programs, libraries should develop interactive AI sandboxes—digital platforms where librarians can safely explore and experiment with AI tools without fear of errors or judgment. These sandboxes can simulate real-world applications, allowing users to experience the practical benefits of AI and build confidence in its usage. Additionally, personalized learning pathways, informed by librarians’ levels of readiness and technological familiarity, can ensure tailored support. For instance, librarians with low TRI scores on optimism or high discomfort may benefit from immersive, gamified training sessions that progressively reduce perceived complexity.

Fostering positive attitudes requires innovative communication strategies. Libraries could establish AI immersion labs—collaborative spaces where librarians and users co-create solutions using AI. These labs not only demystify AI’s potential but also empower librarians to see themselves as co-creators of AI-enhanced services. Recognition programs, such as a “Librarian Innovator of the Month” award, can further incentivize early adopters to advocate AI within their institutions. This approach leverages PIIT by positioning early adopters as influencers who can drive cultural shifts toward greater AI acceptance.

Optimizing facilitating conditions goes beyond infrastructure upgrades. Libraries should implement real-time AI support assistants, such as chatbots tailored to librarians’ workflows, to provide immediate solutions to technical challenges. Additionally, libraries could adopt adaptive AI systems that adjust their complexity based on user proficiency, ensuring that both novice and advanced users can engage effectively. These systems could incorporate dynamic interfaces that gradually introduce advanced features as users gain familiarity.

Addressing the nuanced effects of technological innovativeness requires a staged rollout of AI tools. Libraries could implement pilot programs that begin with basic AI functionalities, collecting feedback to refine and expand features over time. For highly innovative librarians, leadership could establish AI innovation hubs—dedicated teams tasked with testing cutting-edge AI tools and mentoring others. Meanwhile, those less technologically inclined could focus on integrating foundational AI applications tailored to their specific tasks.

Finally, while top management support showed limited direct effects, its role remains vital in signaling institutional commitment. Management should adopt participatory AI roadmaps, where librarians contribute to planning and decision-making, ensuring alignment between AI initiatives and their professional realities. Visible allocation of resources, such as dedicated AI budgets and staff, reinforces the organization’s commitment to technological advancement.

By implementing these targeted, innovative strategies, libraries can create a culture that supports AI adoption, fosters readiness, and addresses barriers at both systemic and individual levels, ensuring effective and sustainable integration.

5.3. Conclusion and Future Research Directions

This study significantly advances the understanding of artificial intelligence (AI) adoption in academic libraries by extending the Unified Theory of Acceptance and Use of Technology (UTAUT) to include constructs from Personal Innovativeness in IT (PIIT) and the Technology Readiness Index (TRI). The integration of these elements offers a holistic framework that accounts for both organizational and individual factors, highlighting the pivotal roles of facilitating conditions, effort expectancy, and readiness in shaping behavioral intentions and actual AI usage. Notably, the limited direct effects of performance expectancy and top management support underscore the influence of individual readiness and personal innovativeness, challenging traditional assumptions within technology acceptance research.

From a practical standpoint, these findings emphasize the value of AI implementation strategies such as AI “sandboxes”, adaptive AI systems, and immersive learning environments that align with librarians’ specific professional needs and varying levels of technological readiness. Such strategies address both the technical and psychological dimensions of AI adoption, motivating librarians to explore and integrate novel tools while fostering a positive, hands-on learning culture. However, several limitations should be considered when interpreting these findings.

One notable limitation is the potential presence of confounding variables. This study did not explicitly control for individual differences in digital literacy, prior AI exposure, or institutional policies regarding AI use. These factors may influence librarians’ attitudes and behavioral intentions, potentially affecting the relationships tested in the model. Additionally, job roles and responsibilities may shape AI adoption differently, as technical librarians may perceive AI’s relevance differently from reference librarians or administrators. Future research should incorporate these factors as control variables to provide a more precise understanding of AI adoption in academic libraries.

Another limitation concerns measurement constraints, as this study relied on self-reported survey data. Self-reports can introduce biases such as social desirability bias, recall bias, or overestimation of AI engagement. While self-reported behavioral intentions provide valuable insights, they do not always translate into actual behavior. Future studies should consider incorporating objective behavioral measures, such as system-generated AI usage logs, observational methods, or experimental interventions, to validate self-reported findings and provide a more robust assessment of AI adoption.

The cross-sectional nature of this study also presents a constraint, as it captures perceptions at a single point in time. AI adoption is an evolving process influenced by organizational changes, technological advancements, and shifts in user attitudes. A longitudinal research approach would offer a more comprehensive understanding of how librarians’ AI adoption behaviors develop over time, shedding light on how facilitating conditions and personal innovativeness interact to support sustained technology use.

Although this study focuses on academic librarians in China, cultural and institutional factors may influence AI adoption in different contexts. China’s institutional structures, policies, and technological infrastructure may not be directly comparable to those of other countries, potentially limiting the generalizability of findings. Expanding the geographical scope of future research to include librarians from diverse regions with different cultural and organizational settings would help identify how variations in AI maturity levels, governance models, and user expectations impact adoption behavior. Comparative analyses across countries or regions could provide deeper insights into how cultural dimensions, such as collectivism versus individualism or centralized versus decentralized management, influence AI-related behaviors in academic libraries.

Additionally, while this study examines AI adoption at a general level, AI applications in libraries vary widely, including chatbots, intelligent search engines, predictive analytics, and automation tools. Future research should differentiate AI adoption patterns based on specific AI functionalities, assessing how different applications contribute to service efficiency, decision-making, and knowledge management in academic libraries. This would provide a more granular understanding of the facilitators and barriers associated with different AI technologies.

Finally, while this study primarily examines the Chinese context, its findings have broader applicability for institutions integrating emerging AI technologies, including generative AI. As AI continues to evolve, future research should track its influence on task allocation, professional roles, and the overall innovation climate in academic libraries. Longitudinal and cross-cultural examinations could provide valuable insights into how variations in organizational culture, resource availability, and professional norms interact with rapidly changing technological landscapes.

By integrating individual readiness and innovativeness into existing technology acceptance models, this study provides a nuanced perspective on AI adoption in academic libraries. The findings offer actionable insights for library administrators, information professionals, and policymakers seeking to capitalize on AI’s potential while navigating the complexities of organizational culture and professional autonomy. Moving forward, addressing the limitations of self-reported data, adopting longitudinal methods, expanding the geographical scope, and exploring the transformative effects of advanced AI tools will be essential for enriching our collective understanding of AI’s role in the evolution of academic librarianship.