Research on AI-driven innovations and services in academic libraries: A bibliometric and systematic literature review

The number of documents stayed steady and comparatively small; 2014 saw one, 2015 and 2016 saw three, and 2017 saw four. Eleven papers with 116 citations were published during this time, accounting for 0.06% of the total output in this field of study. The most cited articles were “Content based image retrieval based on geo-location driven image tagging on the social Web” (Memon et al., 2014), published in 2014 with 25 citations, followed by an article published in 2016 by (Khabsa et al., 2016), with 21 citations in all ACM/IEEE Joint Conference on Digital Libraries proceedings. Exploratory studies marked this period focused on developing theoretical frameworks for academic libraries (Dabbour & Kott, 2017; Ellern et al., 2015; Zhu et al., 2016), academic search (Khabsa et al., 2016), and proof-of-concept applications in libraries, i.e., linked open data (Sateli et al., 2017), visual analytics (Ventocilla & Riveiro, 2017), behavior analytics (Fernandez & Ponnusamy, 2015), and using AI applications and tools in libraries, for example, recommendation system (Beel et al., 2017; Chadha & Kaur, 2015), data mining, machine learning, and natural language processing (Bohra et al., 2015; Sateli et al., 2017). This foundational stage was critical for laying the groundwork and creating a foundation for subsequent growth.

There is a gradual increase in the number of documents, with 6 in 2018, 7 in 2019, and peaking at 16 in 2020 before slightly dropping to 11 in 2021. Forty papers were published. The most cited article was “The Intelligent Library: Thought Leaders’ Views on the Likely Impact of Artificial Intelligence on Academic Libraries” (Cox, Pinfield, et al., 2019), published in 2019 with 141 citations in Library Hi Tech, followed by the article “UTAUT as a Model for Understanding Intention to Adopt AI and Related Technologies among Librarians” (Andrews et al., 2021) published in 2021 with 77 citations in the Journal of Academic Librarianship. During this period, research likely expanded into perceptions and awareness of artificial intelligence in managing academic libraries (Abayomi et al., 2021; Hervieux & Wheatley, 2021), including AI tools and perspectives of academic librarians (Ali et al., 2020; Wheatley & Hervieux, 2020). There are early natural language processing tools (Wang et al., 2020), linked data integrated with AI in the transformation of library services (Schreur, 2020), enhanced digital libraries with book recommendation systems (Nugraha et al., 2020), and data mining approach that emerged in the library collection and services (Friedman et al., 2021; Giannopoulou & Mitrou, 2020; Li & Daoutis, 2021; Patel et al., 2020; Wang et al., 2020).

Although the number of documents slightly dropped to 11 in 2021, the publications addressed critical topics, reflecting the growing interest in and adoption of advanced technologies within academic and public libraries globally. One of the most highly cited works, “Awareness and Perception of the Artificial Intelligence in the Management of Academic Libraries in Nigeria” (Abayomi et al., 2021), has 23 citations. This study explored how librarians in Nigeria perceive and adopt AI technologies, highlighting gaps in awareness and challenges to implementation. Its high citation count underscores the relevance of AI adoption in developing regions. Similarly, the 42-citation paper “Perceptions of Artificial Intelligence: A Survey of Academic Librarians in Canada and the United States” (Hervieux & Wheatley, 2021) offered information on how academic librarians in North America were utilizing AI in their work. This study is highly cited because it highlighted the value of virtual assistant technologies and training in enhancing library services. The paper “Using AI/Machine Learning to Extract Data from Japanese American Confinement Records” discussed the use of AI in digital collections and archives (Friedman et al., 2021). This study illustrated AI’s multidisciplinary reach in libraries by showcasing the real-world implementation of machine learning to enhance archival data accessibility.

The 2021 publications place a particular emphasis on using AI, data mining, and machine learning to address emerging technologies in academic libraries. These studies concentrate on awareness, perceptions and practical applications, highlighting the field’s transition to digital transformation and user-centered innovations. The high citation counts of individual works demonstrated their major impact on subsequent studies and the rising understanding of technology’s role in changing library systems around the world.

Between 2022 and 2023, the number of research articles grew rapidly from 28 to 39 and 71, respectively. The paper “Artificial Intelligence (AI) Library Services Innovative Conceptual Framework for the Digital Transformation of Academic Education” (Okunlaya et al., 2022), which had 96 citations, is one of the highlights. There are also 92 citations in the paper “The CLEAR path: A framework for enhancing information literacy through prompt engineering” (Lo, 2023). Research on AI, data mining, machine learning, and cutting-edge technology for libraries and information management achieved tremendous breakthroughs in 2022. Twenty-eight papers covering a range of topics, from digital transformation strategies to AI-powered library services, have added to this growing body of knowledge. The fact that these papers received 366 citations shows how important and current they represent.

The significance of investigating how AI might change academic libraries was underlined by one of the most frequently cited publications in (Okunlaya et al., 2022). This paper established a thorough framework for applying AI to enhance library services, engage with initiatives for digital transformation, and address evolving needs in education. Thirty-seven citations were included in another interesting study that examined librarians’ attitudes and capability to use AI technologies (Yoon et al., 2022). These results made AI a valuable tool for professionals and researchers, highlighting the need for training and addressing reservations about integrating it into library operations. With 39 citations, (Kaushal & Yadav, 2022) showed increasing interest in conversational AI and its uses to increase service efficiency and optimize user interactions. As libraries adapt to the needs of patrons in the digital age, chatbots are becoming popular. This study concentrated on the practical implications of implementing them in libraries.

The articles by (Ajani et al., 2022) (25 citations) and (Rodriguez & Mune, 2022) (25 citations) illustrated how AI technologies are applicable globally in a range of institutional and geographic situations. These studies highlighted the difficulties and advantages of applying AI solutions, automated processes, and virtual reference tools in educational contexts. Additionally, by offering state-of-the-art computational techniques for data analysis, content management, and user interaction, the article in (Wang, 2022) explored the innovative application of data mining and expanded the boundaries of traditional academic libraries (Thornley et al., 2022). Emerging concerns included the moral implications of applying AI and cutting-edge information extraction methods for conserving cultural collections, as illustrated in (Bergamaschi et al., 2022).

Publication in AI and its library applications increased significantly in 2023. Thirty-nine articles (346 citations) showed a wide range of interest and influence in areas from ChatGPT applications to AI integration in library workflows. For example, the 26-citation study in (Houston & Corrado, 2023) highlighted its impact on the discussion of generative AI in academic library environments. Similarly, (Brzustowicz, 2023) (9 citations) investigated how AI could transform cataloging systems. Publications from 2023 also emphasized the beneficial use of AI in libraries. The contribution of ChatGPT to enhancing educational environments was examined in another highly cited paper in (Vargas-Murillo et al., 2023) (28 citations), indicating its multidisciplinary applicability in libraries and education. In (Narendra et al., 2025) and (Jha, 2023), the potential of AI to enhance user services and expedite library operations was examined. These studies demonstrated how AI can solve operational problems and increase user satisfaction. For example, In (Clark & Lischer-Katz, 2023) highlighted the challenges associated with implementing AI in libraries. In (Sala, 2023) explored innovative uses of AI beyond traditional domains emphasizing the use of data mining and natural language processing systems to enhance the accessibility of multilingual content.

With 71 articles, research on artificial intelligence and its use in academic libraries increased significantly in 2024. The study covered a wide range of subjects, such as the use of AI in library services, user-centered innovations, ethical issues, and recent developments in generative AI tools. While some articles were highly cited, others, as is common with recent publications, have yet to receive considerable citations. The use of AI-powered adaptive systems for customized library services was a common theme (Sowmiya et al., 2024). AI and machine learning technologies can enhance library services and streamline operations when put into practice (Bhattacharya, 2024). As AI capabilities proliferate in the creative industries, this issue becomes increasingly important.

ChatGPT was one of the generative AI technologies that was discussed significantly in (Torres, 2024). Another interesting study emphasized the importance of AI literacy among librarians and the necessity of training to optimize the effective use of generative AI technology (Lo, 2024). The creative industry makes greater use of AI tools. (Liu et al., 2024) looked at the transparency and ethical concerns with AI-powered search engines, highlighting the importance of responsibility in both their creation and application. The literature makes considerable use of generative AI technologies, including ChatGPT, and explores how they could be useful for tasks like virtual assistance, instructional design, and content generation. Another significant study in (Lo, 2024) focused on the significance of AI literacy among librarians and emphasized the role of training in enhancing generative AI methods. Research on the application of AI in African academic libraries with a particular emphasis on regional and global perspectives was conducted by (Alam et al., 2024). This pattern emphasizes AI’s worldwide applicability and capacity to solve regional academic library issues. Some studies, such as (Harisanty et al., 2024) (45 citations), stood out despite the variety of topics because of their high citation counts, proving their substantial impact on the scholarly community. This work reflects the growing awareness and interest in AI among key stakeholders in the library sector.

Figure 16(a) illustrates the temporal evolution of key research terms with bubble size indicating the frequency of publications associated with each term. The years are shown on the x-axis, and the terms are listed on the y-axis. Figure 16(b) represents the frequency and importance of keywords in AI-related research for academic libraries. More significant words indicate more frequent or prominent topics in the analyzed literature.

Figure 16.

Research topics on AI applications in academic libraries (a) temporal trends (2016-2024), (b) word cloud of author keywords.

Figure 16(a) reveals that early terms of “learning systems,” “behavioral research,” and “search engines” dominated from 2016-2018, reflecting the foundational exploration of AI in user behavior, education, and information retrieval. Post-2020, “artificial intelligence,” “digital libraries,” and “data mining” emerged as dominant themes, indicating the integration of advanced AI technologies in digital transformations and analytics. The emphasis on “artificial intelligence” and “digital libraries” demonstrates the increasing role of AI in automating and enhancing library systems, including cataloging, personalized recommendations, and resource management. The emergence of terms like “data mining” and “natural language processing” reflects advancements in using AI to extract insights from extensive data and improve information retrieval.

The word cloud (Figure 16(b)) visualizes the 50 most frequent author keywords related to the intersection of artificial intelligence (AI) and academic libraries based on word frequency. Using a circular layout, larger words indicate higher occurrence. Prominent terms include “academic libraries,” “machine learning,” “artificial intelligence (AI),” “chatGPT,” “generative AI,” and “digital libraries.” These keywords capture the expanding discussion about incorporating AI-powered technologies into library services. The prevalence of “academic libraries” and “machine learning” indicates a great deal of research being done on using intelligent systems to improve library operations like user interaction, classification, and information retrieval. The presence of “chatGPT” and “generative AI” signals a recent surge in interest in large language models and their implications for knowledge access, content creation, and library instruction. This keyword distribution, viewed thematically, demonstrates how traditional library concerns like information literacy, academic integrity, and user experience are coming together with new AI technologies. “Virtual assistants,” “digital libraries,” and “information services” are some terms that highlight how libraries are changing in the context of digital transformation.

Figure 17 displays the co-occurrence network visualization generated using Biblioshiny (Bibliometirx), employing the Louvain clustering algorithm with association normalization and key parameters: 50 nodes, a minimum of 2 co-occurrences, and isolated nodes removed. The layout is automatic with dot-shaped nodes, and label overlap is suppressed for clarity. The conceptual landscape of artificial intelligence research in academic libraries, as well as thematic clusters, are revealed by this network, which offers a structural representation of how author-defined keywords co-appear in the literature.

Figure 17.

The top 50 author keywords in academic library documents about AI applications. Notably, red cluster 1 is followed by blue cluster 2, green cluster 3, purple cluster 4, and yellow cluster 5.

At the network’s core, “artificial intelligence” and “academic libraries” emerge as the most central and highly connected nodes, indicating their dominance as focal topics. Closely linked to these are terms like “machine learning,” “digital libraries,” “chatGPT,” “natural language processing,” and “information literacy,” all of which form dense, well-connected clusters. This suggests an integrative research trend in which AI technologies are being increasingly applied to enhance academic library services, ranging from intelligent search and automation to educational support and digital infrastructure. Color-coded clusters representing various thematic communities were found using the Louvain clustering algorithm. Terms like “machine learning,” “robotics,” and “automatic” are included in the red cluster around “artificial intelligence,” for example, suggesting a strong emphasis on technological applications. The green cluster reflects a current interest in conversational AI and its implications for education and digital engagement by connecting terms like “generative AI,” “ChatGPT,” and “AI literacy.” The blue cluster around “academic libraries,” “library services,” and “university libraries” reflects traditional library science concerns now being re-contextualized through AI integration.

Table 8 presents the results of the author keyword co-occurrence network, organized into five distinct clusters using the Louvain clustering algorithm. Each cluster is color-coded and includes keywords along with network centrality metrics: Betweenness, Closeness, and PageRank, which help to identify the influence and positioning of keywords within the network structure.

Cluster 1 (Red, 17 items) centers around “artificial intelligence,” which holds the highest betweenness (610.24) and PageRank (0.216), indicating its role as the most influential and connective term in the network. Other key terms include “machine learning,” “digital libraries,” “big data,” “chatbot,” and “cloud computing,” reflecting technological and systems-based applications of AI within library contexts.

Cluster 2 (Blue, seven items) is dominated by “academic libraries” (betweenness = 120.12), which serve as the thematic hub. This cluster reflects the institutional and service-related aspects of AI integration, including “library services,” “access to information,” and “university libraries.”

Cluster 3 (Green, five items) revolves around “chatGPT,” “generative AI,” and “AI literacy.” This cluster captures emerging trends and the rise of conversational AI and educational themes. The relatively high PageRank of “chatGPT” (0.041) suggests its growing prominence.

Cluster 4 (Purple, five items) includes terms like “ethics,” “libraries,” and “natural language processing (NLP),” pointing to a conceptual blend of ethical foundational and computational concerns surrounding AI use in libraries.

Cluster 5 (Yellow, seven items) captures broader and administrative themes such as “AI,” “technology,” “academic librarian,” and “research,” with moderate PageRank and betweenness values suggesting their supporting role in the overall structure.

Five distinct thematic communities are revealed by the network, which reflects the complex research environment at the nexus of academic libraries and artificial intelligence. The clusters are separate but linked, with “artificial intelligence” and “academic libraries” serving as key hubs that connect institutional implementation and technological innovation. Keywords like “chatGPT,” “machine learning,” and “AI literacy” indicate a change in the discourse from infrastructure-focused research to issues of ethics, education, and human interaction.

These findings emphasize the significance of both well-established and new research themes. The dominance of AI-related terms like “machine learning,” “automation,” and “deep learning” in the Red cluster show that the technical applications of AI are still central to scholarly attention. At the same time, the Green and Purple clusters emphasize the growing awareness of ethical and social considerations, especially regarding AI’s influence on education, information access, and professional roles in libraries. The separation of institutional and service-related terms (Blue cluster) suggests that while technological innovation is accelerating, the library profession remains grounded in service delivery, user experience, and information literacy, which are now being reframed in the context of AI.

The co-occurrence network visualization of author keywords reveals a clear and organized clustering of terms, representing various thematic communities associated with academic libraries and artificial intelligence (AI). Semantic coherence, key node centrality, and clear visual separation demonstrate these detected communities’ robustness. The central clusters-most notably those surrounding “artificial intelligence” and “academic libraries”-form the conceptual backbone of the network, with peripheral but connected nodes reflecting subthemes such as AI ethics, machine learning applications, and service-related innovations in libraries.

• Robustness of communities

  The communities are robust partly due to the Louvain clustering algorithm, which is known for its modularity optimization and suitability for complex, weighted networks. In this network, key clusters (e.g., those centered on artificial intelligence, academic libraries, and chatGPT) display a high density of internal connections and minimal ambiguity in their boundaries. High-betweenness nodes like artificial intelligence and machine learning further contribute to the clusters’ structural stability, as these terms serve as connectors that anchor related subfields.

• Sensitivity to resolution parameter changes

  In community detection algorithms like Louvain, the resolution parameter determines the granularity of clusters: lower values tend to produce fewer, broader communities, while higher values yield more, smaller clusters. While this visualization does not explicitly show outcomes under different resolution settings, it can be inferred that the core structure-especially the major hubs and their immediate neighbors, will likely remain relatively stable. Larger thematic cores such as artificial intelligence and academic libraries are unlikely to fragment, as their centrality metrics (e.g., betweenness, PageRank) and high co-occurrence with a wide range of other terms anchor them firmly within the network.

• Stability under different clustering thresholds

  The threshold used in this analysis (minimum co-occurrence of 2 and removal of isolated nodes) balances the inclusion of meaningful terms while reducing noise. If this threshold were lowered, the network would become denser with minor connections and reveal more latent or emerging topics at the cost of visual clarity and thematic distinctiveness. Conversely, raising the threshold would shrink the network and may obscure relevant but less frequent associations, potentially altering the periphery but leaving the core communities essentially unchanged. Thus, the stability of primary communities is expected to hold even under moderate changes in threshold settings.

In summary, the detected communities in the co-occurrence network demonstrate high robustness, with strong internal cohesion and logical thematic alignment. While peripheral terms and small clusters may be sensitive to resolution and threshold adjustments, the major clusters anchored by highly central nodes exhibit structural stability. It indicates that the network’s conceptual structure is sufficiently valid as a model of the literature on AI in academic library research for directing future studies or strategic research planning. This suggests that the network is a credible representation of the theoretical structure of AI literature found in university libraries that may be useful as a guide to future analyses or strategic research.

Thematic mapping and co-occurrence network analysis are two distinct bibliometric techniques that utilize author keywords to explore the intellectual structure of a research field, yet they differ significantly in their analytical focus, methodological approach, and interpretive value. Based on the co-word analysis framework, a thematic map is designed to evaluate the strategic positioning of research themes by plotting them on a two-dimensional space defined by centrality and density. While density shows the internal cohesion of the theme and suggests its stage of development, centrality measures the extent to which a theme interacts with other themes and indicates its relevance to the field as a whole. This technique divides themes into four quadrants – motor themes, basis themes, niche themes, and emerging/declining themes – to provide a macro-level view of the maturity, significance, and evolution of thematic areas within an area of research.

The thematic map (Figure 18.) was constructed using the author keywords field, applying the Louvain clustering algorithm with specific parameters: 250 keywords, a minimum cluster frequency of 5 per 1000 documents, three labels, and zero community repulsion. This visualization provides insights into the structural and strategic positioning of research topics within the intersection of academic libraries and artificial intelligence (AI) by grouping keywords into four thematic quadrants based on density (degree of development) and centrality (relevance to the field).

Figure 18.

Thematic author’s keywords map for AI applications in academic libraries

In the upper right quadrant (Motor Themes)-high centrality and high density-were academic libraries, library services, and chatbots. Once again, these are fairly well-developed and strongly integrated into the broader research landscape, suggesting that they are defined areas of inquiry driving the field forward. Their positioning suggests strong conceptual coherence and strategic importance for academic institutions seeking to modernize library functions through AI-based tools.

The lower-right quadrant (Basic Themes) involves topics that serve as basic components but are not as internally developed, such as artificial intelligence, machine learning, and big data. While these topics are integral in the field, their lower density indicates they are still being developed for library science. This shows a gap between the full theoretical and practical incorporation of AI’s technological possibilities in academic libraries.

The upper left quadrant (Niche Themes) contains specialized topics such as ethics and artificial intelligence (AI) (in a narrow understanding) and academic integrity. These themes exhibit high levels of internal consistency and low levels of association with other theme areas, implying a specialized community of scholarly interest; the identified themes illuminate the ethical, regulatory, and normative aspects of introducing AI into educational and information systems, as its impact grows due to increasing autonomy and potential misuse across applications.

In the lower left quadrant (Emerging or Declining Themes), newer and less developed themes such as ChatGPT, generative AI, and AI literacy have low centrality and density in the network structure. They have low scores since new themes are still developing in their academic discourse. However, given the recent emergence of large language models and generative tools, these keywords are more likely to represent emerging themes that are simply occurring in the literature for the first time. Their current position reflects a good opportunity for early research and policy work.

The findings underscore a dynamic balance between technological adoption, service innovation, and ethical governance in using AI within academic libraries. The strength of the motor and basic themes suggests that research is actively bridging AI capabilities with library needs, while the presence of niche and emerging themes reveals an increasing focus on critical reflection, digital ethics, and user-centered implications. This map illustrates how scholarly attention is beginning to pivot toward how AI is used in libraries and its consequences for academic integrity, information behavior, and literacy education in the digital era.

• Relation between communities in Figure 17 and topics in Figure 18

  Figures 17 and 18 provide visualizations derived from the author’s keywords in academic library documents focused on artificial intelligence (AI) applications. While they share a common data source, the figures use different bibliometric techniques and serve distinct analytical purposes. Figure 17 presents a co-occurrence network map based on keyword frequency and co-appearance relationships, whereas Figure 18 depicts a thematic map derived from centrality and density metrics. To fully leverage the value of both visualizations, it is essential to clarify how the communities in Figure 17 correspond to the strategic themes in Figure 18 and whether they provide complementary or redundant insights.

  In Figure 17, the red cluster (Cluster 1) is dominated by highly interconnected keywords such as artificial intelligence, machine learning, and big data. These terms also appear in Figure 18’s lower-right quadrant (Basic Themes), confirming their centrality in the research field but indicating limited internal development, as reflected by their low density. This overlap illustrates how cooccurrence networks can reveal structural cohesion within a thematic cluster while the thematic map evaluates the strategic maturity of those same terms.

  Likewise, the blue cluster (Cluster 2) in Figure 17, containing keywords such as academic libraries, library services, and university libraries, corresponds to the Motor Themes quadrant in Figure 18, as concepts are shown to be current and highly developed. This supports the metaphorical conceptualization of these topics as structurally significant and mature.

  Moreover, the green cluster (Cluster 3) in Figure 17 includes chatGPT, generative AI, and AI literacy terms that appear in the lower-left quadrant of Figure 18 (Emerging or Declining Themes). This suggests that these topics are structurally connected but still in the early stages of thematic development.

  Clusters 4 and 5 (the purple and yellow clusters in Figure 17), containing keywords such as ethics, academic librarians, and technology, explore several sectors of the thematic map (niche themes and transitional areas) based on their position in terms of centrality and density. To summarize, Figure 17 and Figure 18 offer complementary depictions rather than duplicates.

  The co-occurrence network (Figure 17) emphasizes conceptual relationships and thematic clustering, while the thematic map (Figure 18) strategically evaluates topic importance and maturity. When interpreted together, these visualizations offer a more holistic understanding of how research on AI in academic libraries is structured, what themes drive the field, and which emerging areas may warrant further investigation or policy support.

• Complementary vs overlapping insights

  These two visualizations provide complementary insights rather than redundant or overlapping results. The co-occurrence network (Figure 17) emphasizes the structural and relational patterns among keywords-how concepts group and interact in the literature; the thematic map (Figure 18) classifies these topics based on their strategic importance (centrality) and development stage (density). For example, two keywords may fall within the same cluster in Figure 17 since they often co-occur, while one is more conceptually developed. In Figure 18, they would fall within different quadrants. Thus, we see different but related aspects of topic analysis in each visualization.

The thematic evolution graph illustrates the progression of research themes related to AI applications in libraries across 2014-2018, 2019-2021, and 2022-2024 (Figure 19). The main themes from 2014 to 2018 were “academic libraries” and “machine learning,” which highlighted the emerging phases of AI research in library environments and highlighted foundational technologies and their possible uses.

Figure 19.

Academic libraries’ thematic development of AI technology (2014-2024).

AI, academic libraries, and digital libraries were added to the list of research areas for 2019-2021, illustrating how AI technologies have advanced beyond machine learning to encompass more complete AI concepts. Digital libraries are emerging to use AI to manage digital collections and services as library systems undergo digital transitions. The themes were even more diverse by 2022-2024, with big data becoming a new theme while artificial intelligence and digital libraries continued to be popular. This indicates a rising integration of large-scale data analytics in academic libraries and digital library operations, as well as a greater emphasis on employing AI to manage and extract insights from vast databases.

AI’s ongoing prominence highlights its significance for future studies. The growing emphasis on digital libraries reflects the digital revolution of library infrastructures, where access, management, and personalization are all improved by artificial intelligence. The emergence of big data in recent years highlights the significance of data-driven decision-making for library operations through cutting-edge AI and analytics tools.