Artificial intelligence integration in academic libraries in South Africa: a Web analysis Open Access

Information and communication technologies (ICTs) have long been considered the main drivers of evolution in libraries, enhancing information retrieval, streamlining knowledge organisation and amplifying user education and other services. These ICTs help to bridge the gap between the traditional and contemporary information landscapes ensuring that information is mainly available at the click of a button. The mid-2010s witnessed rapid technological advancements such as machine learning, Internet of Things, cloud computing, Augmented and Virtual Reality and artificial intelligence (AI). This period has been characterised as the fourth Industrial Revolution (4IR) by many. In coining the term, Klaus Schwab described the 4IR as a period that is driven largely by a fusion of technologies blurring the lines between the physical, digital and biological spheres (Schwab, 2016, 2025). It is a period where data and digital technologies play a critical role in human development. Potentially, these 4IR technologies can revolutionarise the way libraries function, affecting all processes and leading to the automation of many of their traditional services. They can enhance information provision through provision of user-tailored services.

Perhaps more than any other types of libraries, academic libraries have been at the forefront of adopting the 4IR technologies to improve the effectiveness and efficiency of their operations and functions (Ocholla and Ocholla, 2020). However, it is noted that the use of 4IR technologies in developing countries like South Africa remains uneven with some technologies more prominently used than others. Makerspaces, research data services and – most notable – AI were among the technologies that show promise for huge impact in the future but whose use remains limited.

The possibility for the integration of AI in libraries presents with it not only endless opportunities but also some complex challenges. Ocholla and Ocholla (2020) believe that AI should feature prominently in academic libraries’ digital transformation agenda. AI can transform the way libraries conduct most of their operations, simplify their processes and enhance efficiency (Kalbande et al., 2024). It has potential to significantly alter the way we discover, learn and work (Wiley, 2025). Zondi et al. (2024) point out that AI brings with it many advantages to academic libraries. These include ubiquity, 24/7 service, efficient service delivery, reduced workload for staff and many others. However, there are concerns about the employment of AI in academic libraries, especially in resource constrained countries of the Global South. These include ethical concerns, privacy issues, limited staff capacity and poor infrastructure to support some of the AI services and tools (Kalbande et al., 2024). If not addressed, then these may further deepen the prevailing digital divides between those who can afford the most advanced AI technologies and infrastructure and those who cannot. It was for this reason that the International Federation of Library Associations and Institutions (2020) cautioned that AI in libraries should be used with clear ethical standards to balance privacy issues and the need for access to information. Further, librarians should be capacitated to understand AI applications and their ethical risks and implications (Filson and Atuase, 2024). Ocholla and Ocholla (2020) exhort South African libraries to find ways to interrogate, popularise and advance AI development and applications. Without embracing these technologies, academic libraries may find themselves in limbo. Additionally, academic libraries must improve the visibility of their digital transformation initiatives on their websites to ensure transparency and demonstrate their ability to keep up with trends.

However, before this study, it was not clear to what extent are these libraries integrating AI in their functions and operations and whether such integration is visible on their websites. Websites remain the doorways through which library resources are accessible (Onyancha and Ocholla, 2007). This study uses Web content analysis to explore the integration of AI in South African academic libraries. The employment of Web content analysis not only provides a glimpse of AI technological applications in South African university libraries but also highlights the visibility of these initiatives on the Web. It is hoped that this study will trigger further dialogue and inquiry into this critical area in other parts of the world, especially in developing countries. In the context of South Africa, Ocholla and Ocholla (2020) broadly explored the adoption of 4IR technologies in academic libraries but not specifically AI. This study is guided by a single objective, which is to glean evidence of AI integration in academic library websites in South Africa. The results of this study offer insights on how libraries in South Africa are embracing AI technologies to enhance their services as well as the visibility of such initiatives on their websites.

In this literature review, an attempt is made to contextualise AI with the study’s purpose and objective in place. Thereafter, a high-level review of literature on the integration of AI in academic libraries follows. The last paragraph seeks to communicate the gap in literature that this study seeks to address.

Although the term AI made its way to the public discourse around the 2010s to the 2020s as a result of significant breakthroughs in the field such as advanced robotics and AI chatbots, ChatGPT and other applications, the first known use of the term dates back to 1955 in a proposal for the study of AI by John McCarthy et al. (2006). At the time of its conception, the term referred to computers that could do some routine tasks. Despite this long history, there is disagreement on what constitutes AI leading to varied formal definitions of the term (The United Nations Educational, Scientific and Cultural Organisation [UNESCO] and United Nations Educational, Scientific and Cultural Organisation and World Commission on the Ethics of Scientific Knowledge and Technology, 2019 (COMEST) (2019)]. However, there are noticeable common patterns in available AI definitions in literature (Cox and Mazumdar, 2024). Miao and Shiohira (2022:9) defines AI as “machines that imitate some features of human intelligence, such as perception, learning, reasoning, problem-solving, language interaction and creative work”. The Organisation for Economic Co-operation and Development (2024) (OECD) refers to AI as machine-based systems that, for explicit or implicit objectives, infers, from the input they receive, learn to generate outputs such as predictions, content, recommendations or decisions that can influence physical or virtual environments. In the context of libraries, the American Library Association (2019) defines AI as the use of deep and machine learning and natural language processing tools to create “intelligent” machines that work and react like humans. AI relies on mimicking human capabilities (Anjila et al., 2021). Rabatseta et al. (2024) concur pointing out that AI involves collaboration between humans and robotic machines to improve job performance, generate increased competence and drive economic progress. AI focusses on human behaviour and how machines imitate human behaviour (Asemi, Ko and Nowkarizi, 2021). In contextualising AI to academic libraries, Tzanova (2024) calls it a synergy between computing and psychology (that is) used to increase librarians’ productivity, quality of services, quality of information yields and quality of communication with patrons. There are many convergencies between the definitions proffered in literature. The definitions are also similar in their recognition of how AI seeks to ultimately behave and act in the same way as human beings. From the OECD definition, it is clear that AI do not currently function independently requiring human intervention and input. AI, therefore, is advanced machine-based tool with limited human capabilities that is used or have potential to be used to improve the productivity of libraries.

The use of AI in libraries remains in infancy (Wheatley and Hervieux, 2019). As a result, the full AI landscape in academic libraries remains uncertain (Gasparini and Kautonen, 2022). However, in the past few years, evidence has started to emerge that academic libraries are leveraging AI to enhance their efficiency. Like in any complex field, studies in this area use different research methods. These include use of interviews (Cox et al., 2019; Filson and Atuase, 2024), surveys (Kalbande et al., 2024), content analysis (Huang et al., 2023; Molaudzi and Ngulube, 2025; Osorio, 2024), systematic reviews (Hodonu-Wusu, 2024; Mabona et al., 2024; Tella, 2023) and literature reviews (Modiba and Ngulube, 2024; Rabatseta et al., 2024; Zondi et al., 2024). Indeed, AI is a complex field requiring different approaches to be fully understood. Literature in this area explores the (potential) uses of AI in academic libraries, the implications of AI to academic libraries, types of AI applications used in academic libraries, academic library strategic responses to the proliferation of AI, barriers to AI use in the libraries and AI literacy and the ways by which libraries can capacitate users to ethically use AI.

Published works show that use of AI can potentially improve efficiencies in reference, cataloguing and acquisitions services of libraries (Mabona et al., 2024). Chatbots and robots, for example, can be trained to handle basic user queries that often form the bulk of queries handled by reference librarians, as well as simplify information retrieval for users through automated book sorting and shelving. In addition, AI can be relied upon to assist in the classification and indexing of information resources. To this end, Raghvaiah and Sreekanth (2023) provide some practical examples from Singapore. PERSIST, the robot of the central library in Singapore, can shelve and retrieve books for clients, while another robot, Bobbie, developed by the Temasek Polytechnic Library, is programmed to answer Frequently Asked Questions. With such humanoid assistance, librarians’ time is freed to focus on other more complex tasks that require deeper engagement, critical thinking, contextual understanding nuances, empathy and logic, things that machines are currently not excelling at. In South Africa, there have been two attempts to use robots that respond to basic user queries. The robot, Libby, was used at the University of Pretoria Library to respond to basic user queries, while robot Sifiso is used at the National Library of South Africa (Modiba and Ngulube, 2024).

Filson and Atuase (2024) believe that AI can have unintended consequences simplifying the path towards academic dishonesty for students and researchers alike. It is critical, therefore, that librarians are sufficiently empowered in AI so that they can deliver AI literacy empowerment workshops at the point of need. Chigwada (2024) advances this view, pointing out that the potential for AI use in academic libraries is vast, permeating all functions such as research support, teaching and learning. However, academic libraries have a responsibility to develop collaborative AI literacy frameworks that can be used to teach users AI literacy that would serve to stem instances of academic dishonesty.

There is a growing realisation that AI in academic libraries can no longer continue to be dealt with in a haphazard manner. It needs to be seamlessly infused in the strategic plans of these libraries, giving it more structure and adding its development as a budget item. Studies are now using the strategic plans of university libraries to determine AI presence in libraries (Huang et al., 2023; Wheatley and Hervieux, 2019). Both studies found no mention of AI in the strategic plans of university libraries in the USA, the UK and China pointing to a gap. However, it was clear that some libraries have started to introduce limited AI applications and services (Huang et al., 2023).

The road towards using AI in academic libraries, however, is not without pitfalls. Some of these challenges are unique to developing countries, while others affect the developed world as well. Limited capacity, inadequate funding and insufficient AI infrastructure are often cited as the biggest barriers to the adoption of AI in libraries. Fear of job loss and unstable power supply are also cited in the developing countries (Omame and Alex-Nmecha, 2020; Oyetola et al., 2023; Raghvaiah and Sreekanth, 2023).

Academic library literature in AI is mostly exploratory with very little in the way of practical implementations. This study goes beyond theoretical discussions and provides compelling evidence of how AI is currently integrated in academic libraries in South Africa.

As there are no relevant standard checklists for evaluating AI integration in academic libraries, the first task was to develop such a checklist using Ms-Excel. This was after an extensive literature review was done to determine the issues often associated with AI in academic libraries. It was important that the checklist be complete to cover all relevant categories of AI services. Eventually, the checklist consisted of 11 categories under which data were collected. Those are: AI guidelines and policies, AI advocacy and announcements, AI literacy training, AI scholarly engagement, dedicated AI LibGuides, LibGuides with AI pages, organising AI-related conferences and other engagement platforms, AI ethics and academic integrity, reference to AI tools, AI-powered resources and services and AI Resources (books, databases, journals, etc.). The description of the various categories can be found in Table 1 which was used as a guiding protocol for the collection of data.

Thereafter, the 26 public universities’ library websites were evaluated using this self-created checklist. The list of public university libraries was obtained from the Committee of Higher Education Libraries of South Africa (2021) “State of South African Academic Libraries Report”. These are: Cape Peninsula University of Technology (CPUT) Libraries, Central University of Technology (CUT) Library and Information Services, Durban University of Technology (DUT) Library, Mangosuthu University of Technology (MUT) Library Services, Nelson Mandela University (NMU) Library and Information Services, North-West University (NWU) Library and Information Service, Rhodes University (RU) Library, Sefako Makgatho Health Sciences University (SMU) Library, Sol Plaatje University (SPU) Library Services, Stellenbosch University (SU) Library and Information Services, Tshwane University of Technology (TUT) Library and Information Services, University of Cape Town (UCT) Libraries, University of Fort Hare (UFH) Library, University of the Free State Library and Information Services, University of Johannesburg (UJ) Library, University of Kwazulu-Natal (UKZN) Library, University of Limpopo (UL) Library Services, University of Mpumalanga (UMP) Library, University of South Africa (UNISA) Library, University of Venda (UNIVEN) Library, University of Zululand (UNIZULU) Library, University of Pretoria (UP) Library, University of the Western Cape (UWC) Library, Vaal University of Technology (VUT) Library, University of the Witwatersrand (Wits) Library and Walter Sisulu University (WSU) Library.

Websites of all these academic libraries’ websites were visited firstly between December 2024 and January 2025, then March 2025. Browsing and searching was then done to identify any signs of AI integration by the academic libraries. Use of library websites to study the integration of technologies is not unique to this study. Huang et al. (2023), Ocholla and Ocholla (2020) and Wheatley and Hervieux (2019) also used website content analysis in similar studies. The researcher could learn from previous studies by replicating some of their keywords to search for AI in academic libraries. The search terms used were AI, robotics, ChatGPT, AI literacy, machine learning, deep learning and AI hub. Additionally, LibGuides of these university libraries were visited and similar searches conducted. To simplify the quantitative analysis of data, where evidence of the existence of a service/resource was found in a website, the number “1” was entered in the checklist, while “0” depicted a lack of evidence of the existence of that particular service/resource from the library’s website. Data were then analysed using the Excel functionalities.

The findings are divided according to the various categories of the checklist used to collect data (Table 2). For the purpose of presenting these findings, the Reference to AI Tools and AI Resources (books, databases, journals, etc.) themes were presented under AI LibGuides. This is because most of this information is captured in LibGuides of the university libraries.

AI policy can assist to provide structure and guide the actions of libraries in adopting AI integration. They can assist libraries to structurally decide what AI tools to adopt and to align those with their library’s mission. Although no academic libraries in South Africa seem to have their own AI policies, there is clearly a realisation that such policies are crucial. The libraries often refer to their universities’ guidelines. Even at university level, none of them seem to have a fully-fledged policy pointing to the infancy of the field. The UWC, NWU, NMU, UKZN, SU and UCT libraries all refer to either the position statements of their universities or guidelines. Not only are local guidelines linked but also national policies, such as the South Africa National Artificial Intelligence Policy Framework, as referred to by NWU Library and Information Service. Additionally, NMU library refer users to the ASSAf and SciELO Guidelines for the Use of AI Tools and Resources in Research Communication. This library also refers users to the international policy framework in the form of the European Commission: Ethical guidelines on the use of AI and data in teaching and learning for Educators. This provides users with a broader perspective of the policy environment on AI. Overall, ten university libraries refer to an AI policy/guideline or the other. Besides the already mentioned, CUT, CPUT, UJ and UP also reference AI policies/guidelines in their websites.

Classified under this were workshops, AI blog posts and other AI announcements. There are eight libraries where evidence of active advocacy was found. These are: NWU, RU, SPU, SU, UKZN, UNISA, UWC and Wits. In September 2023, for example, RU library hosted a workshop on “Artificial Intelligence and Topic Discovery Webinar”. Not to be outdone, the UNISA Library reportedly organised a three-day workshop “Exploring AI and robotics in public libraries” as part of its outreach initiative to that sector. SU and NWU libraries seem to be highly active in the blog environment posting articles on AI research support tools, AI-related conferences attended by staff and AI webinars organised by such institutions.

Ten university libraries demonstrated their understanding and contribution to AI literacy. Unit 6 of the SPU’s information literacy course is on AI Literacy, covering topics such as AI tools, responsible use of AI in academia, everyday use of AI by students, ethical considerations of AI and finally, referring them to the SPU AI guidelines which are presumably unpacked and explained during the training. In other institutions, there is no compelling evidence to demonstrate the SPU’s infusion levels of AI in information literacy training. However, these institutions cover one or two AI-related topics with AI ethics being the most common one. SU, for example, includes “Introduction to the responsible use of generative AI at university”, a session that is part of its #SmartStudents Workshop. Similarly, UP lists “Academic integrity workshop: generative AI in education” among its workshops.

Of these libraries, 12 provide information on this theme. These include libraries of: CPUT, CUT, NMU, SPU, UCT, UFS, UJ, UKZN, UNISA, UP and UWC. As an example, CPUT Libraries’ homepage provides a “student AI Practical Guide” which seeks to explain AI to users, identify some ethical considerations while using AI, link to CPUT AI guidelines, guide users on responsible use of AI and share information about useful AI tools and resources. Regarding ethical considerations, the guide links to a model that identifies different possibilities for harm as a result of unethical use of Large Language Models. For its part, CUT Library provides information about ethical considerations of AI on its “Artificial Intelligence Literacy - Use of AI Generative Tools” LibGuide, not only referring users to the CUT AI guide but also unpacking instances where AI use can be considered unethical. Similarly, NMU Library and Information Services’ “Artificial Intelligence (AI)” LibGuide lists ethical considerations of AI as those associated with the tendency to hallucinate, cause plagiarism, use copyrighted content, inflict harm through discrimination, share harmful content, share private information and lack of fairness and transparency.

Under this theme information about engagement of staff in AI intellectual debates through publishing in academic platforms and presenting papers at conferences was captured. There were 11 university libraries where relevant information depicting their intellectual and scholarly engagements in AI integration was found. Those are: DUT, MUT, NMU, NWU, SU, UKZN, UL, UNISA, UP, UWC and WSU libraries.

Most scholarly engagement has been through the three IFLA IT Section’s AI conferences hosted in South African university libraries. Following the third IFLA IT Section AI Symposium hosted by NWU in 2023, a special issue dedicated to this conference was published in volume 90 issue 2 of the South African Journal of Libraries and Information Science. Library staff have contributed to other scholarly platforms. For example, a paper co-authored by Anele Mabona, NMU Library’s Manager: Digital Innovation Systems and Web Management and two academics at that university, was presented and published as part of the conference proceedings of the IST-Africa 2024 Conference, an IEEE conference. Staff also contribute to internal platforms such as the lightning talk on “Artificial Intelligence: navigating the next decade @ DUT Library” by DUT Library’s Brian Reynolds (2023) during the DUT’s Second Research Colloquium in 2023.

Classified under this theme were conferences, seminars, webinars and other engagement platforms on AI. There were nine instances where evidence of organisation of such platforms was found. As indicated above, three IFLA international conferences on AI were hosted by UP Library in 2021, NWU Library and Information Service in 2023 and UKZN Library in 2025.

In 2023, SU Library and Information Service organised a panel discussion under the title: “A saviour or monster in our midst? AI in higher education”. Discussions focussed on the role of AI in higher education with specific focus on libraries. During the same year, Rhodes University Library organised a webinar on “Artificial Intelligence and Topic Discovery” by Dr Kirstin Krauss from WWIS. Another webinar was organised by NWU Library and Information Service on “AI Tools for Research and Content Discovery in the Fifth Industrial Revolution” by Dr Mashilo Modiba. UJ Library launched its digital scholarship under the theme “Unlocking the Power of Artificial Intelligence to Enhance Research”. The launch description promised to present that library’s cutting-edge technologies and AI tools. All these initiatives are meant to empower librarians and researchers on AI aspects.

The UCT’s National Immunization Technical Advisory Group (2025) describe LibGuides as “online repositories of literature and other resources that created by librarians to collect, organise and guide through information on a particular topic”. These mini-websites can communicate the services and resources supported by libraries. The interest of this theme was AI dedicated LibGuides at these university libraries. The Reference to AI Tools and AI Resources (books, databases, journals, etc.) themes were combined under this section because of the alignment noticed during data collection. Most information about AI tools and resources is captured in the LibGuides of these university libraries.

There are six university libraries with AI dedicated LibGuides in South Africa. Those are: CUT, NMU, UCT, UFS, UJ and UWC. In addition, CPUT, SPU, SU, UFH, UKZN, UNISA and UP have pages and links to AI in other LibGuides. LibGuides contain aspects such as guidelines and policies, responsible use of AI, benefits of AI, AI tools and resources and ethical applications of AI. Among the resources commonly mentioned are: writing tools, chatbots, AI image assistants, AI research tools, video AI and image AI tools. In addition, recommended library resources such as books, journals and databases are often provided. Generally, 14 university library websites refer to AI tools, while 20 refer to an AI resource or the other.

The researcher was also interested on AI resources and services provided by the university libraries. There were 11 libraries that mentioned AI-powered resources and services in their websites. Those were: CPUT, DUT, NMU, NWU, SU, UCT, UFS, UJ, UMP, UP and UWC libraries.

Despite the excitement about robotics in academic libraries, robotics do not feature prominently in South African academic libraries. Only two university libraries mention robotics. Those are the UP and NWU libraries. The UP Library website mentions Libby, who is described as the first library client service robot on the African continent while NWU Library and Information Service refers to a project where the library is designing a client service robot called Kletsbek-Kitso Library Robot in collaboration with the Engineering Faculty. According to the information provided, this robot will perform mundane tasks. There was no evidence of similar work being done elsewhere.

Mention is made of other AI-powered resources and services in these libraries. These include subscriptions to AI Scopus at Wits, RFID and Primo Research Assistant. Wits Library describes Scopus AI “as an intuitive and intelligent search tool based on generative artificial intelligence (GenAI) that enables users to navigate the vast amount of information available in Scopus”. Currently, Wits Library seems to be the only university library that has activated Scopus AI subscription. In announcing its subscription to Primo Research Assistant, SU described it “as a generative AI tool that enables users to research a topic in natural language, using library and expanded academic electronic resources”. CPUT also subscribe to this AI-powered resource.

Academic libraries also mention use of open-source AI-powered tools such as Cockatoo and Grammarly. UCT Libraries, for example, reported how they used Cockatoo to transcribe some of their audiovisual collections to efficiently assign metadata descriptions (Mvungi and Walker, 2024). Despite the poor quality of the video, Cockatoo was able to transcribe the videos with 99% accuracy in English. Because of the success of the project, the library plan to use cockatoo for transcribing other audiovisuals in future.

Key themes were drawn from the results of this study that demonstrate academic libraries’ engagement of AI to improve efficiency. Despite the excitement about AI, it was clear from the results of this study that AI integration in academic libraries is in its nascent stages with very little by way of AI policy development. Mabona et al. (2024) advise that libraries should play a proactive role in AI policy formulation and contribute to ethical AI use at universities.

Galvanised by international developments, these libraries have only recently taken up AI advocacy roles using social media blogs and other platforms. There is a growing recognition by the libraries of the need for their user communities to be empowered on aspects of AI. Therefore, some of the libraries have started to roll out AI literacy training infused within the IL training. Chigwada (2024) advised academic libraries to develop frameworks for AI literacy training. Clearly, AI ethics and academic integrity are worrisome for the integration of AI in academic libraries (Filson and Atuase, 2024). In their websites, these South Africa academic libraries demonstrate efforts that can help decrease instances of academic dishonesty through awareness creation. It is clear that ethics and academic dishonesty that take advantage of the opportunities provided by AI needs a clear strategy from academic libraries including through collaborative AI Literacy frameworks as advised by Chigwada (2024).

Despite perceived AI infrastructural challenges, it is commendable that these academic libraries are contributing to intellectual discourse within the field through hosting conferences, webinars, workshops and publishing and presenting in journals, conference proceedings and books. These are critical to stimulating debates and sharing best practices in the field.

These results demonstrated a lack of deeper engagement with advanced AI such as robots and AI chatbots. This is not surprising, as South Africa is often slow to adopt new technologies because of a number of challenges including limited capacity, inadequate funding and insufficient AI infrastructure.

These results, therefore, point to the need for better strategic investments in AI infrastructure. Furthermore, there needs to be proper strategies for capacity building and imparting of AI skills to librarians and users alike. These, accompanied by the development of AI policy frameworks will assist academic libraries to maximise the benefits of AI and minimise the ethical concerns.

This study sought to explore the integration of AI in academic libraries in South Africa. Websites of university libraries were analysed for evidence of AI integration. The results show that South African academic libraries have been slow in adopting advanced AI applications pointing to an AI environment that is still in its early stages. However, the results also highlight instances where South African academic libraries have shown strong AI inclinations including through contributing to intellectual discourse and advocacy in this area.

There are glaring weaknesses such as a weak AI policy environment, poor AI infrastructure for supporting advanced AI applications and shortage of technical skills. Therefore, it is critical that universities and libraries develop policies and invest strategically in AI infrastructure. This will help academic libraries to fully leverage the advantages of AI adoption and improve the efficiency of research, teaching and learning support. Additionally, it is critical that librarians and users are capacitated to take advantage of the AI applications.

These results point to a need for collaborative effort to offer AI literacy to students. This will heighten ethical use of AI for research, teaching and learning. In addition, there needs to be ongoing debates and discussions on the best practices to adopt AI in academic libraries. Given that the field is new, there is a lot that practitioners and scholars can learn from each other and among each other through case studies. If properly embraced, then AI can improve efficiencies in academic libraries’ processes and free staff time so that they can focus on more complex functions that require understanding of different contextual nuances, critical thinking and innovative solutions. Therefore, it is critical that these libraries place AI at the centre of their digital transformation strategy.