AI literacy has become an essential competency for navigating an increasingly digital world, encompassing more than just the ability to understand and use AI technologies. The concept of AI literacy is relatively new and is often viewed as an advanced form of digital literacy (Yang, 2022). In its basic definition, AI literacy is the ability to understand, interact with, and critically evaluate AI systems and AI outputs. Hossain et al. (2025) conceptualise AI literacy as a multidimensional construct comprising students’ familiarity with AI, their knowledge of its applications, and their ethical perceptions of its use. These three dimensions collectively provide a comprehensive measure of students’ AI literacy levels (p. 3).

While AI literacy is often situated within the broader umbrella of digital literacy, it is conceptually distinct. Digital literacy generally refers to skills in navigating, evaluating and producing information in digital environments (Law et al., 2018; Lee, 2014). In contrast, AI literacy emphasises competencies related specifically to artificial intelligence technologies, including an understanding of how AI systems function, how to apply AI tools, and how to critically evaluate their outcomes (Long & Magerko, 2020). UNESCO (2022) defines AI literacy as equipping individuals with “the skills, knowledge, and ethical understanding necessary to navigate the complexities of AI technology”. Correspondingly, the World Economic Forum (2024) emphasises the essentials of AI literacy in preparing learners to “responsibly use emerging technology for the good of society.”

From a legal perspective, the EU AI Act defines AI literacy as the “skills, knowledge, and understanding that allow providers, deployers and affected persons … to make an informed deployment of AI systems [and] gain awareness about the opportunities and risks of AI” (European Union “Article 3: Definitions”, 2024). Beyond technical proficiency, AI literacy fosters critical thinking and a deeper understanding of AI’s broader social, cultural and political implications, as its increasing pervasiveness continues to shape societal interactions, governance and cultural landscapes (McKee, 2024). Similarly, Ng et al. (2024) advanced the ABCE model, which organises AI literacy into affective (motivation, self-efficacy), behavioral (engagement, collaboration), cognitive (knowledge, evaluation, creation), and ethical dimensions. These perspectives demonstrate that AI literacy goes beyond technical proficiency to include critical awareness and ethical reflection, which are especially salient for LIS education.

Building on this, Hossain (2024) introduces the concept of AI citizenship as a complement to AI literacy. While AI literacy focuses on understanding and engaging with AI technologies, AI citizenship emphasises informed and ethical participation in an AI-driven society. According to Hossain, AI citizens are expected not only to use AI responsibly but also to understand their rights, advocate for transparency and contribute to public debates on fairness and accountability. This framing highlights the civic responsibilities of future librarians, positioning them as both competent users and ethical stewards of AI in their professional environments.

AI adoption in Africa presents both opportunities and challenges, particularly in education, where it has the potential to enhance learning outcomes while also requiring careful governance and policy frameworks. The UNESCO-Eastern Africa Sub-Regional Forum on Artificial Intelligence (EARFAI) Report 2024 highlights the need for AI literacy, inclusive education and ethical AI integration across the continent. Key recommendations include: i) implementing AI literacy programmes to equip students, teachers and the public with foundational AI knowledge and critical thinking skills; ii) promoting AI competency frameworks for students and teachers to ensure ethical and responsible AI usage; iii) providing AI training programmes specifically targeting youth, women, people with disabilities, and marginalised groups to reduce digital inequalities. The report further calls for AI policies in education, enhanced digital equity and the establishment of AI research hubs to foster responsible AI adoption in African educational institutions.

The African Union’s (2024) AI Strategy prioritises embedding AI literacy into education systems, training teachers in ethical AI use and ensuring inclusive access for marginalised groups. Furthermore, the document promotes the development of multilingual AI tools, research centers and ethical frameworks aligned with UNESCO guidelines (UNESCO EARFAI, 2024). Governments must fund infrastructure and partnerships to equip youth for an AI-driven future under Agenda 2063 and the Sustainable Development Goals (SDGs) (African Union, 2013). A case study by Nyakundi et al. (2024) underscores AI's dual role in Africa: as a tool for educational transformation and a challenge requiring strong governance, ethical safeguards and inclusive policies. The study also advocates for integrating AI education into curricula and providing specialised AI training for educators to facilitate adoption. Additionally, it stresses the importance of transparent and fair AI regulations to address issues of bias, accountability and student data privacy.

Despite its potential to drive economic growth and educational advancements, AI adoption in Africa remains limited. Many countries lack the essential infrastructure and policy frameworks needed for their effective deployment (Nyakundi et al., 2024). In the higher education sector, AI implementation reflects a complex balance between potential benefits and persistent structural challenges, requiring strategic investment and policy interventions to maximise its impact.

As Africa advances in AI-driven education, the continent needs to adopt a context-specific approach that aligns with its educational values and priorities (Nyakundi et al., 2024). While global practices offer insights, AI adoption should be tailored to Africa’s unique needs, ensuring equitable access, ethical integration and meaningful innovation. Libraries and librarians play a critical role in promoting AI literacy by curating resources, facilitating digital training and supporting AI research (Cox, 2023a; Hossain et al., 2025). As knowledge hubs, they can bridge the AI literacy gap, equipping students and educators with essential skills for responsible AI engagement. Through advocacy and training, librarians can help ensure AI supports education, knowledge equity and lifelong learning in Africa.

In an evolving information landscape marked by the ascendancy of AI technologies, libraries worldwide are strategically pivoting to cultivate AI literacy as a core competency (Chun Ru & Tang, 2025). The integration of AI into libraries and librarianship has evolved significantly since its nascent applications in the 1970s (Weinryb-Grohsgal, 2020). While early AI adoption focused on automating routine tasks, contemporary GenAI tools—such as SciSpace, Elicit and ResearchRabbit—now facilitate advanced literature searches, summarisation and analysis, demonstrating the transformative potential for library workflows (Cox, 2023a, 2023b). However, the rapid proliferation of these technologies has exposed systemic competency gaps among LIS professionals, particularly in resource-constrained African contexts (Nyakundi et al., 2024). These gaps underscore the urgent need for structured AI literacy programmes that address both technical proficiency and ethical dilemmas, such as algorithmic bias and data privacy (Cox, 2023b; Ibrahim, 2025).

In Sub-Saharan Africa, studies highlight AI’s dual role as a catalyst for efficiency and a source of institutional tension. For instance, Echedom and Okuonghae (2021) emphasise AI’s capacity to enhance library services through natural language processing and machine learning, positioning it as indispensable for navigating the Fourth Industrial Revolution. Similarly, Zondi et al. (2024) and Molaudzi and Marutha (2025) document AI-driven improvements in operational efficiency and service personalisation in South African academic libraries, though infrastructural constraints and staff readiness remain critical barriers. In Zambia, Alam et al. (2024) report librarians’ strong foundational understanding of AI but note persistent challenges, including resistance to technological change, budgetary limitations and the need for advanced expertise.

The ethical implications of AI adoption loom large in African library contexts and Ibrahim (2025) calls it an 'ethical maze'. Studies from Nigeria and Zimbabwe reveal divergent attitudes: while AI tools like ChatGPT are increasingly used for content generation, concerns about misinformation, plagiarism and job displacement persist (Ajani et al., 2022; Chigwada & Pasipamire, 2024; Fasola, 2024). For example, Nigerian librarians acknowledge AI’s potential to minimise human error in repetitive tasks but express scepticism about its threat to traditional library roles (Ajani et al., 2022). These findings align with broader calls for ethical frameworks to govern AI deployment, balancing innovation with accountability (Ibrahim, 2025; Zondi et al., 2024).

African librarians are increasingly positioned as advocates for responsible AI integration. Akakpo (2024) argues that librarians must lead AI literacy initiatives by developing guidelines for ethical GenAI use and integrating digital literacy into curricula. This aligns with Okojie and Awoyemi’s (2024) emphasis on policy advocacy and the need for AI-literate professionals to drive sustainable adoption. Ibrahim (2025) further contends that libraries bear an ethical responsibility to ensure AI serves as a societal good, necessitating competencies in critical thinking and ethical reasoning among LIS practitioners.

Despite growing scholarship, few studies holistically examine AI literacy among LIS students—future library professionals—and to the best of our knowledge, no such study has been conducted in the African context. Existing research often isolates factors like tool usage or infrastructure, neglecting the interplay of familiarity, proficiency and ethical awareness. This study addresses this gap by analysing the current state of AI literacy among LIS students across five African nations, offering actionable insights for curriculum development, policy formulation and ethical LIS education.

Purposive sampling was employed to recruit a diverse cohort of respondents from LIS programmes across academic institutions in Ghana, Nigeria, Rwanda, South Africa and Uganda. While purposive sampling does not ensure random selection, it is suitable for targeting specific populations (Etikan et al., 2016), in this case, LIS students, to explore their AI literacy.

Recruitment challenges were experienced in South Africa owing to logistical issues, including limited institutional access and lower response rates. The questionnaire was administered between March and April 2024 via student email lists and university class groups on social media platforms such as Facebook and WhatsApp Groups. The survey was also distributed to various LIS institutions through personal and professional networks, and the authors directly contacted LIS faculty members at multiple universities via their official websites. The sample included students from both public and private universities, representing different levels of LIS education, ensuring a well-rounded and diverse dataset.

The survey yielded 437 responses, 220 from Ghana, 131 from Nigeria, 23 from Rwanda, 53 from South Africa and 10 from Uganda. A detailed demographic breakdown is provided in Table 3. The gathered responses were analysed using statistical analysis tools, such as SPSS and Google Sheets. Descriptive statistics were carried out to show the variability. To analyse AI literacy differences based on country, study level, gender, and English proficiency, various statistical methods were used. The Games-Howell Post Hoc Test was chosen for cross-country comparisons due to unequal sample sizes and non-normal distributions (Alzubi, 2024; Choukou et al., 2024). Multinomial Logistic Regression (MLR) assessed how these variables influenced AI literacy levels (Basic, Intermediate, Advanced, Expert), by estimating the probability of individuals belonging to each category.

This study adhered to ethical guidelines ensuring anonymity, confidentiality and voluntary participation. Informed consent was obtained, and participants were clearly informed of the study’s purpose, data usage and privacy. The research complies with relevant institutional ethical standards and ethical principles governing studies involving human subjects.

This study has several methodological limitations. It relied on self-reported data and a purposive, voluntary sample, which might introduce bias and limit representativeness, as more digitally engaged professionals may have been overrepresented. The questionnaire was distributed across all universities in the participating countries offering LIS or similar degrees, rather than to a targeted set of institutions, which may also have affected the balance of responses. These factors should be considered when interpreting the findings.

The sociodemographic data are the critical part of the study and are presented in Table 3. The responses were collected based on five variables, namely country, university type, gender identity, current study level and English Language Competency (ELC). In terms of participation by country, Ghana topped the list, followed by Nigeria, South Africa, Rwanda and Uganda. Most of the respondents (65%) are enrolled in public universities (state or provincial government-funded) with a near-equal gender distribution and different study levels, with the majority being undergraduate students (72.3%).

Participants’ English competency was important in this research, as proficiency is indispensable for navigating the predominantly English-language AI tool ecosystem in order to facilitate self-directed learning and crafting precise, effective prompts. Most respondents self-rated their ELC ranges between ‘Intermediate (B1) and Upper Intermediate (B2) levels.

To determine the AI literacy levels of LIS students by country, the study utilised a self-reported assessment, allowing participants to situate themselves on a continuum from 'No Experience' to 'Expert' in AI. Not only are Uganda’s participant numbers the lowest of the participating countries, but they also have less AI literacy at this time, as indicated by the ranking in AI capacity in the ‘Government AI Readiness Index 2024 from Oxford Insights’, as shown in Table 1.

The findings of this study reveal a nuanced picture of AI literacy. While Ghana tops the list in terms of the percentage of LIS students who self-describe their AI literacy at the 'Advanced' or 'Expert' levels (as shown in Figure 1), the overall mean AI literacy scores, as presented in Table 5, show a different ranking. Uganda (x̄, 3.5) and Rwanda (x̄, 3.2) demonstrate higher mean scores, followed by South Africa (x̄, 2.7), Ghana (x̄, 2.5) and Nigeria (x̄, 2.3). This suggests a possible disconnect betweeen self-perception and actual competency, which is a subject for further research.

Table 4 shows the results of one-way ANOVA revealing significant AI literacy differences among the five participating countries (F = 5.196, p < 0.001, η2 = 0.046) and respondents’ gender (F = 10.497, p = 0.001, η2 = 0.024) and ELC (F = 9.583, p < 0.001, η2 = 0.082), whereas the results also reveal no significant differences based on respondents’ university type (F = 1.835, p = 0.161, η2 = 0.008 ) and/or level of study (F = 1.234, p = 0.297, η2 = 0.082). The effect size (η2) indicates that ELC has the strongest influence among the significant variables. Based on the results, it is observed that demographic variables such as country, gender and ELC play crucial roles in AI literacy differences, while university and study level have limited influences.

Further analysis (as shown in Table 5) was carried out using the Games-Howell post hoc test of pairwise comparisons for AI literacy across participating countries. The analysis, which includes mean differences, standard errors, significance levels and 95% confidence intervals, reveals statistically significant disparities at the α = 0.05 level. Specifically, Ghana and Nigeria demonstrated significantly lower mean AI literacy scores compared to Rwanda and Uganda. Comparatively, Ghana exhibited a significantly lower mean score than Rwanda (M = -0.642, p = 0.045) and Uganda (M = -0.968, p = 0.012). Similarly, Nigeria's mean score was significantly lower than Rwanda's (M = -0.830, p = 0.008) and Uganda's (M = -1.156, p = 0.003).

In contrast, pairwise comparisons (see Table 5) involving South Africa did not yield statistically significant differences at the α = 0.05 level, indicating that South Africa's AI literacy scores were not significantly different from those of the other countries examined. These findings underscore significant inter-country variations in AI literacy, suggesting a requirement for targeted educational and developmental initiatives to address observed disparities and enhance digital competence, particularly in Ghana and Nigeria.

Table 6 presents a multinomial logistic regression analysis examining factors influencing AI literacy levels, with "No Experience" as the reference category. Key findings indicate that country and university enrollment significantly impact AI literacy at the Expert level (p = 0.042 and p = 0.006, respectively), with university attendance greatly increasing the likelihood of expert-level proficiency (Exp (B) = 4.564). For Advanced AI literacy, country (p = 0.017) has a negative association, while ELC is a strong positive predictor (p < 0.001, Exp(B) = 2.552).

Similarly, ELC remains a significant factor at the Intermediate level (p = 0.026, Exp(B) = 2.051), whereas country, university, gender and level of study show no significant effects. At the Basic literacy level, no variables reach statistical significance, though ELC (p = 0.063) shows a potential positive influence. Overall, these findings emphasise the role of national context, university education and language proficiency in shaping AI literacy levels, while gender and level of study appear to have limited effects.

This study also aims to identify the specific AI tools and applications that LIS students find most beneficial for their studies and professional development. Additionally, it examines the barriers they encounter when integrating AI technologies into their academic and anticipated professional work.

Table 7 delineates the distribution of AI tool utilisation among respondents, presenting both absolute user counts (n) and corresponding percentage distributions. ChatGPT exhibits the highest adoption rate with 39.6% of total AI tool usage, signifying its preeminent position. Google Bard (Gemini) and Bing demonstrate substantial usage, with 20.6% and 19.0% of respondents respectively employing these platforms. Quillbot, a tool designed for paraphrasing and writing assistance, is utilised in 28.6% of cases. Scholarcy and Perplexity AI show moderate usage, facilitating academic summarisation and information retrieval. Elicit and Sourcely exhibit lower adoption rates, indicating limited engagement. Notably, 39 respondents (4.9%) reported non-usage of AI tools, accounting for 8.9% of cases. Moreover, the respondents also use "Other AI tools" (1.6%), including Grammarly, Snapchat, Claude AI, Siri, Systran, Undetectable.ai, Consensus, AI Chat, Wordtune, Jenni AI, Question AI, etc.

Table 8 analyses the purpose and frequency of AI usage among respondents, categorised into four levels: Never (1), Rarely (2), Occasionally (3) and Frequently (4), presenting mean values (x̄) and standard deviations (σ). Information searching is the most frequent use (x̄=2.02, σ=0.858), with 78.5% of respondents using AI at least occasionally. Understanding concepts (x̄=1.84, σ=0.894) and answering questions (x̄=1.71, σ=0.913) are also common. Generating research ideas (x̄=1.68, σ=0.921) and finding related sources (x̄=1.72, σ=0.922) show substantial engagement. Translation (x̄=1.37, σ=1.013) and article summarisation (x̄=1.39, σ=0.974) are less frequent, with more respondents selecting "Never" or "Rarely." Proofreading (x̄=1.54, σ=1.104) and improving research productivity (x̄=1.58, σ=0.965) are moderately used. Overall, AI is primarily used for retrieving and understanding information, while translation and summarisation are less common, highlighting varying adoption rates across different research tasks.

The key challenges faced by African LIS students in AI adoption revolve around ethical concerns, technical complexities and resource limitations as illustrated in Table 9. The most significant issue is understanding the ethical implications of AI application and usage in library services, with 14.5% of responses highlighting this concern, including privacy issues (7.4%). A lack of AI learning resources and limited exposure to real-world AI applications further hinder students’ abilities to effectively engage with AI. Technical barriers, such as grasping AI concepts and understanding AI’s mathematical foundations, add to the difficulty. Additionally, integrating AI into traditional library practices (9.8%) and interpreting AI-generated data for decision-making (9.2%) pose practical challenges. Furthermore, assessing AI’s impact on user experience and managing AI-related ethical dilemmas indicates the need for structured AI ethics training.

An objective of this study was to examine LIS students' ethical perspectives on AI use, including their understanding of privacy, bias and misinformation concerns. Of surveyed LIS students (Figure 2), most (69.8%) expressed moderate to significant ethical concerns, with a notable minority reporting extreme apprehension. A small fraction noted minimal or no concerns. Results underscore a general recognition of AI-related ethical challenges among LIS students, with varying degrees of concern, thereby emphasising the necessity for the development and implementation of ethical frameworks and guidelines for the integration of AI in library service environments.

The data presented in Table 10 highlight critical ethical concerns in implementing AI for library services, emphasising privacy, security and fairness. Key issues include risks to system security and patron data privacy, alongside ethical challenges in AI-driven personalised recommendations and algorithmic bias in information retrieval. Transparency of AI decisions, equitable access to information, ethical cataloguing practices and legal/copyright complexities further underscore the need for robust ethical frameworks to guide responsible AI adoption in libraries.