Artificial intelligence in Greek libraries: Perceptions,challenges,and readiness for adoption

Abstract

This study examines how Greek librarians understand, engage with, and perceive Artificial Intelligence (AI) technologies. Using a cross-sectional online survey, data were collected on AI-related knowledge and familiarity, personal and professional use, perceived impact across operational domains, challenges to adoption, institutional and individual readiness, and the provision of AI literacy instruction. Respondents reported moderate to good overall understanding, with higher familiarity and use in personal contexts than in workplace settings. Common personal applications included text generation, editing, translation, and information seeking, while professional use focussed on assisting user queries, suggesting search terms, plagiarism detection, and developing instructional materials. Adoption of chatbots and AI-supported cataloguing remained limited. Perceived impact was strongest in areas related to user and staff training, as well as frontline services. Most challenges were rated above the midpoint—particularly infrastructure limitations, funding constraints, staff training needs, and ethical or legal concerns—while fear of job loss was relatively low. AI literacy instruction was infrequent across institutions. Comparisons by library type revealed no significant differences in perceived impact or challenges, though small but significant differences emerged in readiness and professional use. Based on these findings, the authors propose targeted actions to support the successful integration of AI in library services.

Keywords

Artificial intelligence librarians’ perceptions descriptive survey Greece libraries

Introduction

Artificial Intelligence (AI) is rapidly changing the library landscape, with scholars and practitioners exploring both its possibilities and its constraints. The broad availability of tools like ChatGPT has created new avenues for experimentation while highlighting the methodological and ethical issues that must be tackled (Formanek, 2025). A growing body of literature emphasises the impact of AI on almost every aspect of library work (Das and Islam, 2021; Mannheimer et al., 2024; Zhu et al., 2024). Although many studies focus on sociocultural issues and theoretical perspectives, increasing attention is being given to the practical applications of AI and its potential to enhance library services and operations (Vrindha and Syamili, 2025). Metadata and knowledge organisation have been significant areas for experimentation, as cataloguing represents one of the most resource-demanding tasks in libraries (Asula et al., 2021; Chou and Chu, 2022). Experiments with AI tools demonstrate their ability to create MARC records and suggest subject headings faster than human specialists; however, they often produce results of lower quality, and they have difficulties with complex, hierarchical vocabularies such as LCSH, underscoring the ongoing necessity for human monitoring and quality control (Asula et al., 2021; Brzustowicz, 2023; Chou and Chu, 2022; Chow et al., 2024). In user services, AI-powered tools enhance accessibility and alleviate routine tasks, yet they also present concerns regarding reliability, privacy, and maintenance. For instance, chatbots provide 24/7 assistance for common enquiries and help navigate electronic resources, though challenges persist in terms of accuracy, maintenance, and managing user expectations (Li and Coates, 2025; Panda and Chakravarty, 2022; Rodriguez and Mune, 2022; Yan et al., 2023). Recommender systems analyse patterns in user behaviour, personalise the discovery process and inform decisions related to collections development (Hu and Zhang, 2025; Xiao and Gao, 2020), but they raise privacy concerns and possible biases in machine-learning algorithms (Ekstrand et al., 2025; Gulsoy et al., 2023). Assistive technologies, such as speech translation, text recognition, and hands-free control, promote accessibility and inclusivity while requiring robust protective measures (Chauhan, 2024).

Although AI has begun to change libraries, the pattern of this transformation remains unclear. Some libraries are actively using AI technologies, others are exploring possible applications or considering implementation, while some have little interest in adopting them (Huang and Dong, 2024; Lo and Hudson Vitale, 2023). Effective and sustainable implementation is not solely a matter of technology. Studies repeatedly stress the importance of ethical standards, clear policies, risk management strategies and continuous professional development to equip staff with AI-related skills (Monyela and Tella, 2024; Moulaison-Sandy and Coble, 2024; Oyighan et al., 2024; Sussmeier and Henry, 2025). Prior work also indicates that librarians’ knowledge, perceptions, attitudes and expectations significantly influence the integration of AI in library services (Andrews et al., 2021; Juego and Masalinto, 2025; Yakubu et al., 2023)

Thus, and in response to Cox’s (2023) call for more empirical research on the role of AI in libraries, the aim of the present study is to investigate the perspectives of AI within libraries among information professionals in Greece. Despite the strong interest of the Greek library community in the potential and effects of AI in libraries—reflected by the hosting of recent conferences on the topic—there is a lack of empirical evidence regarding the practices and opinions of librarians. Addressing this gap, the present study surveys Greek library professionals to map current patterns of AI use and document perceptions and expectations. Gaining a deeper insight into librarians’ views of AI could be beneficial for implementation strategies and educational initiatives.

Literature review

Research on librarians’ knowledge, views, and feelings about AI offers insights into the way they comprehend and interact with it. Overall, the literature suggests an uneven and fragmented understanding of AI among librarians. According to some studies, self-assessed knowledge ranges from moderate to high (Lo, 2024; Lund et al., 2020). Many practitioners associate AI with Natural Language Processing (NLP), machine learning, or intelligent machines that can think and behave like people (Begum and Elahi, 2025; Harisanty et al., 2024; Yoon et al., 2022).

Nevertheless, complex concepts, like deep learning and generative adversarial networks, are less understood (Lo, 2024). Librarians often lack a shared understanding of what defines AI (Harisanty et al., 2024), and they have limited confidence in their ability to explain basic concepts or evaluate specific tools (Alam et al., 2024). Moreover, there is evidence that many librarians failed to identify AI tools as components of their services, despite these technologies being integrated into library systems and processes for some time (Hervieux and Wheatley, 2021)

Irrespective of their level of knowledge, librarians tend to hold positive views of AI and increasingly see it as a driver of innovation. They believe that it provides numerous advantages for both library personnel and patrons. They view reference services as the area most likely to benefit from AI (Abayomi et al., 2021; Begum and Elahi, 2025; Yoon et al., 2022).

There is a consensus that a key advantage of AI is its ability to enhance the overall user experience. AI-based search technologies, such as NLP and machine learning, enable more efficient and precise retrieval of information (Alam et al., 2024; Kisilowska-Szurmińska, 2025; Lund et al., 2020; Subaveerapandiyan and Gozali, 2024). In addition, ChatGPT is recognised for its ability to help users in performing complex database searches and speeding up the retrieval of information (Patra et al., 2025; Rahman and Islam, 2024). The literature also emphasises the value of personalisation: some librarians are excited about 24/7 support through chatbots and virtual assistants (de Leon et al., 2024; Kaushal and Yadav, 2022), while others point out that one of AI’s major advantages is customised user support (Eiriemiokhale and Sulyman, 2023; Shahzad et al., 2024; Yang et al., 2024).

Another key area of perceived benefit is operational efficiency and automation. Librarians consistently identify AI as very effective for routine and repetitive tasks, such as circulation and shelving, allowing staff to focus on more specialised and challenging responsibilities (Abayomi et al., 2021; Alam et al., 2024; Bøyum and Khosrowjerdi, 2025; Eiriemiokhale and Sulyman, 2023; Harisanty et al., 2024; Lulu-Pokubo and Okwu, 2025; Lund et al., 2020; Rahman and Islam, 2024; Yoon et al., 2022). It can also streamline technical services workflows, particularly by increasing cataloguing accuracy and efficiency and simplifying the creation of metadata (Begum and Elahi, 2025.; Bøyum and Khosrowjerdi, 2025; Kisilowska-Szurmińska, 2025; Lulu-Pokubo and Okwu, 2025; Lund et al., 2020). Beyond routine operations, librarians recognise AI as a useful tool for library administration, decision-making, and collection management. According to several studies, it is valued for advanced data analysis, which helps libraries to understand usage trends and make evidence-based decisions (Alam et al., 2024; Yoon et al., 2022). AI applications are also acknowledged as beneficial in managing human resources and physical infrastructure, thereby enhancing efficiency and supporting strategic planning (Subaveerapandiyan and Gozali, 2024).

Professional development represents an additional domain in which AI is considered beneficial. Yang et al. (2024) found that subject librarians appreciated AI as a valuable aid in staying updated on recent developments and research findings, which in turn strengthened their expertise and confidence. This implies that in addition to changing library operations and services, AI is also facilitating librarians’ ongoing learning and growth.

Alongside optimism, librarians raise several ethical, organisational and professional considerations related to the adoption of AI. Ethical issues remain a dominant concern, with librarians frequently citing risks such as bias, discrimination, lack of transparency, privacy violations, data security, and threats to intellectual freedom (Bøyum and Khosrowjerdi, 2025; de Leon et al., 2024; Grote et al., 2024; Hlatshwako and Tsabedze, 2024; Kisilowska-Szurmińska, 2025; Rahman and Islam, 2024; Subaveerapandiyan and Gozali, 2024).

Another challenge is the quality, accuracy and reliability of automatically generated content (Bøyum and Khosrowjerdi, 2025; Lo, 2024). As Cox et al. (2019) caution, there is a danger that “robo-content” may be incorporated into library collections or discovery systems, undermining the library’s role as a source of quality information. Fears about job security and changing job roles also arise, though their intensity varies across contexts (de Leon et al., 2024; Kaushal and Yadav, 2022; Kisilowska-Szurmińska, 2025; Lulu-Pokubo and Okwu, 2025; Lund et al., 2020; Patra et al., 2025; Subaveerapandiyan and Gozali, 2024).

At the institutional level, challenges include unclear policies, insufficient infrastructure, and lack of support (Begum and Elahi, 2025; Bøyum and Khosrowjerdi, 2025; Hussain and Khan, 2025). Adoption is further impeded by financial barriers like inadequate budgets, high costs, and energy demands (Huang and Dong, 2024; Kalbande et al., 2024; Kaushal and Yadav, 2022). Professional obstacles, such as low AI literacy and resistance to change, are also significant (Alam et al., 2024; Cox et al., 2019). Finally, concerns around vendor dependency and marketisation raise questions about autonomy and credibility (Cox et al., 2019; Grote et al., 2024).

Several studies have examined the factors influencing librarians’ intentions to adopt AI. Attitude emerges as the central driver of readiness and intention: those who regard AI as advantageous are more inclined to integrate it into their professional activities (Andrews et al., 2021; Juego and Masalinto, 2025; Lund et al., 2020; Yakubu et al., 2023).

Attitude is shaped by antecedent perceptions of performance expectancy—belief that AI improves job performance, ease of use, and interactivity (Andrews et al., 2021; Hussain and Khan, 2025; Yang et al., 2024). Facilitating conditions, such as access to technical resources, training and reliable infrastructure, add weight, reinforcing intentions to use (Fang et al., 2025). These results suggest that the acceptance of AI is complex and depends on a dynamic interplay between personal beliefs and institutional resources. Therefore, efforts to encourage AI use in libraries should focus equally on cultivating positive perceptions and creating the practical conditions that empower librarians to implement them.

The literature shows a clear trajectory from moderate knowledge and limited hands-on experience to growing confidence in AI’s potential to enhance library services. AI is viewed less as a threat and more as a tool that can complement professional expertise, offering efficiency gains and improved user experience. At the same time, professionals are aware of the challenges: concerns about content quality, privacy, bias, technical reliability, costs and infrastructure persist. They emphasise that successful adoption will require not only robust technology but also careful attention to ethics, training and organisational readiness.

Aim and research questions

The aim of the present study is to explore Greek librarians’ understanding, use and perceptions of AI. In this context, the study seeks to address the following research questions:

To what extent do librarians understand the concept of AI, and how familiar are they with using it?

How do librarians perceive their readiness, the barriers to adoption, and the expected impact of AI on library operations?

Do librarians’ perceptions (readiness, impact, challenges) and reported AI use differ by library type?

To what extent do libraries offer AI user training, and which topics are covered?

Methodology

This study employed a quantitative, descriptive research design based on an online questionnaire, which was distributed in May 2025 via a national library listserv. The survey was addressed to all categories of library personnel in Greece, regardless of library type or position, to capture perspectives across a diverse range of libraries. Before completing the questionnaire, each participant was informed about the purpose of the study, the voluntary nature of participation, and the anonymity of their responses, and provided informed consent prior to proceeding with the survey.

The instrument consisted of closed-ended questions, and it was organised into the following sections:

Demographic information.

Familiarity with AI.

This section of the questionnaire included questions on how well participants understand AI, how familiar they are with using AI tools, and the frequency and reasons for use in both their personal and professional lives.

3. Perceived readiness, barriers, and impact

Respondents were asked to indicate their level of agreement with seven statements regarding their library’s preparedness to integrate AI tools, as well as their own personal readiness, including their confidence in using AI tools, participating in AI-related discussions, and evaluating ethical issues. To assess perceived barriers, respondents were presented with a list of potential challenges and asked to express their agreement with statements such as “There is a lack of technological infrastructure,” “Funding is insufficient,” or “There is a risk of job loss due to AI.” All responses were recorded on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). Respondents also rated the expected degree to which AI would affect certain areas of library operations. Responses were recorded on a five-point Likert-type scale (1 = Not at all, 5 = To a great extent).

4. User training

Respondents indicated whether their library offers training to users on the use of AI tools, and they selected all applicable topics from a list.

A total of 266 valid responses were collected, a sample that is considered adequate for exploratory purposes and provides useful insights into current trends and attitudes within the Greek library sector. Not all respondents answered every question; therefore, the number of valid responses (N) varies slightly across analyses. Quantitative data were analysed using SPSS (Statistical Package for the Social Sciences). Descriptive statistics were used to summarise responses, and, where relevant, inferential analyses, including t-tests and analysis of variance (ANOVA), were employed to examine differences across variables.

Results

Background characteristics of the respondents

The sample consisted of 266 library professionals. The majority of participants were female, with males making up less than 20% of the sample. The educational background was notably high, as most participants reported postgraduate degrees either in LIS or in other fields. Most participants indicated they had over 20 years of experience in libraries, with relatively few reporting less than a decade of experience. As for the type of institutions, academic libraries were the most frequently represented, followed by public and special libraries (Table 1).

Table 1.

Background characteristics of the respondents (N = 266).

Characteristic	n	%
Gender
Female	208	78.2
Male	47	17.7
Prefer not to say	11	4.1
Education
Bachelor’s degree	89	33.5
Master’s in another field	79	29.7
Master’s in library science	64	24.1
Doctorate	20	7.5
Master’s in library science and master’s in another field	6	2.3
Prefer not to say	8	3.0
Professional experience
More than 20 years	173	65.0
11–20 years	45	16.9
Less than 5 years	21	7.9
5–10 years	14	5.3
Prefer not to say	13	4.9
Library type
Academic library	152	57.1
Public library	43	16.2
Special library	42	15.8
Other	8	3.0
Prefer not to say	21	7.9

Respondents were also asked to indicate their areas of work. They reported involvement in a wide range of roles, often covering multiple responsibilities. User services and technical services, including acquisitions, cataloguing, and metadata creation, were the most reported areas. A considerable number of respondents were also involved in collection development, as well as user education and information literacy (Table 2).

Table 2.

Participants’ job areas.

Job area	n	%
User services	145	18.1
Technical services	137	17.1
Collection development and management	102	12.7
Information literacy/user education	100	12.5
Library promotion and outreach	85	10.6
Administration	84	10.5
Digital services	81	10.1
Systems services	61	7.6
Other	8	1

Note. Percentages are of responses; participants could select multiple options.

Knowledge and familiarity

Most respondents reported at least moderate understanding of AI concepts (moderate 41.2%, high 29.0%, very high 9.9%; Valid N = 262), whereas familiarity with specific AI tools was lower (23.7% high/very high, 76.2% moderate or below; Valid N = 261). Access to paid AI tools was uncommon: only 7.9% indicated an organisational subscription, and 7.1% reported a personal subscription. On this basis, we examined how often librarians use AI and for what purposes. Participants demonstrated a slightly higher level of interaction with AI in personal contexts compared to professional settings. However, roughly 20% of the respondents stated they never utilised AI, whether in their personal lives or work activities. Among those who acknowledged using AI, frequency varied considerably, ranging from occasional use (less than once per month or about once per month) to regular engagement several times per week (Table 3).

Table 3.

Frequency of use of AI tools.

Frequency of use	Personal use n (%)	Work use n (%)
Never	45 (16.9)	58 (21.8)
Less than once per month	46 (17.3)	35 (13.2)
About once per month	22 (8.3)	22 (8.3)
2–3 times per month	39 (14.7)	35 (13.2)
1–2 times per week	37 (13.9)	27 (10.2)
3–4 times per week	23 (8.6)	20 (7.5)
Daily or almost daily	27 (10.2)	27 (10.2)
Prefer not to say	27 (10.2)	42 (15.8)
Total	266 (100.0)	266 (100.0)

As presented in Table 4, the most frequently reported reasons for personal use of AI tools include creating and editing text (such as emails, essays, or letters), information seeking, and text translation. In the “Other” category, although most participants mentioned practical reasons, a small number provided more reflective and exploratory uses such as decision support (using AI to evaluate choices in personal dilemmas); exploration and experimentation (using AI tools merely to observe their functionality, assess their response accuracy, or examine their reasoning abilities); creative experimentation (composing songs); and curiosity (interacting with AI purely out of curiosity about its capabilities).

Table 4.

Reasons for personal use.

Reason for use (Personal)	n	%
Creating/editing text (e.g., letters, essays, emails)	153	24
Information seeking	139	21.8
Text translation	114	17.9
Education and self-development	73	11.4
Finding ideas for articles, social media or other content	58	9.1
Creating visuals (DALL·E, or Midjourney)	56	8.8
Writing CVs or cover letters	25	3.9
Entertainment or gaming with AI bots	12	1.9
Other	8	1.3
Total	638	100

Note. Percentages are of responses; participants could select multiple options.

Workplace use of AI mainly focuses on supporting users and maintaining academic integrity, with applications such as assisting with research queries, plagiarism detection, and creating educational materials being most frequently mentioned. In contrast, use in technical services and metadata work was reported significantly less. Tools such as virtual assistants were also very uncommon, while other specialised applications—such as document extraction and subject indexing—were noted even less frequently. “Other” uses of AI included developing promotional activities, designing greeting cards, and locating information about library operations or best practices. Some participants highlighted linguistic support, such as refining foreign-language texts and identifying topics or themes within broad texts. A few noted the use of AI for various technical and administrative functions, including aid with HTML, CSS, and SQL queries, especially in Koha reporting. Some even mentioned using AI to document procurement of digital resources—like databases and journals—and experimenting with AI-powered tools and workflows (Table 5).

Table 5.

Reasons for use in the workplace.

Reason for use (Professional)	n	%
Assisting with user queries and research	60	21.7
Plagiarism detection and prevention	47	17.0
Creation of educational material	38	13.7
Generation of suggested search terms (e.g., via recommender or ChatGPT)	31	11.2
User support through chatbots or virtual assistants	19	6.9
Metadata creation	17	6.1
Metadata enhancement	17	6.1
PDF analysis and bibliographic data extraction	12	4.3
Automatic subject indexing	9	3.2
Other	27	9.7
Total	277	100

Note. Percentages are of responses; participants could select multiple options.

Perceptions of AI

Internal consistency of the scale measuring perceived readiness for AI adoption in libraries was satisfactory (α = 0.807). Based on this result, a new variable was created, representing the mean scores of all items for each participant. One-sample t-test was used to examine whether the mean of the new variable was statistically different from the midpoint of the responding scale (3). Results showed that librarians reported decreased levels of readiness. This difference, apart from being statistically significant, was also very meaningful, as suggested by Cohen’s d value (Table 6). Descriptive statistics for the items are presented in Table 7. Mean scores suggest that institutional readiness, such as implementation plans and policies, was rated lowest, while personal capabilities were rated moderately higher.

Table 6.

One-sample t-test analysis of librarians’ perceptions.

Variable	M (SD)	t (df)	p-value	Cohen’s d	Cronbach’s α
Perceived readiness	2.79 (0.71)	4.84 (258)	0.001	0.702	0.807
Perceived impact	3.57 (0.77)	11.872 (256)	0.001	0.767	0.892
Perceived challenges	3.79 (0.75)	17.014 (256)	0.001	0.908	0.880

Table 7.

Descriptive statistics of perceived readiness for AI adoption.

Readiness indicator	n	Mean	SD
Institutional readiness
Library has technological infrastructure	255	2.75	1.05
Management supports AI use	256	2.74	1.07
Library has or is developing AI policies	254	2.33	1.09
Library has an AI implementation plan	256	2.21	1.03
Personal readiness
I can evaluate ethical issues related to AI	256	3.29	0.86
I feel ready to use AI tools	257	3.10	1.07
I can engage in discussions on AI use	257	3.07	1.05

Similar procedures and analyses were carried out for the six items measuring respondents’ perceptions about the possible impact of AI in libraries. The scale demonstrated excellent reliability (α = 0.892). Participants’ mean scores were statistically significant from the scale’s midpoint (3), indicating that library personnel generally anticipate AI to play a significant role in library operations (Table 6). Mean values of the items ranged from 3.36 (internal processes) to 3.78 (user education). Respondents expected AI to have the greatest impact on user and staff training, followed by technical and user services. Other areas, such as collection management and internal processes management, were rated positively but somewhat lower (Table 8).

Table 8.

Descriptive statistics of perceived areas of AI impact.

Operational domain	n	Mean	SD
User education	250	3.78	0.88
Professional development/staff training	254	3.68	0.93
Technical services	255	3.64	0.95
User services	257	3.60	0.97
Collection development	254	3.37	0.99
Internal processes	255	3.36	0.99

Respondents were also asked to assess the likelihood that various issues may present challenges to the implementation of artificial intelligence in their libraries. The scale demonstrated excellent internal consistency (α = 0.880). The overall mean score across the listed challenges reached statistical significance, a finding indicating that librarians perceive multiple barriers to AI adoption (Table 6). They rated most challenges above the midpoint, particularly infrastructural limitations, insufficient funding, lack of staff training, and ethical or legal concerns (Table 9).

Table 9.

Descriptive statistics of perceived challenges to AI adoption.

Barrier/Challenge	n	Mean	SD
Insufficient funding	256	4.04	1.09
Legal concerns	255	3.98	0.97
Inadequate staff training	257	3.96	0.99
Data privacy and security	256	3.83	1.03
Ethical issues (e.g., bias)	256	3.79	1.03
Lack of technological infrastructure	255	3.73	1.08
Accuracy and quality of AI-generated content	257	3.68	0.96
Staff resistance	256	3.66	1.06
Potential job loss	255	3.47	1.20

Differences among types of libraries

One-way ANOVA was employed to examine whether the mean scores of the three composite variables (perceived readiness, perceived impact and perceived challenges) and the frequency of professional use differed in relation to the type of library. Based on the frequencies of each library type (see Table 1), we focussed on academic (n = 152), public (n = 43) and special (n = 42) libraries. Perceived impact and challenges did not differ by library type (all ps ⩾ 0.21), suggesting broadly similar expectations and concern profiles across academic, public, and special libraries. Significant differences emerged only for the perceived readiness and the frequency of use. Post-hoc comparisons using the Bonferroni criterion revealed that academic librarians reported higher readiness for the integration of AI in libraries than public librarians, with special libraries not differing from either. Similar post-hoc analysis indicated that frequency of use was higher among special librarians in comparison to public librarians, while other pairwise differences were not significant (Table 10).

Table 10.

Analysis of variance results for the four examined variables across the three types of libraries.

Variable	Public	Academic	Special
Variable	M (SD)	M (SD)	M (SD)
Perceived readiness	2.50 (0.65)^a	2.83 (0.67)^a	2.73 (0.70)
AI impact	3.38 (0.94)	3.57 (0.72)	3.66 (0.76)
AI challenges	3.74 (0.74)	3.81 (0.70)	3.74 (0.87)
Frequency of use	2.81 (1.93)^b	3.35 (2.05)	4.14 (2.19)^b

Significant differences between public and academic libraries (p < 0.05).

Significant differences between public and special libraries (p < 0.05).

User education

Participants were asked whether their library offers user education concerning the responsible use of AI tools. As shown in Table 11, most respondents (69.1%) reported that no such education is currently provided, while only 17.0% reported its existence.

Table 11.

User education on AI.

Provision of instruction	N	%
Yes	45	17.0
No	184	69.1
Don’t know/no answer	37	14.0
Total	266	100.0

Among the 45 respondents who indicated that their library offers user education on the responsible use of artificial intelligence (AI) tools, a multiple response analysis was conducted to examine the specific topics covered. As summarised in Table 12, the most frequently selected topics were avoiding plagiarism through AI use and evaluating the reliability of AI-generated information. Other covered areas included identifying misinformation, developing critical thinking skills, understanding the limitations of AI tools, and addressing ethical issues and privacy concerns.

Table 12.

Topics covered in user education.

Instructional topic	N	%
Avoiding plagiarism with AI	39	23.8
Evaluating reliability of information	36	22.0
Identifying misinformation	25	15.2
Developing critical thinking	24	14.6
Understanding limitations of AI	21	12.8
Ethical issues & privacy	16	9.8
Other	3	1.8
Total	164	100

Note. Percentages are of responses; participants could select multiple options.

Discussion

Most respondents indicated moderate to good understanding of AI, echoing findings from previous studies that librarians are generally knowledgeable about AI (Begum and Elahi, 2025; Lund et al., 2020). Only a small fraction reported very good knowledge, while roughly one in five reported low or no knowledge. This pattern suggests a good grasp of core concepts alongside a lower understanding of advanced topics (Lo, 2024). Familiarity and use followed the same trend: most individuals feel fairly familiar, with smaller groups at the high and low ends, and usage was more prevalent in personal settings than in professional ones. This is consistent with earlier findings that personal engagement exceeds professional adoption (Hervieux and Wheatley, 2021) and that organisational use is generally infrequent (Abayomi et al., 2021; Lo, 2024; Yoon et al., 2022). Subscription access to AI tools was uncommon at both institutional and personal levels. Such constraints mirror reports of low premium uptake, which has been associated with lower self-rated AI literacy and confidence (Lo, 2024), and may impede sustained, work-embedded practice.

The everyday use of AI reported by participants for text generation, editing and translation mirrors findings that librarians who do use AI tend to focus on content creation and research support (Lo, 2024). Studies on everyday interactions with AI indicate that such engagement is primarily motivated by productivity, learning, and convenience (Skjuve et al., 2024; Wolf and Maier, 2024). Similarly, the tasks most frequently reported in the literature—such as writing, information seeking, summarisation, and language assistance—align with those identified in the present study. Our results also revealed that the use of AI tools is somewhat more prevalent in personal than in professional contexts, with librarians tending to experiment and adopt these technologies more readily at home than in the workplace. This finding is in line with recent evidence, where personal use modestly exceeds use for work or academic purposes (Bick et al., 2024; Gasaymeh et al., 2025).

Workplace use primarily involved user-support functions, a pattern aligning with previous research that identifies reference services and research assistance as main areas of AI application (Hervieux and Wheatley, 2021; Lo, 2024; Lund et al., 2020; Yoon et al., 2022). In addition, across multiple studies, respondents consistently recognise AI’s potential to improve information retrieval, support research processes, and foster knowledge discovery (Ali et al., 2025; Harisanty et al., 2024; Hervieux and Wheatley, 2021; Lulu-Pokubo and Okwu, 2025; Patra et al., 2025; Yoon et al., 2022). By contrast, the very limited use of virtual assistants—and the similarly infrequent use in cataloguing and metadata creation—is inconsistent with the well-documented, widespread deployment of chatbots and AI in technical services and routine operations (Huang and Dong, 2024; Patra et al., 2025; Rahman and Islam, 2024; Yang et al., 2024).

Our respondents generally anticipated a significant role for AI in library operations, with the strongest expected effects in user and staff training, followed by technical and user services. These results corroborate the findings of a great deal of the previous work on the perceived impact of AI (e.g. Begum and Elahi, 2025; Harisanty et al., 2024; Juego and Masalinto, 2025; Yang et al., 2024). The findings of the present study also confirm an emerging trend that AI is increasingly viewed not only as a technical tool but also as a resource for user education. Pantzar (2023) reported that while many school librarians doubted ChatGPT’s reliability, they acknowledged its value as a tool for teaching information literacy. Similarly, the findings of Shahzad et al. (2024) revealed participants’ perceptions about the potential of AI to promote innovative learning.

Consistent with this emphasis on learning, we examined whether libraries offer AI instruction. Provision of AI literacy training programmes was very limited. Where available, instruction focussed mainly on academic integrity (such as avoiding plagiarism) and evaluating AI-generated information, alongside misinformation, critical thinking, limitations of tools, and issues of ethics and privacy. This scarcity and emphasis on evaluation mirror findings from academic libraries in North America, where only a minority reported hosting workshops for patrons, and the topics revolved around technical applications and sociotechnical issues (Hervieux and Wheatley, 2021). Notably, a small but meaningful group of respondents in our study were uncertain whether their library offers AI literacy instruction. One possible explanation for this might be professional segmentation: employees who work in “behind-the-scenes” departments (such as cataloguing, systems, or acquisitions) might be less exposed to educational offerings and less aware of such services. The ambiguity might also be a result of poor internal communication regarding AI-related endeavours.

Findings regarding limited perceived readiness are in line with Lo (2024), who reported that most academic librarians in the U.S. considered their institutions “not prepared” to integrate generative AI within a year, and many felt personally unprepared. Hussain and Khan (2025) likewise note that, despite optimism, librarians expressed concerns about institutional preparedness, skills gaps, and low AI literacy. In our data, the lowest ratings were clustered around implementation strategies and policies, while perceived AI-related skills received comparatively higher scores. These trends suggest that individual competencies are advancing more rapidly than institutional frameworks. The findings on perceived readiness are closely intertwined with the challenges librarians associated with AI. Participants expected several constraints that could impede or complicate adoption, a finding that confirms previous studies highlighting, among other things, ethical concerns, budget limitations, insufficient infrastructure, and lack of technical expertise (e.g. de Leon et al., 2024; Huang and Dong, 2024; Rahman and Islam, 2024). Overall, these findings suggest that to ensure effective and sustainable implementation, libraries should prioritise technical capacity and budgets, staff development, and clear policy frameworks that leverage the increasing personal proficiency with AI tools. Interestingly, while organisational constraints were salient, concerns about job loss were among the least frequently selected. This finding is consistent with research that portrays AI as a supportive tool that complements professional judgement rather than replaces it (Lund et al., 2020; Patra et al., 2025; Subaveerapandiyan and Gozali, 2024), while it contrasts with studies reporting greater concerns about job displacement, particularly in repetitive tasks (de Leon et al., 2024; Kaushal and Yadav, 2022; Kisilowska-Szurmińska, 2025; Lulu-Pokubo and Okwu, 2025). Variations in professional roles, labour market conditions, and temporal context likely contribute to the divergence, suggesting that job security concerns are not the same everywhere.

Both expectations and concerns were generally similar across different library types, whereas academic librarians demonstrated greater readiness, and usage of AI was more prevalent in special libraries. This disparity in readiness aligns with findings from Yoon et al. (2022), which indicated a higher usage in academic environments, a pattern typically associated with enhanced preparedness. On the contrary, Yoon et al. (2022) noted that public librarians had more optimistic expectations, while our findings did not reveal any differences. These discrepancies could be attributed to context, timing, or measurement methods. The relatively increased usage in special libraries likely stems from their specialised workflows and suitability of tools, an issue that has not been sufficiently explored in the literature. Despite having similar expectations, public libraries appear less prepared and less engaged, suggesting that efforts should focus on targeted investment and support rather than on changing attitudes.

Conclusions

This study offers a preliminary overview of Greek librarians’ perceptions of AI, its expected significance, and the obstacles to its adoption. Participants indicated a fair to good understanding of the concept and greater personal than workplace use, focussed on content creation and research support. They anticipated meaningful impact on various operations—particularly in user and staff training as well as frontline services—but also recognised certain organisational barriers.

These findings have both practical and theoretical implications. From the practice viewpoint, they confirm the need for Greek libraries to focus on the ongoing training and upskilling of staff, and to establish clear, institution-wide frameworks governing the responsible use of AI. Furthermore, integrating AI and GAI education into Library and Information Science (LIS) curricula is critical for preparing future professionals to meet the demands of a rapidly evolving digital landscape. Empowering librarians with the knowledge and competencies they need to effectively utilise AI tools is crucial—not only for internal operations but also for guiding patrons in their use. The development of appropriate infrastructure, along with transparent policies and guidelines, will minimise uncertainty and promote consistent practices. While librarians should be involved in discussions surrounding policies, the responsibility for creating and updating these frameworks falls to library leadership. Policies ought to be regularly assessed and modified to keep pace with technological progress and shifting ethical considerations. Establishing a unified regulatory framework for AI usage across libraries in Greece would further improve consistency and accountability. Collaboration with AI tool developers, in addition to interlibrary cooperation, will be essential for sharing experiences, exchanging insights, and creating a supportive professional community. Collectively, these actions can support data-informed decision-making and ensure the ethical and effective use of AI across library functions.

From the theoretical perspective, this study offers an addition to the overall body of knowledge regarding AI adoption in libraries, by extending our understanding of AI acceptance to a national context that has been underrepresented in the literature. The situation in Greece provides a clear example of how institutional structures—such as policies, infrastructure, and managerial support—interact with individual factors, including technological competence and ethical awareness, to shape librarians’ opinions, attitudes and readiness for AI adoption. This contextualised approach emphasises that AI integration in libraries is not uniform but influenced by organisational culture, resources, and national conditions. Accordingly, the findings offer valuable insights for comparative research across diverse cultural and organisational settings, enriching theoretical frameworks that explain how technological innovation is adopted within the information profession.

Limitations and future research

This research has several limitations that should be considered when interpreting its findings. The use of a specific national listserv for survey distribution may have led to selection and nonresponse biases, favouring individuals who are subscribers and willing to participate in such studies. Furthermore, relying on self-reported data raises concerns about various biases, including social desirability and response bias, which could impact the accuracy and objectivity of the results. The cross-sectional design provides a snapshot at one point in time, restricting the ability to observe changes over time or establish causal relationships. Additionally, the rapid advancement of AI technology suggests that the relevance of the findings could decrease quickly as new technologies emerge. Lastly, the study is limited to the Greek context, and its conclusions may not be applicable to other settings.

As this study was exploratory and descriptive in nature, it did not employ a formal theoretical framework. Its primary aim was to establish baseline evidence on how Greek librarians perceive and engage with artificial intelligence, as no prior empirical research on this topic exists in Greece. Future investigations could build on these findings by adopting established theoretical models—such as the Technology Acceptance Model (TAM) or the Unified Theory of Acceptance and Use of Technology (UTAUT)—to examine the specific determinants and predictors of AI adoption in Greek libraries. Applying such frameworks would allow for hypothesis testing and provide deeper insight into the mechanisms underlying librarians’ technology-related attitudes and behaviours. Future work should also combine longitudinal surveys with qualitative designs to capture the evolution of perceptions and practices over time. Moreover, evaluating the impact of training interventions on librarians’ readiness and actual use of AI tools could help identify effective strategies for professional development. In addition, future studies should seek to connect subjective perceptions with objective indicators, offering a more robust assessment of AI integration. Given the relatively higher uptake observed in special libraries, closer examination of their workflows and operations could yield valuable insights, while comparative analyses across library types may further illuminate how institutional context shapes attitudes and adoption patterns.

Footnotes

ORCID iDs

Aspasia Togia

Nikolaos Tsigilis

Eleftheria Koseoglou

Ethical considerations

The study involved an anonymous online survey of library professionals in Greece. As the questionnaire focussed on professional practices and workplace experiences of library employees, the subject matter was deemed to be of low sensitivity and posed minimal risk to participants. The study was conducted in accordance with relevant institutional and national guidelines for research with human participants.

Consent to participate

Before accessing the questionnaire, all potential participants were informed about the purpose of the study, the voluntary nature of their participation, and the procedures for protecting their data. All respondents provided informed consent after being informed of the study’s purpose and data-protection procedures.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Author biographies

Aspasia Togia is an Associate Professor at the Department of Library, Archival and Information Studies, International Hellenic University, Greece. She has authored numerous peer-reviewed publications on topics related to information behaviour, information literacy, scholarly communication, and research methods in LIS. Her research explores the ways in which information practices shape learning, research, and knowledge production. She has a strong interest in empirical approaches to understanding information use across diverse contexts.

Nikolaos Tsigilis is a Professor at United Arab Emirates University. He has extensive teaching experience, specializing in research methods, psychometrics, statistics, and educational assessment. His research focuses on test and instrument development and validation in education and psychology, with over 100 peer-reviewed publications. He has participated in numerous funded projects and received national and international awards.

Eleftheria Koseoglou is an information professional at the Aristotle University of Thessaloniki Library & Information Centre, Greece. She holds an MSc in Information Management in Libraries, Archives, and Museums, an MSc in Cultural Organizations Management, and bachelor’s degrees in Philology and Library Science. Since 2001, her work has focused on web content management, metadata quality control, and the development of electronic library services, including integrated discovery systems and scholarly publishing platforms. Her professional interests include information literacy, plagiarism detection systems, and quality assurance in academic libraries. She is also involved in institutional evaluation initiatives and international staff training activities.

References

Abayomi

Adenekan

Abayomi

, et al. (2021) Awareness and perception of the artificial intelligence in the management of university libraries in Nigeria. Journal of Interlibrary Loan Document Delivery & Electronic Reserve 29(1–2): 13–28.

Alam

Subaveerapandiyan

Mvula

, et al. (2024) AI literacy and Zambian librarians: A study of perceptions and applications. Open Information Science 8(1): 20220166.

Ali

Naeem

Bhatti

(2025) Artificial intelligence (AI) applications and usage among the LIS professionals of Pakistan. Journal of Librarianship and Information Science 57(3): 787–800.

Andrews

Ward

Yoon

(2021) UTAUT as a model for understanding intention to adopt AI and related technologies among librarians. Journal of Academic Librarianship 47(6): 102437.

Asula

Makke

Freienthal

, et al. (2021) Kratt: Developing an automatic subject indexing tool for the National Library of Estonia. Cataloging & Classification Quarterly 59(8): 775–793.

Begum

Elahi

(2025) Readiness to adopt artificial intelligence (AI) in libraries: A study from the lens of information professionals of Bangladesh. Digital Library Perspectives 41: 666–682.

Bick

Blandin

Deming

(2024) The Rapid Adoption of Generative AI. On the Economy Blog (Federal Reserve Bank of St. Louis). Available at: https://www.stlouisfed.org/on-the-economy/2024/sep/rapid-adoption-generative-ai (accessed 20 July 2025).

Bøyum

Khosrowjerdi

(2025) ‘It’s such a blackbox function’: Norwegian academic librarians’ ambivalent perceptions towards AI use in academic libraries and academia. Information Research: An International Electronic Journal 30(CoLIS): 91–104.

Brzustowicz

(2023) From ChatGPT to CatGPT: The implications of artificial intelligence on library cataloging. Information Technology and Libraries 42(3): 1–22. Available at: https://doi.org/10.5860/ital.v42i3.16295 (accessed 20 July 2025).

10.

Chauhan

SPS

(2024) Enhancing Accessibility In Special Libraries: A Study on AI-Powered Assistive Technologies For Patrons With Disabilities. arXiv preprint arXiv:2411.06970. Available at: http://arxiv.org/abs/2411.06970 (accessed 20 July 2025).

11.

Chou

Chu

(2022) An analysis of BERT (NLP) for assisted subject indexing for project Gutenberg. Cataloging & Classification Quarterly 60(8): 807–835.

12.

Chow

EHC

Kao

(2024) An experiment with the use of ChatGPT for LCSH subject assignment on electronic theses and dissertations. Cataloging & Classification Quarterly 62(5): 574–588.

13.

Cox

(2023) How artificial intelligence might change academic library work: Applying the competencies literature and the theory of the professions. Journal of the Association for Information Science and Technology 74(3): 367–380.

14.

Cox

Pinfield

Rutter

(2019) The intelligent library: Thought leaders’ views on the likely impact of artificial intelligence on academic libraries. Library Hi Tech 37(3): 418–435.

15.

Das

Islam

MSU

(2021) Application of Artificial Intelligence and Machine Learning In Libraries: A Systematic Review. arXiv preprint arXiv:2112.04573 [cs.DL]. Available at: https://arxiv.org/abs/2112.04573 (accessed 20 July 2025).

16.

de Leon

LCR

Flores

Alomo

ARL

(2024) Artificial intelligence and Filipino academic librarians: Perceptions, challenges and opportunities. Journal of the Australian Library and Information Association 73(1): 66–83.

17.

Eiriemiokhale

Sulyman

(2023) Awareness and perceptions of artificial intelligence among librarians in university libraries in Kwara State, Nigeria. Indonesian Journal of Librarianship 7(2): 107–118.

18.

Ekstrand

Razi

Sarcevic

, et al. (2025) Recommending with, not for: Co-designing recommender systems for social good. ACM Transactions on Recommender Systems. Available at: https://doi.org/10.1145/3759261 (accessed 20 July 2025).

19.

Fang

Alam

(2025) Usage intention of AI among academic librarians in China: Extension of UTAUT model. Sustainability 17(7): 2833.

20.

Formanek

(2025) Exploring the potential of large language models and generative artificial intelligence (GPT): Applications in library and information science. Journal of Librarianship and Information Science 57(2): 568–590.

21.

Gasaymeh

Abu Qbeita

AlMohtadi

, et al. (2025) Exploring education students’ use of ChatGPT for academic and personal purposes: Insights from a developing country context. Frontiers in Education 10: 1580310.

22.

Grote

Lysholm Faber

Gasparini

(2024) Artificial intelligence in PhD education: New perspectives for research libraries. LIBER Quarterly: The Journal of the Association of European Research Libraries 34(1): 1–29.

23.

Gulsoy

Yalcin

Bilge

(2023) Robustness of privacy-preserving collaborative recommenders against popularity bias problem. PeerJ Computer Science 9: e1438.

24.

Harisanty

Anna

NEV

Putri

, et al. (2024) Leaders, practitioners and scientists’ awareness of artificial intelligence in libraries: A pilot study. Library Hi Tech 42(3): 809–825.

25.

Hervieux

Wheatley

(2021) Perceptions of artificial intelligence: A survey of academic librarians in Canada and the United States. Journal of Academic Librarianship 47(1): 102270.

26.

Hlatshwako

Tsabedze

(2024) Unlocking the future: Exploring librarians’ perspectives and readiness for artificial intelligence integration in Eswatini libraries. Journal of Web Librarianship 18(3): 149–168.

27.

Huang

Dong

(2024) Research on the application pathways and challenges of AIGC in libraries. In: Proceedings of the 2024 5th International Conference on Education, Knowledge and Information Management (ICEKIM 2024), Chengu, China, 19-21 April 2024, pp. 958–969. Dordrecht: Atlantis Press. Available at: https://doi.org/10.2991/978-94-6463-502-7_103 (accessed 20 July 2025).

28.

Zhang

(2025) Big data analytics in library services with AI: Personalized content recommendations and catalog optimization. IEEE Access 13: 88412–88420.

29.

Hussain

Khan

(2025) Librarian acceptance of artificial intelligence in academic libraries of Islamabad: An application of the technology acceptance model. Technical Services Quarterly 42(3-4): 1–21.

30.

Juego

AJP

Masalinto

(2025) Perception, attitude, and readiness in artificial intelligence adoption among academic librarians in the Bicol Region Librarians Council (BRLC). International Journal of Innovative Science and Research Technology 10(4): 4157–4162.

31.

Kalbande

Yuvaraj

Verma

, et al. (2024) Exploring the integration of artificial intelligence in academic libraries: A study on librarians’ perspectives in India. Open Information Science 8(1): 20240006.

32.

Kaushal

Yadav

(2022) The role of chatbots in academic libraries: An Experience-based perspective. Journal of the Australian Library and Information Association 71(3): 215–232.

33.

Kisilowska-Szurmińska

(2025) Artificial intelligence in academic libraries – a tool, a collaborator, an adversary? A Delphi study of university librarians in Poland. Journal of Academic Librarianship 51(5): 103114.

34.

Coates

(2025) Academic library online chat services under the impact of artificial intelligence. Information Discovery and Delivery 53(2): 192–205.

35.

(2024) Evaluating AI literacy in academic libraries: A survey study with a focus on U.S. employees. College & Research Libraries 85(5): 635.

36.

Hudson Vitale

(2023) Quick Poll Results: ARL Member Representatives on Generative AI In Libraries. ARL Views Blog (Association of Research Libraries) 9 May. Available at: https://www.arl.org/blog/quick-poll-results-arl-member-representatives-on-generative-ai-in-libraries/ (accessed 20 July 2025).

37.

Lulu-Pokubo

Okwu

(2025) Librarians’ awareness towards the use of artificial intelligence technologies for sustainable library services. Business Information Review 42(1): 20–28.

38.

Lund

Omame

Tijani

, et al. (2020) Perceptions toward artificial intelligence among academic library employees and alignment with the diffusion of innovations’ adopter categories. College & Research Libraries 81(5): 865–882.

39.

Mannheimer

Bond

Young

SWH

, et al. (2024) Responsible AI practice in libraries and archives: A review of the literature. Information Technology and Libraries 43(3): 1–29. Available at: https://doi.org/10.5860/ital.v43i3.17245 (accessed 20 July 2025).

40.

Monyela

Tella

(2024) Leveraging artificial intelligence for sustainable knowledge organisation in academic libraries. South African Journal of Libraries and Information Science 90(2): 1–11.

41.

Moulaison-Sandy

Coble

(2024) Leveraging AI in cataloging: What works, and why? Technical Services Quarterly 41(4): 375–383.

42.

Oyighan

Ukubeyinje

David -West

, et al. (2024) The role of AI in transforming metadata management: Insights on challenges, opportunities, and emerging trends. Asian Journal of Information Science and Technology 14(2): 20–26.

43.

Panda

Chakravarty

(2022) Adapting intelligent information services in libraries: A case of smart AI chatbots. Library Hi Tech News 39(1): 12–15.

44.

Pantzar

(2023) It makes things up like crazy’: Swedish school librarians’ perceptions on ChatGPT. PhD dissertation, Lund University, Sweden. Available at: https://lup.lub.lu.se/student-papers/search/publication/9116871 (accessed 20 July 2025).

45.

Patra

Gaitanou

Oladokun

, et al. (2025) Exploring the integration of AI in library services: Perspectives and considerations. Business Information Review 42(2): 133–142.

46.

Rahman

Islam

(2024) The impact of ChatGPT for enhancing knowledge management in university libraries. Journal of Web Librarianship 18(4): 177–196.

47.

Rodriguez

Mune

(2022) Uncoding library chatbots: Deploying a new virtual reference tool at the San Jose State University Library. Reference Services Review 50(3/4): 392–405.

48.

Shahzad

Khan

Iqbal

, et al. (2024) Identifying university librarians’ readiness to adopt artificial intelligence (AI) for innovative learning experiences and smart library services: An empirical investigation. Global Knowledge, Memory and Communication. Available at: https://doi.org/10.1108/GKMC-12-2023-0496 (accessed 20 July 2025).

49.

Skjuve

Brandtzaeg

Følstad

(2024) Why do people use ChatGPT? Exploring user motivations for generative conversational AI. First Monday 29(1). Available at: https://doi.org/10.5210/fm.v29i1.13541 (accessed 20 July 2025).

50.

Subaveerapandiyan

Gozali

(2024) AI in Indian libraries: Prospects and perceptions from library professionals. Open Information Science 8(1): 20220164.

51.

Sussmeier

Henry

(2025) Mind the gap!: How do we ethically bridge the divide between the cataloging/metadata community and the world of AI? Journal of Library Metadata 25(4): 241–259.

52.

Vrindha

Syamili

(2025) Navigating the AI landscape in libraries: A PRISMA-based systematic analysis of AI applications in libraries. Journal of Web Librarianship 19(1): 45–61.

53.

Wolf

Maier

(2024) ChatGPT usage in everyday life: A motivation-theoretic mixed-methods study. International Journal of Information Management 79: 102821.

54.

Xiao

Gao

(2020) Connecting the dots: Reader ratings, bibliographic data, and machine-learning algorithms for monograph selection. Serials Librarian 78(1-4): 117–122.

55.

Yakubu

Kassim

Husin

(2023) Conceptualizing hybrid model for influencing intention to adopt cloud computing in North-Eastern Nigerian academic libraries. Journal of Academic Librarianship 49(4): 102747.

56.

Yang

Ding

Chen

, et al. (2024) Factors affecting the use of artificial intelligence generated content by subject librarians: A qualitative study. Heliyon 10(8): e29584.

57.

Yan

Zhao

Mazumdar

(2023) Chatbots in libraries: A systematic literature review. Education for Information 39(4): 431–449.

58.

Yoon

Andrews

Ward

(2022) Perceptions on adopting artificial intelligence and related technologies in libraries: Public and academic librarians in North America. Library Hi Tech 40(6): 1893–1915.

59.

Zhu

Zhao

(2024) Research on the impact of ChatGPT and other NLP technologies based on GPT models on the construction and development of intelligent libraries in higher education. In: Proceedings of the ACM International Conference on Artificial Intelligence and Future Education (AIFE), Shanghai, China, 1–2 November 2024, pp. 119–125. New York: ACM. Available at: https://doi.org/10.1145/3708394.3708416 (accessed 20 July 2025).