Exploring factors associated with nursing students’ artificial intelligence literacy: insights from a national mixed methods study

Abstract

Background

Integrating artificial intelligence (AI) into healthcare is rapidly expanding, yet research on nursing students’ AI literacy (AIL) remains limited. This study assessed AIL levels, identified factors associated with AIL, and explored students’ perceptions of AI use.

Methods

The mixed-methods study used a non-probability snowball sampling technique, involving 383 undergraduate nursing students in Türkiye, and utilized the Descriptive Characteristics Form and the Artificial Intelligence Literacy Scale (AILS). Quantitative analyses included t-tests, ANOVA, Pearson correlations, and multiple linear regression. Qualitative data underwent thematic analysis.

Results

The mean age of participants was 21.52 (SD ± 3.31; range, 18–40), and 83.3% of them were female. Participants’ mean AILS score was 110.80 (SD ± 40.98; range = 31–216) with 41.22 ± 19.67 for technical understanding, 40.49 ± 15.57 for critical appraisal, and 29.09 ± 10.68 for practical application. Higher comfort with new technologies (β = 0.352; t = 7.06; p < 0.001) and having witnessed/experienced ethical concerns (β = 0.094; t = 1.98; p = 0.048) were significantly associated with higher AIL. Seven themes were identified: (1) Support and Convenience, (2) Time Management, (3) Learning and Academic Development, (4) Ethical and Security Concerns, (5) Misinformation and Trust, (6) Laziness and Dependency, and (7) Creativity and Thinking.

Conclusion

Enhancing AIL skills among nursing students is essential for the ethical and rational use of AI. Educators should guide nursing students to develop strategies to improve their technical, appraisal, and practical AIL skills. The present study may contribute to identifying educational strategies to support nursing students’ readiness for AI in nursing education.

Trial registration

Not applicable.

Introduction

Recent technological advancements and the rapid digitalization of healthcare have facilitated the integration of artificial intelligence (AI) into numerous disciplines, particularly within nursing practice [1]. In clinical settings, AI technologies are used to support patient monitoring, workflow optimization, documentation, and decision support, enhancing the quality of patient care and reducing nurses’ workload [2]. The successful implementation of these AI technologies depends not only on the availability of technological systems but also on healthcare professionals, especially nurses, understanding, critically evaluating, and responsibly engaging with AI-related technologies [3]. Nursing students, who represent the future workforce, are expected to graduate with these competencies in both educational and professional contexts [2–4].

Among these competencies, AI literacy (AIL) has emerged as an important foundational capability. It is broadly defined as the combination of knowledge, recognition, and skills needed to interact meaningfully and critically with AI technologies across various contexts [5]. AIL includes three core dimensions. The first is knowledge, which refers to the capacity to critically assess AI technologies, understand how they are designed, interpret their outputs, and anticipate potential consequences. Secondly, recognition involves identifying suitable applications for AI and acknowledging its inherent strengths and limitations. Thirdly, skills refer to proficiency in designing, implementing, or interacting with AI tools and systems in accordance with ethical, technical, and contextual standards [6]. Taken together, AIL supports individuals in using AI technologies safely, efficiently, and ethically across multiple spheres, including education and healthcare practice [7].

AIL includes understanding core AI concepts, critically appraising AI outputs, recognizing limitations and risks, and using AI in ethically and contextually appropriate ways [2, 4, 8]. For nursing students, such literacy is increasingly relevant as AI is entering clinical environments [9] and generative AI (GenAI) tools have become readily accessible in higher education [10]. However, an important distinction should be made between general AI use in educational contexts and the competencies required for clinical AI applications. While students interact with GenAI tools in academic settings, the safe integration of AI in healthcare requires additional competencies. These competences include critical appraisal of algorithms, understanding clinical workflows, and addressing ethical considerations within clinical practice [11]. Therefore, future nurses must go beyond mere digital confidence and cultivate a proactive, critically reflective stance, rigorously evaluate AI-generated outputs, and identify potential risks and biases [12, 13]. This makes AIL competence a crucial issue for undergraduate nursing students to address, including practical, ethical, and professional questions related to transparency, bias, privacy, accountability, and appropriate human oversight [14].

GenAI tools have become particularly prominent in higher education due to their potential to enhance both teaching and learning outcomes [15–17], and are expected to become even more prevalent in nursing curricula in the near future [18]. These tools function as easily accessible learning aids that support academic tasks and self-directed study [15–17]. Nursing students use them to summarize information, brainstorm ideas, improve writing, translate text, and obtain rapid explanations [18]. However, GenAI introduces important pedagogical and ethical concerns [19]. These include issues related to plagiarism and data privacy [17], academic misconduct, over-reliance on AI leading to diminished critical thinking and clinical decision-making skills, and the unintentional propagation of misinformation [20]. For nursing students preparing for a profession grounded in ethical judgment and evidence-based practice, these concerns make AIL especially relevant [21]. Therefore, GenAI represents not only a convenient educational resource but also a pedagogical phenomenon that underscores the need for structured guidance in responsible AI use [22].

Recent quantitative evidence indicates that the urgency of AIL in nursing education is not only conceptual but increasingly supported by empirical research [23, 24]. In higher education, the use of general AI has expanded rapidly, with over 70% of university students using these tools for academic purposes, but almost half are not very proficient, and only a small proportion of students learn how to use general AI through formal courses or conferences, and structured educational support remains limited [21]. A systematic review and meta-analysis also found that the pooled proportion with relatively good AI knowledge was 0.44, whereas the pooled proportion with positive attitudes toward AI was 0.65, indicating a clear gap between receptiveness and actual literacy [25]. A recent cross-sectional study among nursing students also explored that 92.2% reported positive attitudes toward AI, yet 69.6% demonstrated only moderate readiness, and more than half reported discomfort or negative perceptions toward AI use [26]. Consistent with these findings, a systematic review revealed that nursing students generally demonstrate moderately positive attitudes but only moderate literacy and readiness, with prior AI training and stronger computer skills associated with more favorable outcomes [27]. Furthermore, a recent bibliometric review found that only 2.9% of publications in the broader medical field explicitly addressed AIL [28]. Similarly, few available studies have shown that undergraduate nursing students generally demonstrate moderate proficiency in AIL, with notable discrepancies across different subdomains such as technical understanding, critical appraisal, and practical application [4, 29].

Although research on AIL in nursing education remains limited, existing studies suggest that nursing students’ AIL may vary across demographic, experiential, and attitudinal factors [4, 29]. Factors such as innovative mindset and AIL have been shown to significantly predict career and skill self-efficacy [4]. Furthermore, certain characteristics, such as being male, being in the third year of study, receiving formal AI training, and regularly using AI tools, have been associated with higher AIL levels [29]. Despite the emerging findings, there is still limited evidence on which factors most strongly relate to AIL among nursing students, especially in contexts where students are actively encountering GenAI in academic life [30, 31]. More research is needed to clarify not only students’ AIL levels but also the educational and experiential conditions associated with higher or lower readiness [30–32]. Also, current nursing curricula often place insufficient emphasis on informatics and AI training, raising questions about whether students are adequately prepared to engage with evolving digital health tools [33]. Addressing this gap requires not only enhancing AIL training but also systematically assessing current literacy levels among nursing students and identifying the factors that influence their proficiency. There is a clear need for structured educational interventions [18] on safe, ethical, and evidence-based application of AI tools in clinical decision-making and academic settings [34].

Concerns are growing about the adequacy of nursing students’ preparedness to work with AI systems. However, there are limited studies that comprehensively examine the factors influencing nursing students’ interaction with GenAI tools. In light of these gaps, the present study aims to (1) assess nursing students’ levels of AIL, (2) identify the demographic, experiential, and perceptual factors that influence AIL, and (3) explore students’ experiences and views regarding the use of AI tools, particularly conversational agents such as ChatGPT. This study extends prior research on AIL by providing discipline-specific evidence from nursing education. In addition, using a national sample and the validated AILS, the study offers a systematic assessment of nursing students’ AIL levels and associated factors. In this study, AIL is conceptualized as a foundational, cross-context competence that supports nursing students’ ability to understand, critically appraise, and responsibly use AI technologies. However, because undergraduate nursing students currently encounter AI widely in academic settings through GenAI tools such as ChatGPT, the qualitative component of this study focuses particularly on students’ experiences with these educational uses rather than on direct clinical or diagnostic AI. This study not only goes beyond descriptive assessments of AIL but also addresses a critical, underexplored distinction between use and proficiency. By combining quantitative and qualitative analyses, this study aims to provide a more comprehensive understanding of the current state of AIL among nursing students and explore their lived experiences with GenAI tools and inform educational strategies to enhance AI readiness in nursing education. The findings of this study contribute to developing strategies to improve AIL among nursing students and can inform studies on integrating AI into the nursing curriculum. In this respect, this study facilitates the development of strategies to enhance human-technology interaction and contribute to a more conscious integration of AI in healthcare.

Therefore, the research questions are as follows:

1. What is the current level of AIL among nursing students?

2. Which demographic, experiential, and perceptual factors are significantly associated with nursing students’ AIL?

3. How do nursing students describe their opinions, experiences, and concerns regarding the use of AI tools such as ChatGPT in their educational journey?

4. How do the qualitative findings help explain, enrich, or contextualize the quantitative patterns identified in nursing students’ AIL?

Materials and methods

Aim

The study aimed to assess nursing students’ AIL levels, identify the factors associated with AIL, and explore their experiences and views on the use of AI tools, particularly conversational agents such as ChatGPT.

Design

This study employed a convergent parallel mixed-methods design, utilizing both quantitative and qualitative methods. In this design, qualitative and quantitative data are collected simultaneously and analyzed separately [35]. Then, all data are integrated and interpreted to provide a comprehensive and robust understanding of the research topic. While the quantitative phase included a questionnaire to assess students’ AIL levels and identify the factors associated with AIL, the qualitative part included open-ended questions about AI use. In addition, the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist was used for cross-sectional studies in the quantitative Sect [36]., and the Consolidated Criteria for Reporting Qualitative Research (COREQ) was followed for the qualitative Sect [37].

Study population and sample

The study population consisted of 31,629 undergraduate nursing students [38] enrolled in 184 nursing programs [39] across Türkiye during the 2023–2024 academic year. Due to the lack of a centralized student contact registry and the geographically dispersed population, a snowball sampling technique was used to facilitate recruitment. Using the known population sample size formula with a 95% confidence interval and a 5% margin of error, an expected proportion of d = 0.05, the required sample size was calculated as 380 nursing students. To enhance the methodological rigor of the study, an a priori power analysis (G*Power 3.1.9.7) revealed that 153 participants were sufficient to achieve a power of 0.95 at a moderate effect size (f² = 0.15) and a significance level of 0.05 for the regression model with seven independent variables. This demonstrates that the sample size of 383 participants exceeds the targeted statistical power. The qualitative data were collected through open-ended questions via the same online survey. The qualitative component of this study was designed to complement the quantitative findings by capturing participants’ experiences, perceptions, and contextual interpretations. An exploratory approach using open-ended survey questions was employed to identify recurring themes and enrich the overall interpretation, rather than to develop a formal theory. Accordingly, data saturation was not adopted as the primary criterion for methodological rigor. Instead, the study prioritized capturing a broad range of perspectives from a large and diverse participant group. Emphasis was placed on achieving thematic adequacy to reflect the diversity of viewpoints and facilitate the integration of qualitative insights with quantitative results [40]. Inclusion criteria for both qualitative and quantitative parts of the study were: (1) being a 1st, 2nd, 3rd, and 4th year student enrolled in a nursing bachelor program in Türkiye, (2) agreeing to participate in the study voluntarily, and (3) completing the data collection instrument. Exclusion criteria were: (1) refusing to participate in the study, (2) not being enrolled in a nursing bachelor program, and (3) completing the data collection form incompletely. Ultimately, the study was completed with 383 nursing students who voluntarily participated and completed all data collection instruments. Four hundred incomplete or partially filled forms were excluded from the analysis to maintain data quality. Incomplete responses were defined as forms in which participants did not proceed beyond the initial sections of the survey or failed to complete the items required to calculate the AILS total and subscale scores. Since the primary quantitative outcome was the AILS score, forms with missing AILS data could not be included in the analysis. Additionally, partially completed forms with missing key demographic or AI-use variables were excluded to maintain consistency in descriptive, comparative, and regression analyses. The participants included in the study were from seven different regions of Türkiye; 38.64% from the Marmara Region, 21.67% from the Central Anatolia Region, 10.96% from the Aegean Region, 8.61% from the Mediterranean Region, 7.83% from the Eastern Anatolia Region, 7.31% from the Black Sea Region, and 4.43% from the Southeastern Anatolia Region.

Instruments

The data were collected with the Descriptive Characteristics Form developed by the researchers and the Artificial Intelligence Literacy Scale (AILS) previously adapted to Turkish.

Descriptive characteristics form

The form consisted of 15 questions categorized as follows: Sociodemographic characteristics (4 items), AI usage behavior including status, frequency, and purpose of AI tool usage, any negative experiences associated with AI tool usage (8 items), and Open-ended questions (3 items) addressing students’ views on the contributions, potential risks, and reasons recommending/not recommending its use in the nursing context.

Artificial intelligence literacy scale (AILS)

The AILS was originally developed by Laupichler et al. [41] to assess AI literacy in non-expert populations. The scale was subsequently translated and validated into Turkish by Karaoğlan Yılmaz & Yılmaz [42]. The scale consists of 31 items, grouped into three subscales: Technical Understanding (14 items), Critical Appraisal (10 items), and Practical Application (7 items). Items are rated on a 7-point Likert scale, ranging from 1 (Strongly Disagree) to 7 (Strongly Agree), with higher scores indicating greater proficiency in AIL. No reverse-coded items are present. The total possible score ranges from 31 to 217. The original scale demonstrated excellent internal consistency, with Cronbach’s alpha values of 0.99 for the full scale, 0.98 for both the Technical Understanding and Critical Appraisal subscales, and 0.97 for Practical Application [42]. This study’s overall Cronbach’s alpha was 0.97, indicating high reliability [43].

Data collection

Data were collected between April 1, 2024, and May 20, 2025, following ethical approval. The survey was distributed online using the Koç University Qualtrics platform. The link to the electronic survey was disseminated through social media platforms frequently accessed by nursing students, such as those managed by the Turkish Nurses Association Student Commission and the Student Nurses Association. Additionally, it was emailed via official communication channels of the student association presidents and members. The link was shared on three occasions, spaced 15 days apart, to optimize response rates. Upon accessing the link, participants reviewed the informed consent form and confirmed their willingness to participate by selecting “Yes, I want to participate in the study.” The survey was completed anonymously and took approximately 10–15 min.

Data analysis

Data were analyzed using SPSS (Statistical Package for the Social Sciences) version 27 (IBM Corp., Armonk, NY, USA). The normality of continuous variables was assessed using a combination of descriptive statistics (including skewness, kurtosis, and standard deviation-to-mean ratio), graphical methods (such as Q–Q plots and histograms), and statistical testing via the Kolmogorov–Smirnov test. Categorical variables were summarized using frequencies and percentages (n, %), while continuous variables were reported as means and standard deviations (mean ± SD). The internal consistency reliability of the AILS was evaluated using Cronbach’s alpha coefficient. To examine differences in AIL scores across groups, independent-samples t-tests were used for two-group comparisons, and one-way analysis of variance (ANOVA) was used for comparisons involving more than two groups. The relationship between two continuous variables was analyzed using the Pearson correlation test. Multivariate linear regression analysis was used to examine associations between independent variables and the dependent variable. All statistical tests were two-tailed, and results were considered significant at the p < 0.05 level, with a 95% confidence interval.

Qualitative data obtained from the three open-ended questions were analyzed using a hybrid thematic analysis, approach integrating computational text mining with interpretive qualitative analysis. This semi-automated analytical framework was adopted to enhance methodological rigor, transparency, and analytical consistency while preserving the contextual depth of participants’ experiences. In the initial computational stage, qualitative responses were exported into a structured dataset and processed using Python-based natural language processing (NLP) techniques. NLP libraries, including NLTK and spaCy, were employed for systematic preprocessing procedures such as tokenization, stop-word removal, lemmatization, normalization, frequency analysis, and keyword-in-context clustering. These procedures facilitated the identification of recurring semantic patterns and relationships across the dataset and served as an initial pattern-recognition mechanism to support analytical consistency and reduce subjective bias during early-stage coding [44]. To maintain consistency in identifying common semantic patterns, only terms appearing in at least 5% of responses were considered during the computational screening stage [45].

Following computational preprocessing, qualitative data were analyzed inductively using thematic analysis guided by semantic interpretation principles. The analytical process proceeded iteratively across three interconnected phases: preparation, organization, and reporting. During the preparation phase, researchers familiarized themselves with the dataset and defined the units of analysis. In the organization phase, preliminary computational outputs were reviewed alongside participants’ original statements, followed by open coding, category development, and grouping of conceptually related codes into broader thematic categories. Researchers experienced in qualitative research independently coded the responses and compared emerging categories through iterative discussions until consensus was achieved. Overlapping or recurring categories were refined and reorganized to improve conceptual coherence and thematic consistency. Finally, in the reporting phase, themes and subthemes reflecting participants’ perceptions and experiences regarding AI tools were finalized through collaborative interpretation among the research team.

Importantly, the computational component was used solely to support pattern recognition and organizational structure; final themes were not generated automatically but were developed through researcher interpretation, reflexive discussion, and consensus. For example, responses computationally clustered around expressions such as “easy,” “fast,” and “helpful” initially appeared as a single lexical grouping. However, interpretive review of the original participant narratives demonstrated that these expressions reflected conceptually distinct dimensions, which were subsequently differentiated into the themes of Support and Convenience and Time Management. This iterative process ensured that the final thematic framework reflected participants’ intended meanings rather than relying exclusively on word-frequency patterns.

Rigor and trustworthiness of qualitative data analysis

The rigor and trustworthiness of the qualitative analysis were guided by Lincoln and Guba’s framework, including credibility, dependability, confirmability, and transferability [46]. Given the large number of short open-ended responses, a hybrid computational–interpretive approach was adopted to support the systematic identification of recurring patterns, reduce the influence of individual coder bias during the early stages of analysis, and improve transparency in handling the qualitative dataset. However, the computational stage was used only to assist pattern recognition; the final themes were developed through researcher interpretation and consensus rather than automated generation.

Credibility was enhanced through repeated comparison of emerging themes with the original participant responses and through independent review of coding structures by multiple researchers until consensus was achieved [47–49]. Dependability was supported by systematically documenting all analytical stages, including preprocessing, clustering, coding, category refinement, and thematic development [50]. Confirmability was strengthened by maintaining a data-driven analytical process grounded in participants’ own statements [51], while transferability was supported through detailed reporting of the study context, participant characteristics, and analytical procedures.

To further enhance transparency, reproducibility, and auditability, coding decisions, preprocessing procedures, and thematic refinements were documented and version-controlled throughout the analysis. Qualitative responses were processed using Python-assisted NLP techniques to ensure consistency in text preprocessing and organization. An initial inductive coding framework was developed through manual review by the research team, after which Python-based scripts were used to support systematic coding, keyword frequency analysis, and pattern identification across responses. Emerging codes were iteratively refined and grouped into themes through a collaborative interpretive process. Thus, this hybrid computational-interpretive framework combined the scalability and consistency of computational text analysis with the contextual sensitivity and reflexivity of human interpretation, enabling a rigorous yet flexible exploration of nursing students’ experiences and perceptions regarding AI use.

Ethical consideration

Permission to use the AILS was obtained via e-mail from the corresponding author, who conducted the Turkish validity and reliability of the scale. Ethical approval for the study was granted by the Koç University Human Research Ethics Committee (Decision No: 2024.165.IRB3.072; Date: 8 July 2024). Participants in this study provided informed consent electronically. Anonymity and confidentiality were ensured throughout the research process. Participants were informed of their right to withdraw at any stage without penalty.

Results

Quantitative results

Data from 383 participants were analyzed. The mean age of the participants was 21.52 (SD ± 3.31; range, 18–40) years, and their GPA was 3.02 (SD ± 0.49; range, 1.6-4). The majority were female (83.3%, n = 319), with first-year students comprising the largest group (31.6%). Nearly half of the participants (47%) reported being comfortable with using technology, and 93.2% (n = 357) had previously used AI tools. In addition, the most common purposes for using AI included preparing homework (74.7%), translating text (43.9%), and creating presentations (43.9%). A smaller proportion (13.3%) reported experiencing or witnessing ethical issues while using AI tools (Table 1).

Table 1.

Participants’ descriptive characteristics (N = 383)

Variables		Min-max	X̄±SD
Age		18–40	21.52 ± 3.31
GPA		1.6-4	3.02 ± 0.49
		N	%
Gender	Female	319	83.3
	Male	64	16.7
Class	1st year	121	31.6
	2nd year	90	23.5
	3rd year	89	23.2
	4th year	83	21.7
Comfort level in using new technologies	Very comfortable	75	19.6
	Comfortable	180	47
	Uncertain	105	27.4
	Uncomfortable	23	6
Status of using AI tools	Yes	357	93.2
	No	26	6.8
Frequency of AI use	Never	26	6.8
	Occasionally	147	38.4
	Regularly	141	36.8
	Intensively	69	18
Purpose of AI use	Preparing homework	286	74.7
	Translating	168	43.9
	Learning language	91	23.8
	Creating presentation	168	43.9
	Other	88	23
Experience of ethical problems in the use of AI tools	Yes	51	13.3
	No	332	86.7

SD, standard deviation; X̄, mean, AI, artifical intelligence

Participants’ AILS scores

Participants’ mean AILS total score was 110.80 (SD ± 40.98; range = 31–216) (Table 2).

Table 2.

Participants’ total and subdimension mean scores of artificial intelligence literacy (N = 383)

AILS	Item	X̄±SD	Min.-Max.
Technical understanding	14	41.22 ± 19.67	14–97
Critical appraisal	10	40.49 ± 15.57	10–70
Practical application	7	29.09 ± 10.68	7–49
AILS-Total	31	110.80 ± 40.98	31–216

SD, standard deviation; X̄, mean; AILS, Artificial Intelligence Literacy Scale

Comparison of the mean AILS scores and the participants’ descriptive characteristics

Gender was significantly associated with AIL, with male participants scoring significantly higher than females on the total AILS score (t = 2.348; p = 0.019) and on the Technical Understanding subscale (t = 2.867; p = 0.004). Participants who reported encountering or witnessing ethical problems when using AI had higher AIL scores (t = 2.958, p = 0.003). Notably, participants who reported being “very comfortable” using new technologies scored significantly higher across all AILS subscales (e.g., Total Score F = 26.971, p < 0.001) (Table 3). Similarly, frequency of AI use was positively associated with AIL scores, with the highest scores found among those who reported intensive usage (F = 3.962, p = 0.008).

Table 3.

Comparison of the participants’ artificial intelligence literacy Scale scores by their descriptive characteristics (N = 383)

Variables		Technical Understanding	Critical Appraisal	Practical Application	AILS- Total
	n	X̄±SD	X̄±SD	X̄±SD	X̄±SD
Age (years)	383
r/ P -value		0.011/0.831	-0.030/0.564	-0.037/0.469	-0.016/0.760
Gender
Female	319	39.94 ± 19.66	39.96 ± 15.69	28.72 ± 10.68	108.61 ± 40.97
Male	64	47.59 ± 18.60	43.16 ± 14.79	30.97 ± 10.60	121.72 ± 39.57
t/ P -value		2.867/0.004 **	1.501/0.134	1.543/0.124	2.348/0.019 ***
Class
1st year	121	42.77 ± 19.87	40.44 ± 15.99	29.37 ± 11.33	112.58 ± 42.88
2nd year	90	41.32 ± 18.92	42.43 ± 14.73	29.59 ± 9.79	113.34 ± 39.21
3rd year	89	40.94 ± 20.81	39.98 ± 15.39	28.33 ± 10.93	109.25 ± 41.13
4th year	83	39.15 ± 19.08	39.02 ± 16.11	28.96 ± 10.52	107.13 ± 40.26
F/ P -value		0.562/0.640	0.743/0.527	0.248/0.863	0.454/0.715
GPA	383
r/ P -value		-0.029/0.574	0.095/0.065	0.069/0.177	0.040/0.433
Comfort level in using new technologies
Very comfortable	75	54.73 ± 23.07	50.09 ± 16.13	35.44 ± 10.31	140.27 ± 43.59
Comfortable	180	40.72 ± 17.11	40.62 ± 14.05	29.96 ± 9.14	111.29 ± 34.90
Uncertain	105	35.17 ± 16.41	35.85 ± 13.86	25.16 ± 10.01	96.18 ± 35.65
Uncomfortable	23	28.70 ± 17.61	29.39 ± 16.59	19.57 ± 12.61	77.65 ± 41.54
F/ P -value		21.135/< 0.001 *	18.827/<0.001 *	23.610/< 0.001 *	26.971/<0.001 *
Frequency of AI use
Never	26	35.12 ± 23.76	35.00 ± 18.65	24.31 ± 13.48	94.42 ± 51.49
Occasionally	147	39.35 ± 15.86	38.18 ± 14.67	27.54 ± 9.95	105.08 ± 35.23
Regularly	141	43.14 ± 19.43	42.68 ± 14.09	30.68 ± 9.87	116.50 ± 37.72
Intensively	69	43.58 ± 24.80	43.01 ± 18.01	30.96 ± 11.77	117.55 ± 50.79
F/ P -value		2.069/0.104	3.766/0.011 ***	4.645/0.003 **	3.962/0.008 **
Experience of ethical problems in the use of AI tools
Yes	51	46.57 ± 21.99	46.86 ± 15.22	33.02 ± 10.51	126.45 ± 41.99
No	332	40.39 ± 19.19	39.52 ± 15.42	28.49 ± 10.97	108.40 ± 40.36
t/ P -value		2.095/0.037 ***	3.174/0.002 **	2.846/0.005 **	2.958/0.003 **

*p < 0.001; **p < 0.01; ***p < 0.05; r, Pearson correlation test; t, Independent sample t-test; F, One-way ANOVA test; AILS, Artificial Intelligence Literacy Scores; SD, standard deviation; X̄, mean

Factors associated with AIL of the participants

A multiple linear regression analysis was performed with the variables found significant in the preceding univariate tests to identify factors associated with participants’ AILS scores, using the Enter method. The overall model was statistically significant (F = 18.691, p<0.001) and explained 16.5% of the variance in AIL scores (R²=0.165; Adjusted R²=0.156). The Durbin-Watson statistic (DW = 2.002) demonstrated that residuals were independent, and all tolerance (> 0.89) and variance inflation factor (VIF < 1.13) confirmed the absence of multicollinearity. In addition, heteroscedasticity was assessed using the Breusch-Pagan test (p > 0.05). Two variables emerged as significantly associated factors: self-reported comfort in using new technologies (β = 0.352; t = 7.06; p< 0.001) and prior experience or witnessing of ethical problems related to AI use (β = 0.094; t = 1.98; p=0.048), as shown in Table 4.

Table 4.

Factors associated with artificial intelligence literacy (N = 383)

Variables	B	SE	β	t	P	Partial P	Tolerance	VIF		Model Summary
(Constant)	57.552	8.389		6.861	< 0.001					F = 18.691; p<0.001
Gender (male)	4.787	5.255	0.044	0.911	0.363	0.047	0.963	1.039		DW statistic = 2.002
Comfort level in using new technologies	11.559	1.637	0.352	7.060	< 0.001	0.341	0.891	1.122		R2=0.165
Frequency of AI use	2.343	1.994	0.058	1.175	0.241	0.060	0.923	1.084		Adj. R2 = 0.156
Experience of ethical problems in the use of AI tools	11.361	5.726	0.094	1.984	0.048	0.102	0.978		1.023		Method=Enter

B, Unstandardized coefficients; β, Standardized coefficients; VIF, Variance Inflation Factor; SE, Standard error

DW, Durbin Watson statistic, pr², Correlations square

Qualitative results

Participants’ responses to open-ended questions were grouped under seven main themes: support and convenience, time management, learning and academic development, ethical and security concerns, misinformation and trust, laziness and dependency, and creativity and thinking. These themes reflected students’ perceptions of AI tools such as ChatGPT and their perceived impact on academic learning, ethics, and behavior. These themes summarize participants’ perceptions of AI tools and highlight both facilitating and inhibiting dynamics in their academic and professional development. The thematic constructs cover a spectrum ranging from pragmatic advantages (e.g., support in time management and academic routines) to more complex ethical, psychological, and epistemological concerns. The sub-themes identified further expressed nuanced insights supported by participant narratives, highlighting the various implications of using AI in nursing education contexts. Themes, subthemes, and relevant quotations are presented in Table 5. A visually structured coding tree was developed to present the final themes and their sub-themes, highlighting their hierarchical organization and conceptual relationships (Fig. 1). In addition, a word cloud was generated as a supplementary visualization to provide an overview of frequently occurring terms (Fig. 2); however, the primary analytical insights were derived from the thematic analysis.

Table 5.

Overview of themes and sub-themes identified in participants’ use of GenAI tools

Themes	Sub-Themes	Quotations
Support and Convenience	Ease of Use and Accessibility	“I didn’t need a tutorial. I just opened it and started typing.” “The fact that it’s just like talking to someone makes it feel less stressful.”
	Guidance and Direction	“It tells me where to start. That’s often the hardest part.” “Even when I have writer’s block, asking a question helps me brainstorm.”
	Motivation and Reassurance	“It felt like a 24/7 assistant. Just knowing it’s there reducing anxiety.” “Even if it gives simple answers, it makes me feel like I can manage.”
Time Management	Quick Information Retrieval	“I had a 20-page article. It summarized it in seconds.”
	Streamlining Study Routines	“It saved hours of Googling and reading.”
Learning and Academic Development	Enhanced Understanding	“Sometimes I don’t get it from the slides, but ChatGPT explains it in plain language.” “It helps connect things I didn’t know were related.”
	Language Support and Structuring	“It corrected my grammar and made my sentences more academic.”
	Creative and Critical Thinking Stimulus	“It gave me multiple sides to an issue I hadn’t thought that broadly before.”
Ethical and Security Concerns	Academic Integrity and Plagiarism	“I always worry-am I using it too much? Will my teacher notice?” “It’s tempting to just copy the answer, but I know that’s wrong.”
	Data Privacy and Confidentiality	“I don’t feel comfortable typing in case summaries or patient info.” “What if the system saves everything I write and uses it elsewhere?”
Misinformation and Trust	Verification Challenges	“I caught it making a mistake about a drug dosage.” “Sometimes it gives wrong references. I double-check everything now.”
	Limited Depth in Answers	“It just gives general points. If I want depth, I still need to read books.” “Not very useful for clinical case analysis-too vague.”
Laziness and Dependency	Over-Reliance and Passive Learning	“I started skipping readings and just asked ChatGPT.” “It makes me dependent-I don’t trust my own answers as much.”
	Reduced Motivation to Research	“Before, I used to dig into PubMed. Now, I just ask ChatGPT to summarize.” “It spoils the process of learning. I want fast answers all the time.”
Creativity and Thinking	Idea Generation	“It helped me connect ethical theory with real-life examples.”
	Conceptual Exploration	“The explanations helped me develop a framework for my final project.”

Fig. 1.

Visual representation of themes and subthemes identified through semi-automated and iterative thematic analysis

Fig. 2.

Word cloud of frequently occurring terms derived from nursing students’ qualitative responses

Theme 1: Support and convenience

Participants viewed AI tools such as ChatGPT as accessible academic companions providing non-judgmental, readily available support. In addition, they emphasized the lack of a learning curve, highlighting the intuitive interface of AI tools. This ease was especially valued by students balancing demanding clinical rotations and academic responsibilities, and suggests that simplicity in design may support user engagement, promoting frequent, confident interaction without formal training.

AI was often used as a starting point for academic tasks. When participants experienced difficulty initiating assignments or encountered cognitive blocks, the tool provided structure and initial direction. Participants perceived AI as helping to reduce academic inertia by offering motivational cues and initial frameworks for complex tasks. In addition, participants expressed that AI tools provided a psychological safety net, an immediate source of information that reduced anxiety and enhanced their sense of academic control. These perceptions highlight AI as more than a utilitarian tool; it may function as a cognitive and emotional stabilizer in high-pressure learning environments.

Theme 2: Time management

AI’s potential to enhance time efficiency emerged strongly, especially for students balancing coursework with clinical duties. Participants reported using AI to obtain succinct summaries of complex documents or concepts, thus saving time otherwise spent parsing lengthy texts. This suggests that GenAI tools are positioned by students as accelerators of information access, filling a gap in time-constrained learning contexts. In addition, AI was integrated into daily study habits, helping students prioritize and structure their learning processes.

Theme 3: Learning and academic development

Participants identified AI as an educational facilitator that enhances comprehension and writing skills. They used AI to clarify theoretical content that was otherwise difficult to understand, especially when classroom materials were insufficient. In addition, participants, especially those for whom English is a second language, employed AI for academic writing enhancement. Beyond simplification, students used AI to explore issues from different perspectives, catalyzing critical reflection.

Theme 4: Ethical and security concerns

Ethical deliberations were prevalent in participants’ narratives, focused on integrity, privacy, and professional norms. Participants were cautious about using AI-generated text verbatim, revealing an internal ethical struggle between convenience and authenticity. Participants expressed reluctance to input sensitive academic or clinical data into AI tools. Such concerns echo broader debates on digital sovereignty and ethical stewardship in health education technology.

Theme 5: Misinformation and trust

A recurrent theme was the need for critical engagement with AI outputs due to perceived factual inaccuracies. Participants noted instances where AI provided erroneous or misleading information, particularly in clinical contexts. This suggests a cautious trust dynamic, where students recognize AI’s utility but emphasize the need for verification. Participants observed that AI responses often lacked specificity or depth, necessitating supplemental study.

Theme 6: Laziness and dependency

Participants reflected on the psychological and behavioral consequences of AI use on their study habits. Some participants expressed concern that AI use might encourage intellectual complacency. These views may indicate a perceived risk of cognitive offloading and reduced independent engagement. Participants’ accounts suggest that AI’s speed and convenience may sometimes reduce motivation for traditional information-seeking behaviors.

Theme 7: Creativity and thinking

Despite limitations, participants emphasized the tool’s generative potential in fostering creative exploration. AI was used to brainstorm new perspectives, particularly for assignments requiring interpretive thinking. Participants described how AI helped bridge abstract theories with real-world scenarios, facilitating integrative learning.

Combination of qualitative and quantitative findings

Table 6 presents the integration of quantitative and qualitative findings, highlighting areas of convergence and divergence regarding nursing students’ AIL, patterns of AI use, and the ethical, practical, and interpretive dimensions of their experiences.

Table 6.

Integration of quantitative and qualitative findings on factors associated with AIL of nursing students

Quantitative findings	Qualitative findings	Integrated interpretation
Nursing students demonstrated a moderate overall level of AIL (110.80 ± 40.98), with relatively lower performance in technical understanding (41.22 ± 19.67) than in critical appraisal (40.49 ± 15.57) and practical application (29.09 ± 10.68).	Students described AI mainly as a tool for support, convenience, time management, and academic assistance. They frequently emphasized ease of access, rapid answers, and help with starting tasks, but rarely referred to understanding how AI systems function.	Students’ reliance on AI for efficiency-oriented purposes (e.g., summarizing, generating ideas, saving time) reflects instrumental engagement rather than conceptual understanding. The prominence of “support and convenience” and “time management” illustrates a form of functional literacy, where students use AI effectively for tasks but do not engage with underlying mechanisms (e.g., model behavior, bias, limitations). This convergence suggests that current AIL is shaped more by usage practices than by formal or technical knowledge acquisition.
Comfort with new technologies was the significant associated factor with AIL (β = 0.352, p < 0.001).	Themes such as Ease of Use and Accessibility, Guidance and Direction, and Motivation and Reassurance showed that students perceived AI as an “academic companion” that reduces anxiety, provides direction, and supports task initiation.	The qualitative narratives clarify the mechanism underlying this statistical relationship. Students who are more comfortable with technology describe interacting with AI in a confident, exploratory, and iterative manner, using it not only for answers but also for structuring thinking and overcoming cognitive barriers. The perception of AI as a non-judgmental, always-available assistant appears to lower psychological barriers to engagement. This suggests that technological comfort fosters active experimentation and reflective use, which in turn supports higher AIL. At the same time, a subtle divergence emerges while comfort promotes engagement, it may also increase reliance, indicating that comfort alone does not guarantee critical depth unless guided pedagogically.
Frequency of AI use was associated with AIL in univariate analyses, but it did not remain a significant factor in regression analysis.	Students reported widespread and frequent use of AI for homework, translation, summarization, and presentations. However, they described this use as largely task-oriented, emphasizing speed, convenience, and efficiency. Themes of Laziness and Dependency also emerged, reflecting concerns about overreliance.	The integration reveals a clear divergence between use and literacy. Although AI use is widespread, qualitative findings show that this use is often surface-level and efficiency-driven, rather than reflective or analytical. This helps explain why frequency loses significance in regression: repetitive but shallow use does not translate into deeper AIL competencies. The themes of “laziness and dependency” further nuance this relationship by suggesting that frequent use may even displace effortful learning and critical engagement. Thus, the combined findings indicate that how AI is used matters more than how often it is used.
Students who had encountered or witnessed ethical problems related to AI had significantly higher AIL scores and this remained a significant factor in regression (β = 0.094, p = 0.048).	Themes of Ethical and Security Concerns and Misinformation and Trust highlighted students’ awareness of plagiarism, privacy risks, hallucinated or incorrect information, and the need to verify AI outputs. Some students explicitly described questioning or cross-checking AI responses after encountering inaccuracies or ethical dilemmas.	The two strands strongly converge in demonstrating that ethical exposure is linked to deeper critical engagement. Qualitative accounts show that encountering issues such as misinformation or ethical ambiguity prompts students to adopt more reflective and evaluative practices, including verification and cautious use. This supports the quantitative finding that ethical experience associated with AIL, suggesting that such experiences may function as catalysts for developing critical appraisal skills. At the same time, the relationship may be bidirectional: students with higher AIL may also be more capable of recognizing ethical risks, indicating a reinforcing dynamic between ethical awareness and literacy development.
Male students showed higher AIL scores in univariate analyses, but gender did not remain significant in regression.	The qualitative themes did not reveal clear gender-based differences in how students experienced or interpreted AI use. Narratives across participants reflected similar patterns of using AI for support, efficiency, and academic tasks, regardless of gender.	This partial divergence suggests that while gender differences appeared descriptively, they were not strongly supported when other factors were considered and were not meaningfully reflected in participants’ narratives.
Quantitative findings identified moderate AIL but did not fully capture the nature of students’ concerns about AI use.	Themes such as Laziness and Dependency, Misinformation and Trust, and Ethical and Security Concerns revealed worries about over-reliance, reduced critical thinking, inaccurate outputs, and academic misconduct.	The qualitative findings extend and deepen the quantitative results by revealing that AIL is not only a cognitive construct but also an ethical and reflective competency. While the AILS captures knowledge, appraisal, and application dimensions, the narratives demonstrate that students are simultaneously negotiating trust, responsibility, and appropriate use. This indicates that moderate AIL scores may mask underlying tensions between convenience and ethical awareness. The integration highlights that AI literacy should be conceptualized more broadly to include ethical judgment, trust calibration, and metacognitive awareness, which are not fully captured by quantitative scales alone.
Quantitative results highlighted stronger performance in critical appraisal and practical application than in technical understanding.	Students frequently described checking AI outputs, questioning reliability, and recognizing vague or incorrect responses, and using AI cautiously in sensitive contexts.	The two strands show strong convergence. The qualitative data provide concrete behavioral evidence of the relatively stronger critical appraisal dimension, as students actively describe verifying information and maintaining a cautious stance toward AI outputs. However, a subtle nuance emerges while students demonstrate awareness of limitations and the need for verification, they rarely connect this to an understanding of how AI systems function. This suggests that critical appraisal may be experience-driven rather than grounded in technical knowledge, reinforcing the observed gap between appraisal and technical understanding.

AIL, Artificial Intelligence Literacy

Discussion

This study aimed to investigate nursing students’ levels of AIL, the factors associated with it, and their experiences and perceptions of using AI tools. The study revealed that participants had moderate AIL levels, and that higher comfort with new technologies and reporting exposure to negative or ethically ambiguous AI experiences were associated with higher AIL scores. In addition, qualitative analysis enriched the quantitative findings, illustrating that participants perceived AI not merely as a technological aid but as a multifaceted tool that may support and influence their academic engagement. Beyond these findings, this study also offered a methodological contribution by demonstrating how computational approaches can be systematically integrated with interpretive thematic analysis to enhance rigor, transparency, and reproducibility in AI-assisted qualitative research. Furthermore, this study indicates that widespread use of general AI tools may not necessarily translate into higher literacy, as AIL is shaped by critical and reflective engagement, including exposure to ethical challenges.

The integration of quantitative and qualitative findings offers a more nuanced and comprehensive understanding of nursing students’ AIL. While the quantitative results indicated moderate levels of AIL and identified technological comfort as a significant predictor, the qualitative findings provided deeper insight into how students engage with AI tools within their academic practices. Themes such as support, convenience, and time management highlight the role of AI as a functional academic assistant, whereas themes related to ethical concerns and misinformation reveal an emerging awareness of associated risks. Taken together, these findings suggest that although AI tools are widely adopted, students’ engagement remains predominantly pragmatic and task-oriented, rather than underpinned by deeper technical understanding or critical literacy. Overall, the integration of quantitative and qualitative findings reveals both convergence and divergence. While quantitative results suggest moderate AIL and highlight technological comfort and ethical exposure as key factors, qualitative findings demonstrate that students’ engagement with AI remains predominantly pragmatic, efficiency-driven, and emotionally mediated. Importantly, the data suggest that frequent use alone does not ensure deeper literacy and may even be associated with dependency and reduced critical engagement. These findings highlight that AIL is not merely a function of exposure but is shaped by the quality of interaction, ethical reflection, and the broader pedagogical context in which AI is used.

Most students had already experimented with AI tools, predominantly for preparing homework, translation, and creating presentations. This aligns with the findings from a recent study in New Zealand, where nearly 40% of university students reported minimal to moderate use of ChatGPT for academic tasks, including drafting assignments and writing homework (23.1%), and acquiring content (16.5%) [52]. These convergent results reinforce the view that AI is rapidly becoming a highly ubiquitous learning tool across different cultural settings.

In this study, nursing students demonstrated moderate levels of AIL. This finding is consistent with previous studies among nursing students [4, 29] and suggests that their current knowledge and skills may be insufficient to critically evaluate, adapt, and safely integrate AI tools into clinical practice. This result suggests that educators should guide nursing students and integrate AI tools into the curriculum to improve their AIL [21]. In addition, while the content and total hours of nursing programs in Türkiye align with European standards, the integration of AIL varies across institutions. Differences in institutional infrastructure, faculty expertise, and access to digital and AI resources, as well as English-language resources, may have influenced students’ exposure to AI-related competencies. Furthermore, the evolving nature of the Council of Higher Education (YÖK)’s ethical guidelines [53] and different university policies on AI may further shape the development of AIL. These contextual factors should also be considered when comparing the present study findings with international studies.

In the present study, AILS subdimension scores revealed that technical understanding lags behind critical appraisal and practical application. This pattern is consistent with findings from prior research involving medical students [54], and broader university populations [9, 55], which demonstrate that although AI tools are widely used in educational contexts, students’ understanding of underlying algorithmic principles and core concepts remains limited [55]. Furthermore, these results may indicate conceptual awareness of AI tools among nursing students, while also suggesting gaps in their basic technical competence. One possible explanation for this finding is the lack of formal guidelines and training addressing the use of clinical AI tools and Gen AI in nursing education programs. Therefore, the curriculum needs to provide students with structured exposure to the technical dimension of the use of AI tools such as pipelines, rapid engineering, and model bias testing.

While univariate analyses showed that male students scored higher in AIL, consistent with previous findings [9, 29], regression analysis revealed that gender was not associated with AIL. The reason for this difference may be that regression was not conducted in these similar studies. In addition, Mansoor et al.’s study among students from Asia and Africa supports the notion that gender may not exert a consistent influence on AIL across cultural contexts [7]. Although the R² is low in the regression model, a low R² is widely recognized, especially in cross-sectional analysis [56]. The model accounted for only a modest proportion of the variance in AIL. This level of explanatory power is expected in cross-sectional studies investigating complex educational and behavioral constructs, which are inherently influenced by unmeasured cognitive, contextual, and institutional factors. While this may be considered a limitation, the observed R² value also reflects the multifactorial nature of AIL. These findings highlight the need for future models that incorporate additional dimensions, such as educational structures, formal AI training, and pedagogical context, to achieve greater explanatory depth. Nonetheless, further studies with more balanced gender representation are warranted.

Interestingly, the frequency of AI tool use was not associated with AIL in regression analysis, contrary to studies suggesting a positive correlation between usage and literacy [9, 52]. This suggests that mere exposure to AI tools, particularly for basic or transactional tasks, does not guarantee a deeper understanding.Thus, the frequency of AI use may indicate pragmatic engagement rather than a deeper level of understanding, a pattern further supported by the qualitative findings. These findings suggest that AI was often utilized for efficiency-oriented academic tasks, rather than for reflective learning or the development of technical competencies. To support the development of meaningful literacy, educational programs may need to promote structured, reflective, and purposeful engagement with AI. In contrast, comfort with technology emerged as a factor associated with AIL. This finding aligns with the existing literature, which identifies digital self-efficacy as a key factor in students’ engagement with emerging technologies [57, 58]. In addition, this emphasizes the experiential learning effect, where direct and sustained interaction with GenAI systems may be associated with greater familiarity and evaluative and practical competencies [59], and revealed that these participants were better at using AI tools in a technical sense.

Participants who had personally encountered or observed a negative AI incident had higher AIL, accounting for more than 10% of the total participants. Nevertheless, this result also reveals the necessity of developing students’ AIL not through negative experiences but through education. The higher AIL among students who report encountering ethical concerns related to AI suggests that exposure to ethical challenges may reflect greater interaction with AI tools, which, in turn, leads to higher AIL. From a theoretical perspective, this pattern aligns with Protection Motivation Theory [60], which posits that threat appraisal, such as encountering plagiarism accusations, hallucinated references, or privacy risks, can heighten cognitive awareness and motivate individuals to adopt protective strategies, including more careful and critical use of AI technologies. In this sense, negative experiences may function as important learning moments that encourage students to reflect more critically on AIL. At the same time, the relationship may also operate in the opposite direction: students with higher levels of AIL may be better able to recognize ethical risks. Taken together, these possibilities suggest that ethical awareness and AIL may reinforce each other. From an educational perspective, this finding highlights the importance of proactively integrating ethics-focused discussions and scenario-based learning activities into AI-related curricula. Rather than relying on students to develop ethical awareness solely through problematic experiences, structured educational approaches can help them critically evaluate AI technologies and identify potential risks in a more systematic manner.

Thematic analysis provided additional contextual insights into the cognitive, emotional, and ethical dimensions of students’ interactions with AI. Through thematic analysis, students described GenAI not merely as a utility but as a complex educational presence that simultaneously supports, challenges, and reconfigures their academic experiences. Many described these tools as supporting time efficiency, academic motivation, and conceptual clarity, while others raised critical concerns about overreliance, ethical ambiguity, and the erosion of independent learning practices. The main theme of Support and Convenience frequently emerged alongside other themes such as “Time Management”, “Learning and Academic Development”, and “Ethical and Safety Concerns”, suggesting that students may not experience GenAI in isolation, but rather as part of a dynamic ecosystem that includes cognitive, emotional, and moral dimensions. The relationship between “Support and Convenience” and “Laziness and Dependency” is particularly interesting and may point to an underlying paradox: while students appreciate the immediate help offered by GenAI, they are also aware of its potential to discourage effortful learning and critical engagement. The themes of Laziness, Dependency, Mistrust, and superficial use warrant deeper consideration, as they point to potential unintended cognitive and educational consequences of GenAI use. One useful lens for interpreting these findings is cognitive offloading, the delegation of cognitive work to external tools rather than sustained internal processing. In educational settings, such offloading may enhance short-term efficiency, yet it may also reduce effortful cognitive engagement [61]. In the context of GenAI, concerns have also been raised that overreliance on AI tools may limit students’ critical engagement if not supported by appropriate pedagogical strategies [62]. Relatedly, these findings may also be understood through the lens of automation bias, namely the tendency to over-rely on automated outputs and to question or cross-check them insufficiently [63]. In the context of nursing education, this is particularly concerning because uncritical acceptance of plausible but inaccurate responses may undermine the development of clinical reasoning, evidence appraisal, and professional judgment [64]. From a pedagogical perspective, these risks may suggest that the challenge lies not necessarily in AI use itself, but in the conditions under which it is used. Therefore, educational strategies should move beyond simple ‘permission or prohibition’ debates and instead promote structured, human-centered, and critically reflective use of GenAI. These findings align with the literature documenting both the enabling and inhibiting effects of AI in educational settings [65, 66]. Additionally, a word cloud derived from the combined subthemes further confirmed the salience of emotional and ethical constructs in student experiences by highlighting lexical items such as ‘support’, ‘academic’, ‘motivation’, ‘brainstorming’, and ‘privacy’. The co-occurrence of words such as ‘plagiarism’, ‘comfort’, and ‘validation’ reflects the convergence of trust and concern and suggests that AI literacy needs to encompass more than operational knowledge; it requires a reflective, ethics-based approach.

Students often perceived AI as an academic support mechanism, particularly under time constraints. However, this mode of use may raise concerns about intellectual complacency and reduced critical thinking. Several participants reported copying responses directly from ChatGPT without verifying the source, a behavior previously identified as a potential driver of academic misconduct and plagiarism risks [67]. Ethical concerns also extended to issues of data privacy, with students expressing discomfort in inputting clinical or personal information into AI systems, an issue of growing relevance in discussions of digital trust and algorithmic transparency [68].

The thematic analysis yielded seven main themes and several sub-themes, reflecting a spectrum of experiences from pragmatic benefits (e.g., quick content generation, improved academic language) to complex epistemological and ethical dilemmas (e.g., dependency, misinformation, academic integrity). For example, while many students valued AI for its capacity to provide direction and reduce anxiety, others were concerned about the long-term effects on their autonomy, critical reasoning, and capacity for independent research. This aligns with previous findings emphasizing the trade-offs between convenience and cognitive development [69]. The nuanced understanding of these themes highlights that generative AI in nursing education is not merely a technological tool but a pedagogical phenomenon. As such, curricular efforts to integrate AI must balance its instrumental advantages with an emphasis on ethical conduct, critical reflection, and digital responsibility. Educators should foster not only functional literacy but also meta-cognitive awareness, guiding students to navigate the blurred boundaries between facilitation and dependency, efficiency and epistemic depth.

Strengths and limitations of the study

This study is strengthened by its use of a large, diverse sample and by the integration of both quantitative and qualitative data, which allow for a comprehensive analysis of AIL among nursing students. The use of a validated AIL instrument with high internal consistency and regression modeling to identify factors associated with AIL adds further rigor. Additionally, the inclusion of open-ended responses, analyzed thematically, enriched the findings and increased the study’s internal validity. However, certain limitations must be acknowledged. Data were collected via a non-probability snowball sampling technique, which, while effective for reaching a dispersed population, relied on associations and social media accounts and may have introduced sampling bias by overrepresenting students who are more technologically engaged. Such students may be more likely to experiment with AI tools and may demonstrate higher baseline familiarity with digital technologies. At the same time, students with low digital access and limited use of online platforms may have been underrepresented. Therefore, the findings should be interpreted with caution, particularly regarding their generalizability to broader student populations with more diverse levels of technology access, experiences, and engagement. The inclusion of qualitative data in the study mitigated some limitations of the sampling method, enabling a more in-depth examination of heterogeneous experiences. In addition, the study’s results are based entirely on participants’ responses. A one-year data collection period was necessary to reach a sufficient number of participants from diverse student networks and institutions through voluntary participation and snowball sampling. However, since AI is a rapidly evolving field, temporal changes in the AI landscape during this period may have affected participants’ awareness, exposure, and familiarity with AI tools. Furthermore, excluding numerous missing responses from the analysis may have led to underrepresentation of both low and high levels of digital literacy in the final sample. Although this was necessary because the main outcome measure and key study variables were incomplete, it may have introduced attrition bias. Students who completed the survey may have differed from those who discontinued it in terms of motivation, digital confidence, interest in AI, or familiarity with online survey tools. Therefore, the final sample may overrepresent students who were more willing or able to engage with AI-related questions. Moreover, while this study was limited to the context of nursing education in Türkiye and used a non-probability sampling approach, the findings should be generalized to other settings with caution. Differences in infrastructure, resources, and the level of AI integration across universities’ nursing programs may have affected the findings. However, the presence of qualitative data, the identification of factors associated with AIL through regression analysis, a large national sample, and the use of a validated AIL instrument with excellent internal consistency strengthened the study. It is recommended to study the factors that may affect AIL, which were not addressed in this study. The qualitative design studies, such as focus groups, are also recommended to investigate nursing students’ views in depth. Future research should explore additional predictors of AIL, such as curriculum content, institutional policies, and prior training, and employ longitudinal or experimental designs to assess the impact of targeted educational interventions.

Conclusion

Participants demonstrated moderate levels of AIL, with relatively stronger competencies in critical appraisal and practical application, but more limited technical understanding. Comfort with technology was associated with AIL, highlighting the importance of confidence-building, hands-on learning experiences. Although students who encountered ethical issues with AI exhibited higher literacy levels, these findings suggest that AIL should be developed proactively through structured educational approaches rather than through reactive exposure to problematic experiences. The qualitative findings further revealed that students perceive AI as both a supportive academic tool and a potential source of concern, highlighting benefits such as time efficiency and enhanced conceptual clarity, alongside risks such as overreliance, misinformation, and ethical ambiguity. Taken together, these results indicate that AI functions not merely as an assistive technology but as a transformative element shaping learning processes in nursing education. Beyond these empirical findings, this study also provides a methodological contribution by illustrating how computational approaches can be systematically integrated with interpretive thematic analysis. This hybrid analytical framework enhances rigor, transparency, and reproducibility in AI-assisted qualitative research, offering a scalable and robust approach for future studies in health education contexts.

Considering the increasing integration of AI into healthcare and education, it is essential for educators to actively guide nursing students in developing comprehensive AIL competencies, including technical understanding, critical appraisal, and practical application. The findings highlight the need for structured curricular integration of AI-related content, such as algorithmic foundations, bias awareness, and ethical evaluation, embedded throughout nursing programs rather than treated as supplementary topics. Finally, as AIL in clinical practice represents a critical interface between technological capability and ethical responsibility, strengthening these competencies among nursing students has important implications not only for educational outcomes but also for the quality, safety, and ethical integrity of future patient care.

Recommendation for future research

Future research should prioritize longitudinal and intervention-based designs to evaluate the impact of structured AI education on nursing students’ AIL development over time. Comparative and cross-cultural studies across institutions, countries, or healthcare disciplines may also uncover contextual factors influencing AIL. Further work should examine the role of curriculum design, faculty preparedness, and institutional support in effectively integrating AIL into nursing education. Methodologically, future studies are encouraged to adopt hybrid qualitative–computational approaches to enhance analytical rigor, transparency, and reproducibility. Finally, in-depth qualitative or mixed-methods research is needed to better understand students’ evolving perceptions, ethical reasoning, and emotional responses to AI use in both academic and clinical contexts.

Abbreviations

AI

Artificial intelligence

AIL

Artificial intelligence literacy

AILS

Artificial Intelligence Literacy Scale

DW

Durbin-Watson statistic

GenAI

Generative artificial intelligence

SD

Standard deviation

VIF

Variance inflation factor

Author contributions

Pelin Karaçay: Conceptualization; Data curation; Formal analysis; Methodology; Project Administration; Writing- Original draft preparation; Writing-Review and Editing; Supervision; Completed Revisions Process. Özgen Yaşar: Conceptualization; Data curation; Investigation; Methodology; Writing- Original draft preparation; Writing-Review and Editing; Completed Revisions Process.Polat Göktaş: Conceptualization; Data curation; Formal analysis; Writing- Original draft preparation, Completed Revisions Process. Aycan Küçükkaya: Conceptualization; Data curation; Investigation; Writing- Original draft preparation; Completed Revisions Process. All the authors reviewed the final manuscript.

Funding

No funding was received for conducting this study.

Data availability

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

Declarations

Ethics approval and consent to participate

Ethical approval was obtained by the Koç University Human Research Ethics Committee (Decision No: 2024.165.IRB3.072; Date: 8 July 2024). The authors declared that all procedures performed were in accordance with the ethical standards of the institutional research committee and with the Helsinki Declaration. Participants were provided informed consent electronically and were informed of their right to withdraw at any stage without penalty. Anonymity and confidentiality were ensured.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.