Exploring Cambodian Students’ Attitudes Toward AI Use in Education: The Roles of Perceived Academic Benefits, Accessibility of AI, and School Context

Academic benefits of using AI in education

One of the most discussed aspects of AI in education is its capacity to enhance academic outcomes and learning processes (Hanshaw et al., 2024; Srinivasa et al., 2022; Vieriu & Petrea, 2025; Xu, 2024). Previous research has found that students perceived AI-powered tools as valuable for offering instant access to information and enhancing learning efficiency, which helped students with their assignments and problem-solving tasks (Sol et al., 2025a). Moreover, AI-powered tools, like ChatGPT, have been widely used to enhance scientific writing. This tool’s capability provides unlimited access in research planning, execution, and presentation while expediting the writing process, refining writing styles, and supporting proofreading (Huang & Tan, 2023; Lingard, 2023; Sol & Heng, 2024). Indeed, some AI-enabled tools have the potential to provide virtual tutors, personalized support that assists students in understanding complex concepts and fostering their overall study performance (Baidoo-Anu & Ansah, 2023; Belawati & Prasetyo, 2025). Research has also found that these AI-enabled technologies foster creativity and promote more interactive and engaging learning, especially in language learning (Baidoo-Anu & Ansah, 2023; Ngo, 2023). These intelligent technologies are also utilized to support inquiry-based learning due to their capability for automated support and community building, further enhancing students’ academic achievements (Goel, 2020). These perceived academic benefits are usually a major driver of student interest and initial acceptance of AI-powered tools (Adiguzel et al., 2023; Jo, 2024).

Students’ attitudes and willingness to utilize AI in education

Students’ attitudes toward the use of AI-powered applications and their willingness to use them are considerably influenced by their perceptions of the utility and relevance of AI tools to their learning needs (Abdaljaleel et al., 2024; Almogren et al., 2024; Fošner, 2024). Their attitudes usually reflect their readiness and inclination to use these advanced tools in their learning activities (Chan & Hu, 2023; Kim et al., 2020). Prior studies have demonstrated that a positive attitude is often closely linked to the perceived usefulness and applicability of AI-powered applications in achieving academic goals (Kim et al., 2020). A study by Pande et al. (2023) has shown that high school students generally presented a high and positive attitude toward AI technologies; they appreciated the essential role of these technologies to enhance learning experiences and tailor to varying levels of intelligence. The study, nonetheless, noted that practical applications and awareness to optimize the use of AI tools in educational settings remain critical (Pande et al., 2023). Some studies have found that younger people, aged 26 to 34, were likely to show the most openness toward utilizing AI-powered technologies, and their attitudes were significantly influenced by social group perceptions (Maier et al., 2025). Negative attitudes, nevertheless, may stem from difficulties in use, limited perceived value, or concerns related to the consequences of the use of AI technologies, such as privacy, which potentially serve as barriers to AI use in academic activities (Jo, 2024).

Risks and concerns associated with AI use in education

Despite many positive impacts, students were also likely to demonstrate significant concerns regarding the potential risks of using AI in education (Arslan et al., 2025; Baidoo-Anu & Ansah, 2023; Hung & Chen, 2023). Critical concerns are related to the breaches of academic honesty due to misuse or improper use of AI technologies (Sok & Heng, 2024a). Moreover, students have raised their concerns regarding the fear of overdependence or misuse of AI technologies, which results in reduced higher-order thinking skills (i.e., critical thinking, problem-solving, and creativity skills) (Arslan et al., 2025; Pum, 2025; Sok & Heng, 2023). The rise of AI-enabled tools has also amplified the risk of academic plagiarism, blurring the lines between original students’ works and AI-generated outcomes (Adiguzel et al., 2023; Bozkurt et al., 2024; Pum & Sok, 2024). Studies have also shown that students unconsciously violate academic principles, a concern particularly relevant in K-12 settings where clear guidelines tend to be lacking (Akgun & Greenhow, 2022; Marrone et al., 2024). Furthermore, the collection and use of student data by AI systems have raised growing worries about privacy and security, especially among adolescents (Holmes & Miao, 2023; Owoc et al., 2021; Özer, 2024).

Trust and distrust in AI use in education

Students’ willingness to utilize AI-powered tools and their concerns about the risks associated with the use of these technologies in education are both important aspects of their trust or distrust in using AI. Previous studies highlighted that trust captures students’ confidence in the reliability, fairness, and accuracy of using AI technologies (Kozak & Fel, 2024; Marrone et al., 2024). It also noted that distrust can stem from concerns about misleading information produced by AI or worries about biases embedded in algorithms that have the potential to negatively influence either content delivery or fairness in learning assessments (Adiguzel et al., 2023; Sol et al., 2025a). It has been commonly warned that unchecked biases can undermine trust, which is a critical factor in educational contexts (Lockey et al., 2021). Moreover, trust in AI is of vital importance for its successful utilization; however, many challenges, such as explainability, accuracy, transparency, reliability, and large-scale data extraction, considerably influence students’ trust in AI technologies (Lockey et al., 2021). In this sense, it is imperative to develop students’ AI literacy to build informed trust and assist them in critically evaluating AI outputs (Özer, 2024). While previous research highlighted that perceived risks and trust in AI can influence technology adoption (Choung et al., 2023; Gefen et al., 2003), these factors are often secondary to perceived usefulness and access in shaping attitudes toward adoption of AI tools (Chiu & Chai, 2020; Ladeira et al., 2025; Venkatesh et al., 2003; Zhang et al., 2025).

Accessibility and usability of AI in education

Accessibility and usability of AI-powered tools are crucial determinants of adopting AI in education. This includes various factors such as availability of tools and devices, ease of use, and reliable internet connectivity (Almaiah et al., 2022). As prior studies have shown, AI technologies perceived as convenient and available are more likely to be adopted and integrated into learning processes (Choung et al., 2023; Sok et al., 2025). In addition, students’ motivation and engagement to use varying AI systems are directly influenced by the user-friendliness of AI interfaces and their wide accessibility (Pyae et al., 2023). However, there have been great disparities in technological access, with an emphasis on developing nations, raising major challenges to utilize AI technologies equitably (Baidoo-Anu et al., 2024; Goel, 2020).

In the Cambodian context, these challenges are pronounced due to the country’s status as a developing nation (World Bank, 2024) with over 70% of the population living in rural areas and limited infrastructure (Kepios, 2025). The digital divide is a critical barrier, with rural areas experiencing unstable internet connections (Loma Technology, 2025; Sok & Bennett, 2025). This aligns with global concerns highlighted in recent reviews by Mariyono and Alif Hd (2025), who stated that the digital divide and access to technologies remain critical concerns in rural areas, and therefore, the adoption of AI in education should be focused on equity and accessibility. While AI-driven adaptive systems can adapt content to diverse student needs, the infrastructural limitations and ethical concerns like algorithmic bias prevent these tools from fully adapting (Mariyono & Alif Hd, 2025; Pum & Sok, 2024). Rural Cambodian schools often lack sufficient devices and trained educators (Kepios, 2025; Loma Technology, 2025). Similarly, López Costa’s (2025) study in rural Catalan schools found that due to the digital divide and ethical concerns such as plagiarism and reduced critical thinking, practical implementation remains limited, even if teachers reported moderate to high AI knowledge.

The potential of AI as personalized learning platforms can accommodate diverse learning styles and paces of the learner in their learning (Bozkurt et al., 2024; Mariyono & Alif Hd, 2025; Pum & Sok, 2024). However, it is essential to address the divide barriers and unstable internet connectivity issues to integrate the AI tools effectively and inclusively. To promote equity, López Costa (2025) has recommended some strategies, such as optimizing resource allocation and enhancing training mechanisms. In this regard, these strategies could be adapted in Cambodia through partnerships with universities or government initiatives like the Cambodian Education Strategic Plan (Ministry of Education, Youth and Sport, 2024). Moreover, the usability of AI interfaces plays a significant role in shaping students’ attitudes. User-friendly designs can enhance engagement, yet in rural Cambodian schools, where exposure to technology is limited, this requires not only accessible tools but also culturally relevant training programs. In this sense, AI should be seen as a support tool for teaching and learning rather than replacing the teachers (Mariyono & Alif Hd, 2025), implying that Cambodian educators need support to leverage AI effectively.

The influence of contextual factors on AI use in education

Previous research has found that students’ perceptions were significantly shaped by the learning context (Darwin et al., 2023; Han et al., 2025; Monib et al., 2025). Research has found that socioeconomic conditions, geographical locations, and school environment play a significant role in influencing how students use AI tools for their education (Kim & Lee, 2024; Pellas, 2023; Santos et al., 2024). In addition, regarding the digital divide, accessibility, reliable infrastructure, and stable internet connection were also consistently found to be critical barriers for equitable and accessible AI use (Ahmed, 2024; Baidoo-Anu et al., 2024; Ngo, 2023; Kim et al., 2011; Raihan et al., 2024), which directly impacted students’ ability to leverage AI for learning purposes, advocating for AI training programs and infrastructural improvements to ensure inclusive AI adoption in education (Sol et al., 2025a; Özer, 2024).

School locations can also serve as a critical contextual factor that shapes students’ attitudes toward the use of AI-enabled tools in academic activities. As a recent study has shown, students in urban schools, where technological resources and AI-integrated curricula are usually more prevalent, tend to present more positive attitudes toward the use of AI technologies compared to those in remote schools (Sok et al., 2025). This discrepancy significantly stems from the varying levels of exposure to these innovative technologies, the availability of teacher training in AI-supported pedagogy, and institutional support for educational technology (Baidoo-Anu et al., 2024; Ngo, 2023). Studies in the Global South have confirmed that remote teachers and students frequently encounter greater challenges in accessing these advanced technologies, thus resulting in lowering their familiarity and confidence in AI use (Kim & Wargo, 2025; Li, 2025). The shortage of sufficient preparedness and infrastructure among teachers in rural settings further exacerbates these critical challenges, limiting students’ opportunities to benefit from AI-driven learning (Kamaghe, 2026; Ngo, 2023). To bridge this technological gap, there is a need for policy initiatives to provide equitable access to AI-based education in all geographical locations (Sol et al., 2025a).

Risks and trust are typically acknowledged as key drivers influencing students’ adoption of AI technologies; however, the primary focus of this study is on the three aforementioned predictors. Drawing on prior studies, it is expected that higher perceived academic benefits, greater accessibility, and more advantaged school locations will be positively associated with students’ attitudes toward AI in education, providing a clear theoretical basis for the study’s research hypotheses.

Research design

This quantitative research utilizes a survey design to investigate the relationship between students’ attitudes toward AI use and perceived academic benefits, accessibility of AI, and school location. The design emphasizes objectivity and replicability, with clearly defined variables and standardized research instruments to ensure reliability and validity of the research (Harwell, 2011; Lim, 2024). This design enables a rigorous examination of causal relationships between variables, contributing to a deeper understanding of the phenomenon under study.

Sample and sampling strategies

This study utilized convenient and simple random sampling strategies (Emerson, 2015; Stratton, 2021). First, the convenience sampling was used to select schools and classes in which the target students were enrolled. Second, the simple random sampling was utilized to select the students for the study. Based on these sampling strategies, a total of 315 students completed the online survey. These students were enrolled in high schools in the capital city and provinces in Cambodia. After checking and cleaning the data, nine responses were excluded from the analysis because of their incompleteness and the presence of an unstable or inconsistent pattern. Thus, 306 responses were included in the analysis in this study. Among these participants, 62.1% were female, and 67% aged from 16 to 18 years. Almost two-thirds of them (61%) studied in Grade 10, while the rest studied in Grade 11 (21.56%) and Grade 12 (16.66%). In terms of school locations, roughly three-quarters of the students (73.52%) were from the capital city (Phnom Penh), while only 26.47% were from various provinces across Cambodia. A vast majority of the respondents (82.50%) were from families with middle-income socioeconomic statuses, whereas only 0.60% of them were of very high socioeconomic statuses. A detailed summary of the demographic information of the respondent was previously reported in (Sok et al., 2025).

Research instrument

The primary research instrument utilized in this study was an online questionnaire. It was structured into two main components: the first section collects participants’ demographic information, while the second section is designed to assess students’ perceptions of AI use in education. The first section gathered essential demographic details such as gender, age, grade level, English proficiency, the type and location of their schools, and their family socioeconomic status. The second section, comprising 32 items, measured students’ perceptions of AI use in education, specifically regarding the attitudes toward AI, accessibility of AI, perceived academic benefits of AI, risks and concerns of AI, and distrust in AI. Responses were recorded on a 5-point Likert-type scale, ranging from 1 (Strongly disagree) to 5 (Strongly agree). The development of this questionnaire was initially led by the third author, drawing upon insights from relevant existing literature (Baidoo-Anu et al., 2024; Ngo, 2023; Pum & Sok, 2024). Subsequently, the co-authors reviewed and refined the instrument. This collaborative process aimed to ensure its validity, relevance, clarity of language for both Khmer and English versions, and appropriate alignment with the study’s central research question. The authors’ extensive backgrounds in teaching provided valuable expertise in crafting a questionnaire suitable for the participants’ knowledge and understanding level.

The primary aim of the questionnaire was to examine the participants’ perceptions regarding their attitudes toward AI, which consisted of seven items such as ‘I think that using AI tools, like ChatGPT, is a wise choice in the future.’ The accessibility of AI had four items such as ‘I think that the use of AI tools, like ChatGPT, is typical among my classmates and school friends.’ For the academic benefits of AI, there were 10 items, for example, ‘AI tools help to simplify by explaining in detail about difficult terms, concepts, and topics.’ For risks and concerns of AI, there were seven items such as ‘I have fear of reducing my critical thinking and creativity skills.’ Finally, there were three items concerning distrust in AI, for example, ‘I receive falsifying or incorrect responses produced by AI tools, like ChatGPT (Sok et al., 2025).

Data collection procedures

Data for this study were collected using an online survey administered through Google Forms. Before the actual data collection began, a pilot test was conducted with 10 upper-secondary school students. Based on the results of the pilot test, a few questionnaire items were revised to match the participants’ language abilities. The revised questionnaire was then administered via Google Forms. The survey link was disseminated via various social networking sites, commonly used in Cambodia, including Telegram (groups and channels), Messenger, and Facebook pages. To facilitate data collection, the researchers solicited assistance from contacts working in upper-secondary schools, particularly teachers, to further distribute the survey link to potential student participants. The target upper-secondary school students voluntarily and anonymously responded to the questionnaire. Ethical data collection was prioritized in this study. The cover page of the questionnaire clearly stated that participation was entirely voluntary, with the right to withdraw at any time without the need to provide a reason. By submitting their responses, the participants consented to the use of their responses for research and publishing purposes. The average time required to complete the survey was 15 minutes. In addition, to ensure ethical standards, approval to conduct the study was sought and granted by the schools involved in this research. Approval from classroom teachers was also sought and received prior to the administration of the questionnaire.

Data analysis

Four types of data analysis were involved in this study: (a) confirmatory factor analysis (CFA) to examine the construct validity of the adapted and developed scales in the sample under study; (b) scales’ reliabilities, using Cronbach’s alpha coefficient; (c) bivariate correlations between study variables; and (d) hierarchical multiple regression to examine how demographic factors (i.e., gender, grade level, age, level of English proficiency, type of institution, school location, and socioeconomic background), perceived benefits of AI (i.e., academic benefits of AI and accessibility of AI), and challenges regarding AI use (i.e., risk and concern, distrust), are associated with students’ attitudes toward AI use in education. Construct validity was evaluated through convergent and discriminant validity. Convergent validity is acceptable when the factor loading exceeds 0.50; construct reliability (CR) is 0.70 or higher; and average variance extracted (AVE) is 0.50 or higher (Hair et al., 2019). Discriminant validity is ensured when the inter-construct correlations are smaller than the corresponding square root values of AVE (Fornell & Larcker, 1981).

CFA was estimated by maximum likelihood robust (MLR), which is robust to handle non-normal data (Lei & Wu, 2012). Model fit was evaluated based on the following indices: the chi-square (χ²)/degree of freedom (df) (acceptable when not exceeding 3), the Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) (acceptable when exceeding 0.92), the Standardized Root Mean Residual (SRMR) (acceptable when not exceeding 0.08, ideally with CFI above 0.92), and the Root Mean Squared Error of Approximation (RMSEA) (acceptable when not exceeding 0.07, ideally with CFI of 0.92 or higher) (Hair et al., 2019; Hu & Bentler, 1999; Kline, 2016). For the estimation of CFA, Jeffreys’s Amazing Statistics Program (JASP) version 0.19.3 was used, while multiple regression analysis was conducted using SPSS version 27.

Confirmatory factor analysis

Table 1 presents the statistical values of the whole measurement model. The standardized factor loadings ranged between 0.52 and 0.85. The CRs for the academic benefits of AI and accessibility of AI were 0.93 and 0.74, with the AVEs of 0.55 and 0.42, respectively; the CRs for risks and concerns of AI, distrust in AI were 0.88 and 0.75, with the AVEs of 0.51 and 0.50, respectively; the CRs for attitudes toward AI was 0.87, with the AVEs of 0.50. These statistical values indicated acceptable convergent validity (Hair et al., 2019). Moreover, the square root of the AVE of each latent construct was greater than 0.50 and was also higher than the correlation between that latent construct and other latent constructs in the analysis. These statistical values demonstrated acceptable discriminant validity (Fornell & Larcker, 1981; Hair et al., 2019). The CFA results also revealed that the whole measurement model fits the empirical data well, χ2 (395) = 767.193, p < 0.001, CFI = 0.97, TLI = 0.97, SRMR = 0.051, and RMSEA = 0.056 (90% CI = 0.050–0.061), indicating that all the adapted scales could match the Cambodian upper secondary school context.

Descriptive statistics and bivariate correlations

Table 2 presents means, standard deviations, skewness, kurtosis, and bivariate correlations. The correlation analysis revealed that almost all variables had small to strong correlations (from 0.13 to 0.68), indicating a lack of concern regarding multicollinearity problems (Kline, 2016). Attitudes toward AI, accessibility of AI, and academic benefits of AI were positively correlated with one another, while risks and concerns of AI and distrust in AI were strongly positively correlated with each other. The positive constructs showed weak or no significant correlations with the negative constructs.

Multiple regression

Table 3 shows the results of hierarchical multiple regression analysis. A hierarchical multiple regression analysis was conducted to investigate the extent to which academic benefits of AI, the accessibility of AI, and school location predict students’ attitudes toward AI use in education. Three models were tested sequentially.

In Model 1, the academic benefits of AI were entered as a sole predictor. The model was statistically significant, F(1, 304) = 266.67, p < 0.001, and accounted for approximately 46.70% of the variance in students’ attitudes toward AI use in education, R² = 0.467. Academic benefits of AI were a significant positive predictor (β = 0.68, p < .001), indicating that students who recognized AI benefits also tended to report stronger positive attitudes toward AI use in education.

In Model 2, accessibility of AI was added to the regression. The model explained a significantly greater proportion of the variance, R² = 0.564, F(2, 303) = 195.88, p < 0.001. Both academic benefits of AI (β = 0.52, p < .001) and accessibility of AI (β = 0.35, p < .001) were significant predictors, suggesting that students who perceived AI benefits and had access to AI tools also reported more positive attitudes toward AI use.

In Model 3, a school location was added. The results from Model 3 indicated that roughly 57.1% of the variability in students’ attitudes toward AI use in education can be explained by the combined predictors, which included academic benefits of AI, accessibility of AI, and school location, R² = 0.571, F (3, 302) = 134.05, p < 0.001. In this model, all three predictors demonstrated significant positive effects on students’ attitudes toward AI use in education. Specifically, the academic benefits of AI had the strongest influence (β = 0.51, p < 0.001), followed by accessibility of AI (β = 0.36, p < 0.001). Notably, while school location is statistically significant, the effect size is small (β = 0.09, p = 0.03). This suggested that students’ attitudes toward AI use in education were significantly and positively influenced by the perceived academic benefits of AI and accessibility of AI, while school locations tended to play a minor yet notable role.