LLM Chatbots in Academic Writing: Usability and Ethical Concerns Among Bangladeshi University Students

Wahid bin Ahsan, Fariha Islam
Department of Human-Centered Design
Userhub

Abstract

The rise of large language model (LLM) chatbots, such as ChatGPT and Gemini, is transforming academic writing support in resource-constrained higher education systems like Bangladesh’s. This qualitative study explores how 12 Bangladeshi university students experienced and evaluated the usability and ethical implications of these tools during a structured academic writing task. Using moderated one-on-one sessions, think-aloud protocols, and Bangla-language interviews, the analysis identified ChatGPT’s intuitive interface, mobile accessibility, and time efficiency as key strengths. However, participants faced challenges with prompt formulation, repetitive outputs, and Gemini’s less intuitive, search-oriented design. While both tools aided ideation and drafting, concerns emerged about fabricated citations, over-reliance, and diminished critical thinking, raising academic integrity issues. These findings highlight the need for context-sensitive AI literacy, clear institutional policies, and user-centered, transparent HCI design. The study offers a comparative perspective from a Global South context, characterized by limited digital resources and evolving AI adoption, contributing actionable insights for policy, pedagogy, and design.

Keywords: LLM Chatbots, ChatGPT, Gemini, Usability, Academic Writing, Ethical Concerns, AI Literacy, Technology Acceptance, Human-Computer Interaction, Bangladesh, Higher Education, Qualitative Study

Introduction

The transformative potential of large language model (LLM) chatbots like ChatGPT and Gemini is reshaping academic writing, particularly in higher education systems with limited digital tools, faculty support, and writing resources, such as Bangladesh’s. Here, students face substantial barriers, including inadequate linguistic support, underdeveloped curricula, and unequal access to digital learning resources (Alam et al., 2023; Muniruzzaman & Afrin, 2024; Rahman et al., 2024). Prior work has shown that Bangladeshi students often adopt AI tools like ChatGPT for accessibility and convenience, though concerns persist around ethical uncertainty and over-reliance (Talukder & bin Ahsan, 2025). Unlike broader AI engagement explored in virtual learning contexts (Talukder & bin Ahsan, 2025), this study focuses on task-specific behaviors in academic writing, examining how students interact with LLMs during structured tasks. These tools have also been found to enhance writing processes such as ideation and structuring in global academic contexts, acting as scalable, interactive alternatives to traditional feedback (Baber et al., 2024; Klimova & de Campos, 2024). Yet, their growing adoption raises pressing concerns related to usability, cognitive erosion, and academic ethics—especially in settings lacking institutional AI literacy frameworks (Kayalı et al., 2023; Naidu & Sevnarayan, 2023).

Grounded in the Technology Acceptance Model (Davis, 1989), writing process theory (Flower & Hayes, 1981), and ethical human-computer interaction (HCI) principles (Friedman & Hendry, 2019), this study investigates how Bangladeshi university students perceive and experience the usability of ChatGPT and Gemini during a structured academic writing task, framing students’ interactions with LLMs and grounding usability and adoption in theoretical contexts. The central research question asks: How do university students experience and perceive the usability of LLM chatbots in supporting academic writing tasks? To explore this, four sub-questions examine usability factors (SQ1), writing processes (SQ2), adoption influences (SQ3), and ethical concerns (SQ4).

Using a qualitative, task-based design with moderated 1:1 sessions and Bangla-language interviews, this study contributes to educational HCI research by: (1) comparing ChatGPT and Gemini’s usability to inform adaptive design; (2) situating findings within Bangladesh’s resource-constrained academic context; and (3) proposing actionable, HCI-informed recommendations to address usability and ethical challenges in LLM adoption.

Literature Review

This section synthesizes recent literature (2022–2025) on large language model (LLM) chatbots’ role in academic writing, framed by the Technology Acceptance Model (TAM), ethical human-computer interaction (HCI) principles, and writing process theory. It highlights usability, writing support, adoption, and ethical challenges, emphasizing the need for context-sensitive research in under-resourced settings like Bangladesh.

Usability Factors in LLM Chatbots

Usability—encompassing ease of navigation, cognitive load, and responsiveness—shapes student engagement with LLMs. Shaikh et al. (2023) highlight ChatGPT’s intuitive interface and effectiveness in grammar-related tasks. Nguyen et al. (2024) emphasize iterative interaction, showing that repeated exchanges promote reflective learning and improve writing outcomes. However, challenges persist. Kayalı et al. (2023) report difficulties in prompt formulation—the ability to craft clear, effective inputs for relevant outputs—which increases cognitive strain, particularly for novice users. These findings underscore the importance of prompt literacy and adaptive design in diverse educational contexts.

In Bangladesh, Talukder & bin Ahsan (2025) highlight AI’s impact on student engagement in virtual learning, noting usability challenges like irrelevant outputs and calls for intuitive interfaces, with preferences for low cognitive barriers and mobile accessibility inferred from global usability studies (Kayalı et al., 2023; Shaikh et al., 2023). Yet, infrastructural constraints and varying digital literacy levels moderate usability, highlighting the need for context-specific design.

Academic Writing Support Outcomes

LLMs support ideation, outlining, and revision but introduce pedagogical challenges. Klimova & de Campos (2024) demonstrate ChatGPT’s aid in comprehension and task completion under time pressure. Conversely, Aljuaid (2024) cautions that LLMs may improve grammatical accuracy without enhancing critical thinking. Nguyen et al. (2024) note that deeper engagement with LLMs fosters better learning outcomes, requiring active student input.

In Bangladesh, students face linguistic difficulties, insufficient instructional scaffolding, and limited writing resources (Alam et al., 2023; Muniruzzaman & Afrin, 2024; Rahman et al., 2024). While Talukder & bin Ahsan (2025) highlight AI’s motivational benefits in virtual learning, this study examines LLMs’ role in structured writing tasks, emphasizing the need for context-specific scaffolding.

Student Adoption and Attitudinal Factors

TAM underscores perceived usefulness, ease of use, and social influence as adoption drivers. Sallam et al. (2024) and Abdaljaleel et al. (2024) link efficiency and peer influence to increased adoption, though output inaccuracies temper enthusiasm. Koivisto (2023) highlights prior familiarity’s role in shaping perceptions, suggesting adoption depends on tailored support in under-resourced settings.

In Bangladesh, Talukder & bin Ahsan (2025) note students adopt AI through peer networks, driven by curiosity and personalized support needs. Affordability and mobile readiness further shape acceptance in contexts lacking institutional support.

Risks, Ethics, and Institutional Oversight

Ethical concerns include hallucinated outputs and skill erosion. Kayalı et al. (2023) note these issues undermine reliability and student development. Privacy risks and unclear institutional policies exacerbate challenges (Naidu & Sevnarayan, 2023; Ullah et al., 2024). In Bangladesh, Talukder & bin Ahsan (2025) report student anxiety over plagiarism and misconduct due to guideline ambiguity, emphasizing the need for transparent frameworks.

Research Gaps

Few studies compare multiple LLM platforms in task-based educational contexts (Nguyen et al., 2024). The long-term cognitive effects of LLM use, such as impacts on critical thinking, remain underexplored (Kumar et al., 2024). While Talukder & bin Ahsan (2025) examine AI engagement in Bangladesh’s digital learning, little is known about task-specific LLM interactions in academic writing. This study addresses these gaps by comparing ChatGPT and Gemini in Bangladesh’s resource-constrained context, offering practice-oriented insights for HCI and education.

Methodology

Research Design

This study employed a qualitative, task-based usability approach to explore how Bangladeshi university students experience the usability and ethical dimensions of large language model (LLM) chatbots in academic writing. Grounded in an interpretive research paradigm, the design prioritizes depth over generalizability, aligning with best practices in usability and educational HCI research. By combining real-time interaction tasks with reflective interviews, the study captured procedural and experiential insights in a resource-constrained context with evolving AI literacy.

Participants

Twelve students were purposively selected to reflect diversity in academic discipline, gender, and prior LLM use. Selection criteria included enrollment as a second-year undergraduate or postgraduate student, recent academic writing experience, and prior exposure to LLMs. Participants’ fields included business, computer science, economics, environmental studies, and microbiology. Two were international Bangladeshi students based in Canada and the USA, adding cross-cultural perspectives on usability. This sample size aligns with recommendations for thematic saturation in usability studies (Guest et al., 2006). Table 1 summarizes participant demographics; some mentioned using DeepSeek or Copilot, but this study focused solely on ChatGPT and Gemini.

Table 1: Demographic Characteristics and Prior LLM Usage of University Student Participants (n=12)

Participant	Age Range	Gender	Discipline	LLMs Used
P01	25–30	Male	MBA	ChatGPT, DeepSeek
P02	20–25	Male	BBA	ChatGPT, Gemini
P03	35–40	Female	MBA (USA)	ChatGPT, Gemini
P04	20–25	Female	BBA	ChatGPT, Gemini
P05	25–30	Male	Computer Science & Engineering	Gemini, ChatGPT
P06	20–25	Female	Economics	DeepSeek, ChatGPT
P07	30–35	Female	MBA (Canada)	ChatGPT, Copilot
P08	20–25	Female	Accounting	ChatGPT, Gemini
P09	20–25	Female	Human Resource Management	ChatGPT, DeepSeek
P10	20–25	Male	Environment & Development Studies	ChatGPT, Gemini
P11	20–25	Female	Microbiology	ChatGPT, Gemini
P12	20–25	Male	Accounting	ChatGPT, Gemini

Note. LLM = Large Language Model. P03 and P07 were international Bangladeshi participants based in the USA and Canada, respectively. DeepSeek and Copilot were mentioned by participants but not evaluated in this study, which focused on ChatGPT and Gemini.

Procedure

Data collection involved 45–60-minute sessions, conducted in person or via secure video conferencing. After informed consent and a screening questionnaire, participants completed a writing task using ChatGPT and Gemini in randomized order. The task required generating an outline for a short essay on “The role of AI in education.” Think-aloud protocols and screen sharing documented interaction patterns and usability behaviors, followed by semi-structured interviews in Bangla to capture reflections on tool performance, writing processes, and ethical considerations.

Data Analysis

Interview data were thematically analyzed using a bilingual coding approach. Two native Bangla-speaking researchers independently coded Bangla transcripts using English thematic categories developed through iterative review, ensuring linguistic fidelity and cultural sensitivity. Codes were clustered around four analytic domains—usability, writing support, adoption dynamics, and ethical concerns—reflecting the study’s research questions and conceptual framework.

Ethical Considerations

The study adhered to ethical guidelines per the (American Psychological Association, 2017). Participants received detailed study information, and participation was voluntary with the option to withdraw at any time. Anonymity was maintained through coded identifiers, and data from screen recordings and transcripts were stored on encrypted drives accessible only to the research team.

Findings

Thematic analysis of 12 Bangladeshi university students’ experiences revealed context-specific insights into LLM chatbot usability. These findings highlight tool affordances and constraints within a resource-constrained educational context, reflecting diverse engagement patterns with ChatGPT and Gemini.

User Experiences with LLM Integration

Facilitator: Interface Familiarity

Participants reported varied LLM usage frequencies: two used them daily, four weekly, one monthly, and three per semester. All had experience with ChatGPT; eight had used Gemini. ChatGPT’s mobile accessibility and conversational interface facilitated integration into routines. P11 noted, “I found ChatGPT very convenient on my phone.” In contrast, Gemini’s search-engine-like interface caused initial uncertainty. P10 shared, “I wasn’t sure how to interact with it the first time.” These accounts underscore how interface design and familiarity shape engagement.

Usability and Interaction Dynamics (SQ1)

Facilitator: Structured Output

ChatGPT’s structured output and conversational tone were valued (e.g., P01, P04, P11). P01 stated, “I like how ChatGPT formats the content—like an outline I can follow.”

Barrier: Prompt Formulation

However, usability challenges persisted. Nine participants struggled with prompt formulation, requiring rephrasing for relevant outputs. P02 explained, “It sometimes misunderstands unless I rephrase.” Five found outputs repetitive or generic, indicating a need for prompt literacy. Gemini’s interface was less intuitive for some (e.g., P04, P06), particularly due to its search-oriented design, highlighting the role of adaptive interfaces in usability.

Academic Writing Support and Process Adaptation (SQ2)

Facilitator: Cognitive Offloading

Most participants (n=11) used LLMs for ideation, drafting outlines, or rephrasing, reducing mental effort—a form of cognitive offloading. P06 shared, “It saves time when I’m stuck—it gives me a start.”

Barrier: Fabricated Citations

However, six expressed concerns about over-reliance. P09 noted, “If I always use it, I stop thinking deeply.” Five encountered fabricated citations in complex prompts. P03 recounted, “The references looked real, but when I checked, they didn’t exist.” These findings highlight LLMs’ value in early writing stages but underscore the need for critical engagement and verification.

Adoption Motivations and Tool Comparison (SQ3)

Facilitator: Accessibility

Utility and accessibility drove adoption. Eleven participants preferred ChatGPT for its ease of use and reliability. Gemini was less familiar, partly due to limited access in Bangladesh. P07, in Canada, valued Gemini’s image generation: “It helped with visual explanations.” Cost-free access was crucial for all, especially without institutional support for premium tools. These patterns reflect how infrastructure and exposure shape tool preferences.

Ethical Concerns (SQ4)

Barrier: Plagiarism Concerns

Nine participants raised concerns about plagiarism and originality. P12 remarked, “I worry that I might use something too directly.” Ambiguity in university guidelines stressed several, with P09 noting, “There are no clear rules on what’s allowed.” Six reported reduced independent writing effort when LLMs were available, signaling potential skill erosion. These concerns highlight the need for clear policies and AI literacy programs.

Comparative Thematic Summary

Table 2 synthesizes coded themes based on participants’ experiences. Preferences reflect explicit statements about tool strengths and usability. Unequal exposure to Gemini, due to its lower availability in Bangladesh, limited comparative insights for some participants.

Table 2: Thematic Prevalence and Comparative LLM Use among Bangladeshi University Students

Theme	Mentioned (n)	ChatGPT Positive (n)	Gemini Positive (n)	Observations
Usability Facilitators	12	9	5	ChatGPT appreciated for mobile ease (P11); Gemini noted for output detail (P05)
Usability Challenges	9	6	3	Prompting difficulties common; Gemini less intuitive for some (P04)
Writing Support	11	8	4	ChatGPT aided structure (P04); Gemini helped elaborate (P03)
Writing Concerns	6	4	2	Over-reliance and citation issues more common with ChatGPT
Adoption Drivers	11	8	3	ChatGPT favored for familiarity and free use; Gemini praised for visuals (P07)
Adoption Barriers	8	5	3	Technical inconsistencies noted for both tools
Ethical Concerns	9	6	2	Plagiarism and policy gaps spanned both platforms

Note. Participant counts reflect explicit comments on themes. Unequal exposure to Gemini, due to its lower availability in Bangladesh compared to ChatGPT, influenced comparative preferences.

Discussion

This discussion synthesizes how 12 Bangladeshi university students experienced and evaluated the usability of ChatGPT and Gemini in academic writing tasks. Anchored in the Technology Acceptance Model (TAM), writing process theory, and ethical HCI principles, it addresses the study’s central research question and four sub-questions. The findings highlight LLMs’ potential to enhance writing efficiency while revealing usability, adoption, and ethical challenges shaped by Bangladesh’s resource-constrained context.

Usability Implications

ChatGPT’s ease of interaction, time-saving features, and mobile accessibility (n=12) align with Shaikh et al. (2023), supporting cognitive offloading in early drafting as a usability benefit. However, prompt formulation difficulties (n=9) echo Shaikh et al. (2023), emphasizing prompt literacy as a critical skill. Gemini’s less intuitive, search-oriented interface posed challenges for some, underscoring the need for adaptive design in low-digital-literacy settings. These insights highlight that usability in Global South contexts requires addressing literacy disparities and infrastructural constraints.

Academic Writing Support

LLMs facilitated ideation and structuring (n=11), extending findings on AI’s motivational benefits in Bangladesh’s virtual learning. While Talukder & bin Ahsan (2025) focus on motivational AI use in virtual learning, noting over-reliance (51.6%) and critical thinking concerns, this study reveals task-specific challenges, including over-reliance (n=6) in structured writing, suggesting LLMs may reduce independent effort in cognitively intensive tasks, consistent with Klimova & de Campos (2024). Fabricated citations (n=5) further highlight the need for critical verification, particularly in resource-constrained settings lacking robust academic support.

Adoption Dynamics

Adoption aligned with TAM’s perceived usefulness and ease of use, with ChatGPT preferred (n=11) for accessibility and familiarity. Gemini’s lower familiarity, due to limited availability in Bangladesh, moderated its adoption, though its visual capabilities were valued in specific contexts (e.g., P07 in Canada). These findings reflect Koivisto’s (2023) emphasis on context-dependent adoption and Abdaljaleel et al. (2024) note on inaccuracies as barriers, underscoring the role of infrastructure in shaping preferences.

Ethical and Institutional Challenges

Plagiarism concerns (n=9) and guideline ambiguity (P09: “There are no clear rules”) mirror Naidu & Sevnarayan (2023) and Ullah et al. (2024). Six participants noted reduced independent effort, suggesting potential long-term cognitive implications, as implied by Klimova & de Campos (2024) on over-reliance and Talukder & bin Ahsan (2025)on critical thinking concerns. These issues highlight systemic policy gaps in Bangladesh’s higher education, necessitating AI literacy programs and transparent guidelines to ensure ethical use.

Implications for Practice, Policy, and Design

The findings propose the following:

• Educational Practice: Instructors should integrate prompt literacy and critical reflection into curricula, as Nguyen et al. (2024) suggest, to mitigate over-reliance and enhance engagement.
• Policy Development: Institutions must establish HCI-informed policies clarifying AI use, including plagiarism thresholds and citation verification, aligning with Talukder & bin Ahsan (2025) and Ullah et al. (2024).
• Technology Design: Developers should incorporate transparency features (e.g., source verification) and context-sensitive interfaces, addressing hallucination and usability (Kayalı et al., 2023).

These implications advocate for inclusive LLM integration that balances innovation with pedagogical and ethical integrity in resource-constrained contexts.

Limitations and Future Research

This study offers valuable insights into the usability of LLM chatbots among Bangladeshi university students, contributing a task-specific, Global South perspective. However, its limitations provide opportunities for future research. The purposive sample (n=12) ensures contextual depth but limits generalizability to broader populations. While ChatGPT and Gemini were the primary focus, incidental mentions of DeepSeek and Copilot by three participants suggest potential for cross-platform comparisons. The task-based design captured immediate usability patterns but did not assess long-term impacts on critical thinking or writing development. Additionally, excluding educator perspectives constrained triangulation, as instructors could reveal institutional expectations or policy gaps influencing student LLM use.

Future research should expand to include diverse LLM platforms, longitudinal studies, and a wider range of users. Incorporating educator input will enhance understanding of institutional dynamics. Advanced HCI methods, such as iterative testing across varied writing tasks, and diary-based studies tracking long-term cognitive effects will further inform context-sensitive, ethically grounded AI policies in higher education.

Conclusion

This study provides context-rich insights into the comparative usability of ChatGPT and Gemini among Bangladeshi university students, addressing user experiences, writing support, adoption patterns, and ethical concerns in academic writing tasks. LLMs offer significant benefits—time-saving, accessibility, and structured outputs—but challenges persist, including prompt formulation difficulties, output inconsistencies, and unclear institutional guidelines on ethical use.

Building on prior research on AI-facilitated engagement in virtual learning (Talukder & bin Ahsan, 2025), this study shifts focus to task-specific behaviors, revealing that LLM use in structured academic writing may reduce independent effort for some users, potentially impacting critical engagement. This distinction highlights the need to balance motivational adoption with sustained cognitive performance.

The findings contribute to educational HCI by: (1) advancing comparative knowledge of LLM usability in a resource-constrained Global South context; (2) emphasizing the need for targeted AI literacy programs to promote responsible use; and (3) exposing institutional policy gaps that shape ethical adoption. As LLMs become integral to academic practice, understanding their usability and limitations in specific educational contexts will ensure equitable, ethical, and effective integration.

Acknowledgements

We extend our sincere gratitude to the university student participants whose insights and engagement were essential to this study’s success. We thank Userhub for their support in coordinating research logistics and ensuring secure data handling. AI tools enhanced language clarity and presentation under the research team’s critical oversight to maintain fidelity to the study’s findings.

Declaration of Interest

The authors declare no conflicts of interest.

References

Abdaljaleel, M., Barakat, M., Alsanafi, M., Salim, N. A., Abazid, H., Malaeb, D., Mohammed, A. H., Hassan, B. A. R., Wayyes, A. M., Farhan, S. S., Khatib, S. El, Rahal, M., Sahban, A., Abdelaziz, D. H., Mansour, N. O., AlZayer, R., Khalil, R., Fekih-Romdhane, F., Hallit, R., … Sallam, M. (2024). A multinational study on the factors influencing university students’ attitudes and usage of ChatGPT. Scientific Reports, 14(1), 1983. https://doi.org/10.1038/s41598-024-52549-8

Alam, M. J., Ogawa, K., & Islam, S. R. Bin. (2023). e-Learning as a Doubled-Edge Sword for Academic Achievements of University Students in Developing Countries: Insights from Bangladesh. Sustainability, 15(9), 7282. https://doi.org/10.3390/su15097282

Aljuaid, H. (2024). The Impact of Artificial Intelligence Tools on Academic Writing Instruction in Higher Education: A Systematic Review. https://papers.ssrn.com/abstract=4814342

American Psychological Association. (2017). Ethical principles of psychologists and code of conduct. American Psychological Association. https://www.apa.org/ethics/code/

Baber, H., Nair, K., Gupta, R., & Gurjar, K. (2024). The beginning of ChatGPT – a systematic and bibliometric review of the literature. Information and Learning Sciences, 125(7/8), 587–614. https://doi.org/10.1108/ILS-04-2023-0035

Davis, F. D. (1989). Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology. MIS Quarterly, 13(3), 319. https://doi.org/10.2307/249008

Flower, L., & Hayes, J. R. (1981). A Cognitive Process Theory of Writing. College Composition and Communication, 32(4), 365. https://doi.org/10.2307/356600

Friedman, B., & Hendry, D. G. (2019). Value Sensitive Design. The MIT Press. https://doi.org/10.7551/mitpress/7585.001.0001

Guest, G., Bunce, A., & Johnson, L. (2006). How Many Interviews Are Enough? Field Methods, 18(1), 59–82. https://doi.org/10.1177/1525822X05279903

Kayalı, B., Yavuz, M., Balat, Ş., & Çalışan, M. (2023). Investigation of student experiences with ChatGPT-supported online learning applications in higher education. Australasian Journal of Educational Technology, 39(5), 20–39. https://doi.org/10.14742/ajet.8915

Klimova, B., & de Campos, V. P. L. (2024). University undergraduates’ perceptions on the use of ChatGPT for academic purposes: evidence from a university in Czech Republic. Cogent Education, 11(1). https://doi.org/10.1080/2331186X.2024.2373512

Koivisto, M. (2023). Tutoring Postgraduate Students with an AI-Based Chatbot. International Journal of Advanced Corporate Learning (IJAC), 16(1), 41–54. https://doi.org/10.3991/ijac.v16i1.35437

Kumar, H., Xiao, R., Lawson, B., Musabirov, I., Shi, J., Wang, X., Luo, H., Williams, J. J., Rafferty, A. N., Stamper, J., & Liut, M. (2024). Supporting Self-Reflection at Scale with Large Language Models: Insights from Randomized Field Experiments in Classrooms. Proceedings of the Eleventh ACM Conference on Learning @ Scale, 86–97. https://doi.org/10.1145/3657604.3662042

Muniruzzaman, S. Md., & Afrin, S. (2024). Improving academic writing skill: Difficulties encountered by undergraduates of English studies in Bangladesh. Englisia: Journal of Language, Education, and Humanities, 11(2), 215. https://doi.org/10.22373/ej.v11i2.19918

Naidu, K., & Sevnarayan, K. (2023). ChatGPT: An ever-increasing encroachment of artificial intelligence in online assessment in distance education. Online Journal of Communication and Media Technologies, 13(3), e202336. https://doi.org/10.30935/ojcmt/13291

Nguyen, A., Hong, Y., Dang, B., & Huang, X. (2024). Human-AI collaboration patterns in AI-assisted academic writing. Studies in Higher Education, 49(5), 847–864. https://doi.org/10.1080/03075079.2024.2323593

Rahman, Md. M., Hossain, M. A., & Uddin, M. M. (2024). Challenges in Academic Writing for Undergraduate Students in Bangladesh. European Journal of Teaching and Education, 6(4), 49–77. https://doi.org/10.33422/ejte.v6i4.1418

Sallam, M., Elsayed, W., Al-Shorbagy, M., Barakat, M., El Khatib, S., Ghach, W., Alwan, N., Hallit, S., & Malaeb, D. (2024). ChatGPT usage and attitudes are driven by perceptions of usefulness, ease of use, risks, and psycho-social impact: a study among university students in the UAE. Frontiers in Education, 9. https://doi.org/10.3389/feduc.2024.1414758

Shaikh, S., Yayilgan, S. Y., Klimova, B., & Pikhart, M. (2023). Assessing the Usability of ChatGPT for Formal English Language Learning. European Journal of Investigation in Health, Psychology and Education, 13(9), 1937–1960. https://doi.org/10.3390/ejihpe13090140

Talukder, N. M., & bin Ahsan, W. (2025). The Impact of AI on Student Engagement in Virtual Learning within Bangladesh’s Higher Education Sector. Userhub Journal. https://doi.org/10.58947/journal.jtgx34

Ullah, M., Bin Naeem, S., & Kamel Boulos, M. N. (2024). Assessing the Guidelines on the Use of Generative Artificial Intelligence Tools in Universities: A Survey of the World’s Top 50 Universities. Big Data and Cognitive Computing, 8(12), 194. https://doi.org/10.3390/bdcc8120194