The New Writing
Partner? EFL Students’ Real-Time Use of GenAI Tools in Academic Composing
Tran Thuy Khanh-Quynh[1]
& Ngo Cong-Lem2*
|
*Correspondence: lemnc@dlu.edu.vn |
|
1 Department
of Foreign Languages, Foreign Trade University - Ho Chi Minh City Campus,
Vietnam 2
Faculty of Foreign Languages, Dalat University,
Vietnam |
Abstract
While
existing research has largely focused on students’ perceptions of GenAI tools,
there remains limited empirical evidence of how EFL university students engage
with these tools during real-time writing processes, particularly within Asian
higher education contexts. Addressing this gap, the present study investigates
how Vietnamese EFL university students use GenAI tools when completing academic
essay tasks and examines the functions these tools serve across different
stages of writing. Drawing on 39 screen-recorded writing sessions from students
enrolled in academic English courses, the study adopts a qualitative,
process-oriented approach to analyse GenAI use during composition. The analysis
showed that students used GenAI across three interrelated domains of writing:
content construction, where GenAI supported idea generation, summarisation, and
content development; language mediation, where it assisted with vocabulary,
translation, paraphrasing, and proofreading; and process regulation, where it
provided feedback, explanations, and guidance during drafting and revision. The
findings further revealed several pedagogical and ethical issues, including
uneven evaluative engagement with GenAI output, blurred authorship boundaries,
and ethical ambiguity when GenAI was used to generate or shape writing content
and structure. While some students engaged with GenAI reflectively, others
incorporated generated, translated, or paraphrased content with limited visible
transformation, raising questions about authorial control, acceptable
assistance, and students’ engagement with core writing processes.
Keywords:
Generative
AI, EFL learners, academic integrity, Vietnam, writing
1. Introduction
The emergence of generative artificial
intelligence (GenAI) tools such as ChatGPT has intensified debates across
higher education, particularly regarding their influence on academic writing.
Within days of its release, ChatGPT attracted more than one million users
(Teubner et al., 2023), signalling both the accessibility of these systems and
the immediacy with which students began integrating them into their academic
routines. Recent research shows that students now turn to GenAI not only for
language correction but also for idea generation, content development,
information synthesis, and feedback, practices that collectively reshape how
writing is conceptualised and carried out in universities (Cummings et al.,
2024; Xiao et al., 2026). For students working in linguistically demanding
environments, especially those in English-as-a-Foreign-Language (EFL) settings,
such affordances may offer timely scaffolding that supports both comprehension
and production (see a systematic review, Teng, 2024).
At the same time, growing dependence on GenAI has
raised concerns about academic integrity, reduced cognitive engagement, and the
potential marginalisation of student agency in writing processes (Brickhill et
al., 2025; Hyland, 2026; Zhai et al., 2024). Specifically, scholars warn that
relying on GenAI for paraphrases, summaries, or sentence expansions can
diminish creativity by bypassing the iterative drafting processes essential for
knowledge construction and critical reflection (Cong-Lem & Thuy-Dung, 2026;
Ramoni et al., 2024; Subaveerapandiyan et al., 2025).
Furthermore, students utilising GenAI for multimodal composing may engage in
pattern-based elaboration rather than original synthesis, potentially weakening
the development of a distinct authorial voice (Liu et al., 2024). Compounding
these cognitive and creative risks, educators also express serious concerns
about work authenticity, cautioning that students may submit unverified,
hallucinated, or fully AI-generated text without comprehension, thereby turning
a potentially productive mediational tool into a shortcut that bypasses core
learning activities (Liu et al., 2024; Mohammadkarimi,
2023; Pratiwi et al., 2025; Rowland, 2023; Seo, 2024). Ultimately, scholars are
left to navigate a critical pedagogical paradox: while GenAI has the potential
to scaffold writing development, it risks circumventing the learning process
and undermining the very skills students are expected to master (Pratiwi et
al., 2025).
Despite expanding interest, empirical
understanding of how students actually use GenAI tools during the writing
process remains limited. Much of the existing literature relies on surveys or
interviews, which offer insight into perceptions and ethical concerns but
rarely capture real-time writing behaviour (Almanea,
2024; Ofem et al., 2025). Few studies trace GenAI use across multiple stages of
writing, from brainstorming and drafting to revising and editing, and even
fewer do so in naturalistic, classroom-based contexts. As a result, little is
known about the functional roles GenAI plays in shaping students’
moment-to-moment writing decisions.
The present study addresses these gaps by
examining how university students in Vietnam, an underrepresented EFL context,
engage with GenAI tools during academic essay writing. Drawing on 39
screen-recorded writing sessions collected in English for Academic Purposes
(EAP) classes, the study provides direct behavioural evidence of students’
GenAI use across the entire writing arc. This process-oriented approach moves beyond
self-reported perceptions and captures how students integrate GenAI into their
real-time decision-making, linguistic problem-solving, and text construction. The
study offers a nuanced understanding of how GenAI tools mediate authentic
writing practices in a rapidly digitising EFL environment. Through direct
observation of students’ composing processes, the research contributes evidence
on the opportunities and risks that GenAI introduces into academic writing,
insights that are increasingly vital for pedagogy, assessment, and policy
development in higher education.
2. Literature Review
2.1
AI Tools in Academic Writing: Expanding Functional and Pedagogical Support
GenAI tools are reshaping how university students
approach academic writing. Powered by large language models (LLMs), tools such
as ChatGPT, Wordtune, and QuillBot
are increasingly used to scaffold composing processes through idea generation,
language enhancement, and structural refinement (El-Garawany,
2024; D. Zhao, 2025). Rather than acting merely as digital dictionaries or
grammar checkers, these tools function as interactive writing partners that
offer real-time support for planning, organising, and revising texts,
affordances that are especially valuable for EFL learners, who often face
linguistic and cognitive constraints during composition (Liu et al., 2024;
Wang, 2025).
Research has shown that GenAI writing tools can
offload lower-order demands such as word retrieval, error correction, and
syntactic structuring, enabling learners to focus more on global concerns like
argumentation and coherence (Dong, 2023; Gayed et al., 2022). Their integration
is not only technological but also pedagogical: they align with principles of
strategy-based, feedback-rich writing instruction (Strobl et al., 2019) and may
help address practical constraints in many EFL contexts, such as large class
sizes and limited teacher feedback (Han & Li, 2024). This evolving writing
ecology has prompted calls for a reconceptualisation
of digital literacy. Kalantzis and Cope (2025) argue that AI literacy,
including prompt design, critical filtering of GenAI output, and ethical
judgment, is now a foundational component of contemporary writing education.
2.2
Cognitive, Creative, and Ethical Concerns in GenAI-Assisted Writing
Despite these affordances, GenAI introduces
notable risks. Scholars have raised concerns about reduced cognitive engagement
and diminished creativity, particularly when students rely heavily on
GenAI-generated paraphrases, summaries, or sentence expansions in ways that
bypass the iterative drafting processes essential for knowledge construction
and critical reflection (Ramoni et al., 2024; Subaveerapandiyan
et al., 2025). Liu et al. (2024) further show that students using GenAI for
multimodal composing may engage in pattern-based elaboration rather than
original synthesis, potentially weakening the development of authorial voice.
Educators have also expressed concerns about the
authenticity and integrity of student work. Studies report that some learners
use GenAI to generate entire paragraphs or rewrite content without adequate
comprehension (Mohammadkarimi, 2023), blurring the
boundaries between acceptable support and inappropriate substitution (Rowland,
2023). As generative tools become more sophisticated, distinguishing
student-driven writing from machine-driven production becomes increasingly
challenging.
Another emerging concern is information accuracy.
Seo (2024) highlights the phenomenon of GenAI hallucination, where models
produce plausible yet fabricated content. This places new evaluative burdens on
students, who must verify sources and assess the credibility of generated text.
Ethical GenAI use therefore requires explicit instruction in checking factual
accuracy, maintaining originality, and applying citation norms appropriately (Bieliauskaitė, 2021).
2.3
A Conceptual Framework of GenAI-Mediated L2 Writing
To understand students’ GenAI-supported academic
writing practices, the present study draws on a conceptual framework synthesised
from previous literature on L2 academic writing competence, AI literacy,
self-regulated learning, and academic integrity. Rather than serving as a rigid
coding scheme, this framework provides sensitising concepts for interpreting
how students used GenAI to support, regulate, or substitute aspects of the L2
writing process (see Figure 1).
At the centre of this framework is L2 academic
writing competence, understood as students’ ability to construct coherent,
meaningful, and rhetorically appropriate academic texts in English. This
competence includes not only grammatical accuracy and lexical range, but also
discourse organisation, argument development, source engagement, and authorial
voice (Xie & Lei, 2022; Yoon & Tabari, 2023). In the present study, L2
academic writing competence provides the core lens through which GenAI use is
interpreted. GenAI may support students’ writing development by scaffolding
linguistic expression, idea development, and revision. However, when used
uncritically, it may also reduce students’ opportunities to engage in the
cognitive, rhetorical, and linguistic work through which such competence is
developed. Around this central construct, AI literacy, self-regulated learning,
and academic integrity offer complementary lenses for understanding students’
engagement with GenAI.
AI literacy, as articulated by Kalantzis and Cope
(2025), encompasses far more than technical proficiency. It involves evaluative
judgment, ethical awareness, and the ability to craft effective prompts and
interpret GenAI outputs critically. From this perspective, learners are not
passive recipients of machine suggestions but active decision-makers who assess
the relevance, quality, and appropriateness of GenAI-generated support. This
emphasis on learner agency resonates with the broader orientation of Digital
Applied Linguistics (DAL), where technology use is understood not simply as
interaction with digital tools but as participation in human-mediated, socially
organised knowledge practices (Teng & Shen, 2026). In the GenAI era,
therefore, AI literacy can be seen as part of learners’ capacity to engage
critically and collaboratively in digital language practices, where knowledge
is continuously constructed, circulated, and refined through shared purposes,
participatory engagement, and collaborative intellectual work.
If AI literacy explains the critical capacities
students need to engage with GenAI, self-regulated learning theory helps
explain how these capacities are enacted during the writing process. This view
intersects meaningfully with self-regulated learning theory (Zimmerman, 2000),
in which writing is understood as a goal-directed process involving strategic
planning, ongoing monitoring, and adaptive revision. GenAI use can be seen as a
regulatory behaviour, sometimes compensatory (to overcome linguistic limitations),
sometimes strategic (to optimise workflow), and sometimes expedient (Teng,
2025). The quality of such regulation, however, depends on students’
metacognitive awareness and their ability to evaluate GenAI suggestions without
becoming dependent on them.
Figure
1. A Conceptual Framework for Understanding GenAI-Mediated L2 Academic Writing
Development
Academic integrity constitutes the third dimension
of the framework. In contexts where institutional policies on GenAI remain
unclear or unevenly implemented, students are often required to make their own
judgments about what forms of assistance are appropriate. This points to the
need for a contextualised understanding of academic integrity, one that
acknowledges the ethical uncertainty students may experience as they navigate
emerging technologies, peer norms, and assessment demands (Bieliauskaitė,
2021; Rowland, 2023). This dimension underscores that GenAI-supported writing
practices are shaped not only by individual choices but also by institutional
and socio-cultural constraints.
In summary, this framework understands
GenAI-supported L2 academic writing as a practice centred on the development
and enactment of L2 academic writing competence. AI literacy, self-regulated
learning, and academic integrity provide complementary lenses for understanding
how students evaluate GenAI outputs, regulate their use of GenAI across the
writing process, and negotiate the ethical boundaries of acceptable assistance.
Together, these dimensions help explain whether GenAI functions as a meaningful
scaffold for students’ academic writing development or reduces their engagement
with the linguistic, rhetorical, and cognitive work required to become
competent L2 academic writers.
2.4
Gaps in the Literature and Study Rationale
Although GenAI tools are now embedded in writing
instruction, much of the existing research focuses on learner perceptions or on
evaluating final written products, with limited insight into how students
actually use GenAI during writing (Malik et al., 2023; Utami et al., 2023).
Self-report data and controlled experiments provide valuable perspectives but
cannot reveal the moment-to-moment decisions and strategies that shape
tool-mediated composing.
There remains a substantive gap in tracing GenAI
engagement across multiple stages of writing, from prewriting to drafting and
revising, particularly in authentic, high-stakes academic contexts. Recent
studies have begun to examine GenAI’s influence on revision practices
(Warschauer et al., 2023; X. Zhao, 2023) and multimodal composition (Liu et
al., 2024), but few have used screen recordings to capture authentic,
unscripted interactions with GenAI tools.
The present study addresses these gaps by
analysing 39 screen-recorded writing sessions produced by Vietnamese EFL
university students. This approach offers fine-grained behavioural evidence of
GenAI use during authentic writing tasks, enabling close examination of how
students integrate GenAI into their composing processes. The Vietnamese
context, marked by centralised curricula, exam-oriented instruction (Duong et
al., 2025; Truong et al., 2026), and uneven access to GenAI-related training
(Cong‐Lem et al., 2025; Nguyen et al., 2025), provides a particularly revealing
setting, one that mirrors conditions in many other EFL and Global South higher
education systems.
By combining a novel methodological approach with
a rarely studied context and an integrated conceptual framework, the study
seeks to deepen current understanding of how GenAI shapes academic writing
practices and to inform pedagogical and ethical responses to emerging
challenges with two following research questions:
RQ1:
What functions do GenAI tools serve across different stages of the writing
process of Vietnamese EFL learners?
RQ2:
What pedagogical and ethical issues emerge from students’ GenAI-supported
writing behaviours?
This study is significant in three main ways.
First, it contributes methodological value by moving beyond self-reported
accounts of GenAI use and examining students’ observable writing behaviours
through screen-recorded composing sessions. This allows a more
process-sensitive understanding of how GenAI enters students’ academic writing
as they plan, draft, revise, and manage language-related challenges. Second, by
focusing on Vietnamese EFL learners, the study adds evidence from an
underrepresented Asian higher education context where students’ engagement with
GenAI is shaped by linguistic demands, assessment expectations, and emerging
institutional guidance. Third, the study offers pedagogical relevance for EFL
writing instruction by clarifying the kinds of support students may seek from
GenAI and by foregrounding the need for writing pedagogy that develops
students’ critical AI literacy, self-regulated tool use, and ethical awareness.
In doing so, the study provides timely insights for teachers, curriculum designers,
and institutions seeking to respond constructively to GenAI’s growing presence
in academic writing.
3. Methodology
3.1
Research Design
This study adopted a qualitative, process-oriented
research design to examine how EFL university students engage with GenAI tools
during academic essay writing. The design is grounded in the view that writing
is a dynamic, multidimensional activity shaped by moment-to-moment
decision-making rather than a static product. Screen-recorded writing sessions
were therefore used as the primary data source to capture real-time composing
behaviours, including when and how students accessed GenAI tools, the types of
prompts they produced, and the extent to which GenAI-generated text shaped
subsequent drafting and revision.
3.2
Participants and Context
The study was conducted at a public university in
Vietnam, where English is taught as a compulsory subject across most
undergraduate programmes. Participants were students enrolled in English for
Academic Purposes (EAP) courses, an environment that requires academic writing
in English but remains firmly situated within an EFL context. The courses focus
on essay organisation, source integration, and academic vocabulary, but at the
time of data collection offered no formal instruction on the use of GenAI tools.
A total of 39 Vietnamese students who spoke
Vietnamese as their first language volunteered to participate in the study.
They had been placed in intermediate-to-advanced English courses according to
the university’s internal placement criteria, and their participation in the
research did not affect their course grades or academic standing.
3.3
Data Collection
Participants were recruited through a voluntary
invitation issued at the end of the semester and were asked to submit full
screen recordings of their academic essay-writing process. The writing task was
a take-home academic essay aligned with the course syllabus, which allowed
students to work under familiar conditions and use the digital tools they would
normally employ when writing independently. To support consistency across
submissions, students were provided with standardised instructions for recording
their full screen activity, including browser use, GenAI interactions, and the
writing environment, using any screen-recording software of their choice.
Recordings were screened for completeness and
technical quality. Submissions that omitted portions of the writing process,
contained significant audio–visual disruptions, or failed to capture GenAI
interactions when used were excluded from the dataset. A total of 39 qualified
screen-recording videos met the inclusion criteria and were retained for
analysis. Table 1 presents the distribution of participants across four course
levels, along with essay topics and the number of eligible submissions.
A schematic diagram illustrating the research
design, including participant recruitment, writing task requirements, data
submission workflow, and analytic procedures, is provided in Figure 2 to offer
a visual overview of the study’s methodological structure.
Figure
2. Schematic Diagram of the Research Design
3.4
Data Analysis
All videos were anonymised, labelled by course and
participant code (e.g., BAE2_1, ESP1_3), and analysed using qualitative content
analysis. The primary researcher conducted the initial coding of all
screen-recorded videos, taking detailed notes on visible GenAI tool use (e.g.,
ChatGPT, Grammarly, QuillBot) and the apparent
function each tool served in the writing process, such as summarising,
translating, generating examples, providing feedback, or proofreading. Coding
followed a hybrid inductive–deductive approach. Initial codes were informed by
previous studies (e.g., Warschauer et al., 2023), while additional codes were
developed inductively as new patterns of GenAI use emerged from the data. To
enhance analytic trustworthiness, a second researcher served as an auditor and
critical peer reviewer, examining coded samples, reviewing emerging themes, and
discussing interpretive decisions with the primary researcher. Discrepancies or
uncertainties were resolved through iterative discussion and refinement of the
coding framework.
3.5
Ethical Considerations
At the time of data collection, the researchers’
institution did not have a formal ethics review board. Nevertheless, all study
procedures were designed and implemented in accordance with widely accepted
ethical standards for research involving human participants. Prior to data
collection, students were informed in writing and orally about the nature and
purpose of the study, including the use of screen-recorded writing sessions for
research purposes. Participation was voluntary, and students were assured that
their decision to participate or decline would have no impact on their course
grades or academic standing. They were also informed that they could withdraw
at any point without explanation or consequence. All data were anonymised
during transcription and analysis, with identifying information removed to
protect participants’ privacy. Table 1 presents detailed information about the
participant groups and assignment questions.
4. Findings
4.1
RQ1: What Functions Do GenAI Tools Serve in Students’ Writing
Processes?
Across the 39 screen-recorded sessions, students
engaged with GenAI in ways that shaped the conceptual substance of their
writing, the linguistic realisation of their ideas, and the procedural and
metacognitive work of composing. At the initial coding stage, nine primary
functions of GenAI tool use were identified: (1) idea generation, (2)
vocabulary support, (3) translation, (4) paraphrasing, (5) summarisation, (6)
content generation, (7) feedback provision, (8) tutoring/explanation, and (9)
proofreading (see Table 2). These function-level codes were then organised into
three broader analytic domains reported in this section: content construction,
language mediation, and process regulation. Specifically, idea generation,
summarisation, and content generation were grouped under content construction;
vocabulary support, translation, paraphrasing, and proofreading were grouped
under language mediation; and feedback provision and tutoring/explanation were
grouped under process regulation. Screenshots and transcripts were used to
support interpretation and validate categorisation. Together, these three
interrelated domains formed a dynamic ecosystem of GenAI use, in which students
navigated the affordances and constraints of AI-supported writing in uneven and
sometimes contradictory ways.
Table 1. Information of Participant Groups
|
Participant Group |
English course |
Year |
Majors |
No. of qualified respondents |
Assignment Questions |
|
Group 1 |
BAE2 |
Freshmen |
Business
Administration |
2/5 |
Choose one of the questions and write
an essay to share your opinion. 1/Are video games considered as art? 2/ Is sculpture considered as art? |
|
Group 2 |
BAE3 |
Sophomores |
Business Administration |
1/4 |
Choose one of the
questions and write an essay to share your opinion. 1/ Do you think
gunpowder is the most important invention of the world. 2/ Technology has developed rapidly in the last
century? Do you think its development is slowing now? |
|
Group 3 |
BAE4 |
Sophomores |
Business
Administration |
22/45 |
Write a
summary and lesson(s) you learned from the documentary film Money Monster |
|
Group 4 |
ESP1 |
Juniors |
Business Administration + Finance |
14/70 |
Write a summary and lesson(s) you learned from the documentary film
INSIDE JOB |
4.1.1 AI as a resource for content construction and transformation
A prominent pattern involved students mobilising
GenAI to assist with the conceptual labour of writing. Within this domain,
GenAI was mainly used for idea generation, summarisation, and content
generation, particularly when learners faced uncertainty about how to initiate,
structure, or develop their essays. Students routinely requested topic
contextualisation, argument scaffolds, and structural outlines, using
GenAI-generated responses as springboards for drafting. For example, BAE2_1
queried ChatGPT for arguments supporting the classification of video games as
art and subsequently aligned the first draft with the GenAI-generated
structure. Similarly, BAE4_20 solicited thematic information about Money
Monster to anchor the introductory framing of the essay.
GenAI also mediated students’ engagement with
information-rich or cognitively demanding source materials. Documentary-based
assignments prompted learners to input film titles into GenAI tools to obtain
condensed summaries, thematic breakdowns, or organising frames. ESP1_14 asked
Gemini to divide Inside Job into “five main parts,” while ESP1_1 relied on
NoteGPT to generate a concept map of key ideas. These uses enabled students to
establish a macro-structure for their essays with relative ease. However, their
uptake varied markedly: some students transformed the generated content through
reorganisation, selective adaptation, or further elaboration, whereas others,
such as ESP1_12, incorporated GenAI-generated sentences with minimal
alteration, effectively outsourcing parts of the interpretive process.
Content generation was also evident when students
asked GenAI to produce examples, explanations, introductory material, or
supporting points that could be inserted into their drafts. In these cases,
GenAI functioned not merely as a source of inspiration but as a direct
contributor to the substance and structure of the text. While such uses helped
students access conceptual resources and sustain writing momentum, they also
raised pedagogical concerns. When GenAI-generated ideas were critically
evaluated, adapted, and integrated into students’ own arguments, the tool
appeared to function as a scaffold for content development. When generated
content was adopted with little transformation, however, GenAI risked reducing
students’ engagement with the analytical and meaning-making work central to
academic writing.
4.1.2 AI as a mediator of language use and expression
A second prominent domain of GenAI use concerned
linguistic mediation. Within this domain, students used GenAI primarily for
vocabulary support, translation, paraphrasing, and proofreading, drawing on
these tools to address gaps in lexical knowledge, grammatical accuracy,
expression, and bilingual composing. These practices show that GenAI operated
as a flexible linguistic scaffold, enabling students to move between Vietnamese
and English, refine sentence-level expression, and improve the perceived
fluency of their academic texts.
Students frequently used GenAI for lexical
queries, which varied from simple synonym searches to more elaborate requests
involving usage examples, collocations, and pronunciation support. BAE2_1
exemplified this more exploratory pattern, iteratively refining vocabulary
choices and checking how particular words could be used in context. Others,
such as BAE3_1, treated GenAI as a quicker lexical substitution tool, seeking
straightforward synonyms to improve stylistic variation. These contrasting
patterns suggest that vocabulary support could function either as a site of
language learning or as a more surface-level strategy for textual improvement.
Translation emerged as a central strategy through
which students moved between Vietnamese and English during academic writing.
Students translated vocabulary items, sentences, and, in a few cases, entire
paragraphs. While some, such as BAE2_1, compared multiple translation outputs
to achieve greater semantic precision, others used translation in ways that
raised concerns about authorship and engagement. For example, BAE4_16
translated an online paragraph and inserted the output directly into the essay,
illustrating how translation can be appropriated not only to support expression
but also to circumvent parts of the composing process.
Students also used GenAI tools for paraphrasing,
particularly to reformulate sentences they had already drafted and to refine
the clarity, fluency, and academic appropriateness of their English expression.
In some instances, paraphrasing tools appeared to support students’ efforts to
clarify meaning and adjust expression. In others, however, paraphrasing
functioned more mechanically, allowing students to reword borrowed or
AI-generated text without substantive engagement with the underlying ideas.
This pattern highlights the dual role of paraphrasing as both a potential
language-learning resource and a possible mechanism for distancing students
from deeper textual decision-making.
Proofreading and sentence-level editing were
another important linguistic function of GenAI use. Near submission, many
students submitted whole drafts to GenAI tools for grammatical and stylistic
refinement. ESP1_9 incorporated nearly all of Gemini’s suggested edits, whereas
BAE4_11 selectively adopted changes, maintaining greater authorial control.
These contrasting cases reveal divergent levels of trust, discernment, and
linguistic awareness in students’ use of AI-generated corrections.
Collectively, these practices demonstrate how
GenAI mediated the linguistic complexity of EFL academic writing by supporting
vocabulary development, bilingual expression, reformulation, and textual
polishing. At the same time, the ease and immediacy of such support sometimes
obscured persistent gaps in students’ language proficiency and reduced
opportunities for deeper linguistic noticing. Thus, GenAI’s value as a language
mediator depends not only on the quality of its suggestions, but also on
whether students engage with those suggestions critically, selectively, and
reflectively.
Table 2. Functions of GenAI Use in Participants’
Academic Writing
|
Analytic domain |
Function |
Definition |
|
Content
construction |
Idea
generation |
Students used
GenAI to generate initial ideas, arguments, examples, or perspectives for
developing their essays. |
|
Summarisation |
Students used GenAI to condense source materials, documentary content,
or topic information into shorter summaries or thematic points. |
|
|
Content
generation |
Students used
GenAI to produce substantive textual content, such as explanations,
supporting points, introductory material, or draftable sentences. |
|
|
Language
mediation |
Vocabulary support |
Students used GenAI to search for word meanings, synonyms,
collocations, usage examples, pronunciation, or more appropriate lexical
choices. |
|
Translation |
Students used
GenAI to translate words, phrases, sentences, or longer passages between
Vietnamese and English. |
|
|
Paraphrasing |
Students used GenAI or AI-supported tools to reformulate existing
sentences or passages while preserving the intended meaning. |
|
|
Proofreading |
Students used
GenAI to check or improve grammar, spelling, punctuation, sentence structure,
style, or overall linguistic accuracy. |
|
|
Process
regulation |
Feedback provision |
Students used GenAI to obtain comments or evaluations on clarity,
coherence, organisation, argument development, or overall text quality. |
|
Tutoring/explanation |
Students used
GenAI to seek explanations about language use, writing quality, coherence,
grammar, or revision choices, often through dialogic interaction. |
4.1.3 AI as a tool for regulating and managing the writing process
GenAI also served important procedural and
metacognitive functions, particularly through feedback provision and
tutoring/explanation. In this domain, students used GenAI not only to improve
specific textual features, but also to manage uncertainty, monitor progress,
and make decisions during drafting and revision. These process-oriented uses
often intersected with content and language work, forming the connective tissue
of students’ composing practices.
Many participants relied on GenAI to navigate
local rhetorical challenges, such as writing transitions, clarifying topic
sentences, improving coherence, or generating concluding remarks. BAE4_3, for
example, asked ChatGPT to propose a concluding sentence to “tie the ideas
together,” indicating how GenAI was used to resolve structural and rhetorical
issues during drafting. In such cases, GenAI helped students maintain writing
momentum by offering immediate solutions to problems that might otherwise interrupt
the composing process.
More holistically, students engaged GenAI as a
feedback mechanism by requesting evaluations of clarity, cohesion,
organisation, and argument development. When BAE3_1 submitted an introduction
for critique, ChatGPT provided comments on logical flow and lexical choice.
Some students responded reflectively by testing recommended revisions, asking
follow-up questions, and selectively adapting the feedback. Others, however,
adopted suggestions wholesale without interrogating their relevance or
appropriateness. This variation suggests that feedback provision could support
self-regulated writing, but only when students retained an active evaluative
role in the revision process.
A smaller but pedagogically significant subset of
students used GenAI in tutor-like, dialogic exchanges. In these instances,
learners did not merely ask for corrections but sought explanations, asking
questions such as “Why is this sentence unclear?” or “How does this affect
coherence?” These exchanges, evident in extended interactions by BAE3_1, reveal
emerging metacognitive engagement and suggest GenAI’s potential to function as
a responsive instructional partner. However, such reflective and explanation-seeking
use remained less common than more direct correction- or feedback-seeking
behaviours.
Through these process-regulation functions, GenAI
helped students manage cognitive load, reduce uncertainty, and sustain progress
across the writing task. Yet, as with content construction and language
mediation, the degree of agency and critical judgment varied considerably.
GenAI supported students’ composing processes most productively when it
prompted reflection, monitoring, and informed decision-making. Conversely, when
students treated feedback or explanations as ready-made solutions, GenAI risked
weakening their participation in the evaluative and regulatory work in academic
writing.
In short, these patterns portray GenAI as a
multifaceted mediational resource embedded in learners’ composing trajectories.
Students did not use GenAI for a single, stable purpose; rather, they moved
fluidly between conceptual scaffolding, linguistic support, and process
regulation depending on task demands and momentary writing difficulties. This
interplay illustrates both the versatility and volatility of GenAI as an
academic composing companion: it can expand learners’ strategic repertoires
while also enabling forms of overreliance that displace essential cognitive and
metacognitive work. The heterogeneity observed across students underscores a
central insight from the analysis: GenAI’s pedagogical value is not inherent in
the technology itself but depends on how learners mobilise it, with what
intentions, and under what instructional conditions. This variability
foregrounds the pedagogical and ethical questions examined in RQ2.
4.2
RQ2: What Pedagogical and Ethical Issues Emerge From
Students’ GenAI-Supported Writing Behaviours?
Analysis of students’ real-time engagement with
GenAI tools revealed three interconnected issues shaping their writing
practices: uneven critical AI literacy, blurring of authorship boundaries, and
ethical ambiguity in a low-guidance institutional environment. These issues are
analytically distinct yet empirically intertwined, providing insight into how
students navigate GenAI within an EFL academic writing context.
4.2.1 Uneven evaluative engagement with GenAI output
Students varied widely in their ability to
question, evaluate, and regulate GenAI output. Some students displayed emerging
critical literacy by engaging in iterative, dialogic exchanges with GenAI
tools. For example, BAE3_1 frequently asked GenAI to explain why particular
sentences were weak or how cohesion could be improved, and then revised the
draft based on a reflective understanding of the suggestions. This behaviour
indicates a metacognitive use of GenAI as a learning resource rather than a
correction machine.
However, many students interacted with GenAI in a
more passive, acceptance-based manner. For instance, ESP1_9 submitted an entire
draft to a GenAI platform for proofreading and incorporated most revisions
directly into the text without further interrogation. Similar patterns emerged
when students asked for vocabulary or grammar checks but did not probe the
rationale behind the suggested changes.
This unevenness suggests that GenAI tools may
amplify existing differences in students’ linguistic awareness, metacognitive
ability, and confidence in evaluating textual choices. Without explicit
pedagogical scaffolding, some learners benefit from GenAI as a site of
cognitive apprenticeship, while others adopt the tool’s output uncritically,
weakening their own decision-making in the writing process.
4.2.2 Blurring of authorship boundaries and the potential for
substitution
A second issue concerns the difficulty many
students faced in distinguishing between acceptable support and unwarranted
substitution of core writing processes. Several recordings showed students
importing GenAI-generated content directly into their essays with minimal or no
modification.
A clear instance of this pattern occurred when
ESP1_12 requested a summary of the documentary Inside Job and inserted several
sentences almost verbatim into the essay. Likewise, BAE4_16 located a
Vietnamese paragraph online, asked GenAI to translate it into English, and
embedded the translated paragraph directly into the assignment. These
behaviours illustrate how GenAI can displace interpretive tasks such as
summarising, synthesising, or restating ideas in one’s own voice.
Paraphrasing tools also contributed to blurred
boundaries. For example, ESP1_4 adopted paraphrased output from QuillBot with minimal modification, suggesting limited
engagement with the process of refining meaning or developing stylistic control.
While paraphrasing may reduce plagiarism risk or increase clarity, its
uncritical use risks diminishing the cognitive operations central to L2 writing
development.
Together, these cases reveal a continuum ranging
from legitimate scaffolding (e.g., improving clarity or structure) to the
wholesale substitution of meaning-making. Students’ ability to locate
themselves on this continuum was inconsistent, highlighting the need for
clearer pedagogical cues about where support ends and authorship begins.
4.2.3 Ethical ambiguity in generating writing content and structure
The third issue concerns students’ use of GenAI to
generate or shape the content and structure of their writing, where the
boundary between legitimate support and excessive assistance became less clear.
In several recordings, GenAI was used not only for grammar correction,
vocabulary support, or sentence-level editing, but also for content selection,
idea organisation, summarisation, transitions, and conclusions. These uses were
ethically ambiguous because they involved parts of academic writing closely connected
to interpretation, meaning-making, and authorial control.
For instance, ESP1_14 asked GenAI to segment a
documentary into thematic parts to structure the essay, an approach that could
be interpreted either as useful scaffolding or as outsourcing part of the
analytic work. Similarly, ESP1_1 relied on a GenAI-generated mind map of the
film’s key ideas to guide the organisation of the essay. Other students, such
as BAE4_3, requested GenAI-generated concluding sentences or transitions,
suggesting that GenAI was also used to shape rhetorically meaningful parts of the
text.
Students using GenAI summarisation tools, such as
ESP1_12, also drew on generated summaries when engaging with source content.
From the screen-recorded data alone, it is difficult to determine whether such
practices reflected intentional misconduct, uncertainty, or a pragmatic attempt
to manage task demands. What the data show, however, is that GenAI became
involved in writing decisions where academic integrity, authorship, and learning
are difficult to separate. These observations point to a broader ethical
challenge in GenAI-mediated L2 academic writing: students may use GenAI in ways
that appear supportive at the surface level, while also reducing their
engagement with the interpretive and rhetorical work through which academic
writing competence is developed.
5. Discussion
5.1
Study Contributions
This study examined how Vietnamese EFL university
students engaged with GenAI tools during academic writing through the analysis
of 39 screen-recorded composing sessions. The findings show that GenAI was not
used simply as an external aid for final editing. Rather, it became embedded in
students’ real-time composing practices, shaping how they generated ideas,
resolved language problems, organised content, sought feedback, and revised
their texts. This process-oriented evidence extends prior research that has
relied largely on self-report or final-product analysis (Malik et al., 2023;
Utami et al., 2023) by making visible the moment-to-moment decisions through
which GenAI entered students’ writing processes.
Interpreted through the conceptual framework in
this study (see Figure 1), these findings show that GenAI use was closely
connected to the development and enactment of L2 academic writing competence.
Students used GenAI to support several components of this competence, including
lexical choice, grammatical accuracy, discourse organisation, idea development,
summarisation, and revision. In some cases, GenAI appeared to function as a
scaffold that helped students overcome linguistic constraints and move their
writing forward. For example, students used GenAI to search for vocabulary,
improve sentence clarity, organise ideas, or obtain feedback on coherence.
However, the findings also show that GenAI sometimes reduced students’
engagement with the cognitive, rhetorical, and linguistic work through which L2
academic writing competence is developed. This was particularly evident when
students incorporated generated summaries, translated paragraphs, or
paraphrased output with minimal transformation. Thus, the pedagogical
significance of GenAI lies not only in whether students use it, but in how its
use affects their participation in the work of composing.
A central insight from the study is that students’
GenAI engagement can be understood along a continuum between scaffolding and
offloading. Consistent with previous research suggesting that GenAI can reduce
lower-order cognitive demands and support writing development (Dong, 2023;
Gayed et al., 2022), some students used GenAI strategically by refining
prompts, comparing suggestions, and selectively adapting output. These
practices suggest a form of self-regulated writing in which GenAI was used to
support planning, monitoring, and revision. At the same time, other students
appeared to use GenAI more expediently, relying on generated or transformed
text to complete parts of the task with limited visible evaluation. These
patterns resonate with concerns about over-reliance on GenAI (Teng, 2025; Zhai
et al., 2024), although the screen-recorded data suggest that students’
reliance on GenAI was uneven, varying across tasks, writing stages, and
individual composing practices.
Although the present study did not directly
measure AI literacy as a competence, the screen-recorded data offered important
process-level evidence of how students visibly exercised different levels of
critical judgment during GenAI-supported writing. Some students engaged with
GenAI dialogically by asking for explanations, questioning sentence quality,
and using feedback to revise more reflectively. Others accepted AI-generated
suggestions with little visible evaluation. This contrast suggests that AI
literacy in GenAI-mediated writing should extend beyond technical familiarity
with AI tools. It should be understood as a situated capacity to judge the
relevance, accuracy, and appropriateness of AI-generated output, adapt it to
one’s communicative intentions, and integrate it without losing authorial
control. In this respect, the study supports recent views of AI literacy as
involving evaluative judgment, ethical awareness, and critical participation in
digital writing practices (Kalantzis & Cope, 2025; Teng & Shen, 2026),
while adding behavioural evidence of how such capacities are unevenly enacted
during real-time composing.
The study further highlights the authorship and
academic integrity challenges that emerge when GenAI becomes involved in
content generation, translation, summarisation, and paraphrasing. Previous
scholarship has raised concerns about the ethical ambiguity of GenAI use in
academic writing (Bieliauskaitė, 2021; Rowland,
2023), although much of this work has drawn on policy debates, student
perceptions, or retrospective accounts. The present findings add behavioural
evidence by showing how authorship boundaries became blurred through small,
incremental writing decisions. Practices such as using GenAI-generated
summaries, translating external Vietnamese content into English, or incorporating
paraphrased output from QuillBot demonstrate that
GenAI can become involved not only in surface-level language correction but
also in meaning-making and textual ownership. These practices do not
necessarily indicate intentional misconduct. Rather, they show that academic
integrity in GenAI-mediated L2 writing is often negotiated through situated
decisions about how much AI-generated support is acceptable, how much
transformation is required, and where student authorship remains visible.
A further pattern concerns the potential task
sensitivity of GenAI use. Although task type was not systematically analysed as
a separate variable, the recordings suggest that different assignments invited
different forms of GenAI mediation. Documentary-based summary tasks appeared to
encourage more use of summarisation, thematic segmentation, and content
organisation, whereas opinion-based essays appeared to elicit more idea
generation, lexical support, and argument development. This pattern aligns with
research showing that GenAI-mediated writing practices vary according to task
and genre demands (Liu et al., 2024; Warschauer et al., 2023). In the present
study, it also suggests that assessment design may shape the kinds of GenAI
assistance students seek, highlighting the need for assessment tasks to make
students’ interpretive, source-use, and composing processes more visible rather
than evaluating only the final written product.
In summary, these findings provide empirical
support for the conceptual framework presented in Figure 1 by showing that
GenAI-mediated L2 academic writing is shaped by the dynamic interaction among
writing competence, AI literacy, self-regulation, and academic integrity. L2
academic writing competence remains the central concern because students’ GenAI
use ultimately matters in relation to how they develop and demonstrate control
over language, discourse, source engagement, argument, and authorial voice. AI
literacy shapes whether students can evaluate and adapt GenAI output
critically. Self-regulated learning shapes whether GenAI is used strategically
to support writing or expediently to complete tasks. Academic integrity shapes
how students negotiate authorship, originality, and acceptable assistance. The
study therefore contributes a process-oriented account of GenAI use in EFL
academic writing, showing that GenAI can support writing development when it
extends students’ engagement with composing, but may constrain development when
it reduces their participation in the linguistic, rhetorical, and cognitive
work of writing.
5.2
Educational Implications: Toward Process-Oriented and Critical GenAI Writing
Pedagogy
The findings point to the need for a more
process-oriented approach to GenAI in EFL writing pedagogy. Rather than framing
GenAI use primarily as a question of prohibition or permission, writing
instruction should focus on how AI enters students’ composing processes and how
students can remain intellectually active when using it. The central
pedagogical issue, therefore, is not simply whether GenAI improves the final
written product, but whether it supports students’ engagement with the
linguistic, rhetorical, cognitive, and metacognitive work through which L2
academic writing competence is developed. From this perspective, GenAI should
be treated as a context for cultivating higher-order critical thinking,
including students’ capacity to evaluate AI-generated output, compare
alternative textual possibilities, justify revision decisions, and maintain
authorial responsibility for their writing.
First, AI literacy should be embedded within the
writing process rather than taught as a separate technical skill. The observed
variation in students’ engagement with GenAI output suggests that technical
access to AI tools does not necessarily lead to critical or productive use.
Students need explicit support in interpreting, questioning, and adapting
AI-generated suggestions in relation to their own communicative intentions.
Teachers can design activities in which students compare multiple AI responses
to the same prompt, identify inaccuracies or weak reasoning, explain why
particular suggestions are accepted or rejected, and revise AI-generated text
to better fit their intended meaning. Such activities can help students move
from passive uptake toward more evaluative, reflective, and self-regulated
engagement with GenAI.
Second, writing pedagogy should make authorship
and intellectual control more explicit. The findings show that GenAI may
support local aspects of writing, such as vocabulary choice, grammar
correction, and sentence clarity, but it may also become involved in more
substantive processes, including summarisation, translation, content
organisation, transition building, and conclusion writing. These uses differ in
the extent to which they require student transformation, interpretation, and
accountability. Teachers therefore need to help students distinguish between
using GenAI as a scaffold and allowing it to substitute for core meaning-making
work. Classroom discussion of concrete scenarios, such as brainstorming with
GenAI, translating external content, generating an outline, paraphrasing a
paragraph, or producing a conclusion, can help students develop a more nuanced
understanding of acceptable assistance and authorial responsibility.
Third, assessment design should make students’
writing processes, not only their final products, visible. Since different task
types may invite different forms of GenAI reliance, teachers should consider
how assignment design shapes students’ opportunities for critical engagement.
Tasks involving source comprehension, documentary summaries, or argument
development may require additional process-oriented components, such as
annotated drafts, AI-use reflections, revision logs, source-use notes, or short
oral explanations of writing decisions. These components do not necessarily
prohibit GenAI use. Instead, they help preserve students’ active participation
in interpreting information, constructing meaning, and making informed
rhetorical choices.
Overall, the findings suggest that EFL programs
may benefit from clearer pedagogical and institutional guidance on GenAI use in
academic writing. Rather than treating GenAI mainly as an academic integrity
issue, such guidance should support students’ gradual development as
independent and responsible L2 writers in AI-mediated environments. It should
help students distinguish between uses of GenAI that support language
development, idea refinement, and revision, and uses that may reduce their
engagement with the cognitive, rhetorical, linguistic, and ethical work
required for L2 academic writing development.
6. Conclusion
This study examined how Vietnamese EFL university
students engaged with GenAI tools during academic essay writing by analysing 39
screen-recorded composing sessions. The findings show that GenAI was used
across multiple stages of the writing process, including idea development,
language formulation, content organisation, and revision. Students’ engagement,
however, was not uniform; it varied in purpose, depth, and level of authorial
control. In some cases, GenAI appeared to support students’ development of ideas,
linguistic expression, and revision strategies, while in others it was used
more directly to generate, translate, summarise, or modify content with limited
visible transformation. By focusing on real-time composing behaviour, the study
contributes process-level evidence of how GenAI becomes embedded in EFL
students’ academic writing practices, extending prior research that has relied
largely on self-reported data. At the same time, the findings highlight
important pedagogical and ethical challenges, particularly uneven evaluative
engagement with GenAI output, blurred authorship boundaries, and questions
about how students engage with source material and construct their own
understanding.
Several limitations of this study should be
acknowledged. First, the analysis drew on a relatively small, context-specific
sample from a single institution, which may limit the transferability of the
findings. Second, although screen recordings provided valuable access to
students’ observable writing behaviours, they could not fully capture students’
internal reasoning, intentions, or decision-making processes. Third,
differences in task type and course level may have shaped patterns of GenAI use;
however, these factors were not systematically analysed as separate variables
in the present study. Future research could extend this work by combining
screen-recording data with interviews, stimulated recall, or learner
reflections to better understand the reasoning behind students’ actions.
Comparative studies across institutional contexts, proficiency levels, and
writing task types would also help clarify how different conditions shape
GenAI-mediated writing practices. Such work is needed to develop a more
comprehensive understanding of how GenAI is reshaping EFL academic writing and
the development of L2 writing competence.
Acknowledgements
The authors would like to thank the students who
participated in this study for their time and willingness to share their
writing processes. Their contributions made this research possible.
Declaration
of GenAI Use in the Writing Process
The authors used ChatGPT to improve language,
clarity, and readability. The authors reviewed and edited all AI-generated
content as necessary and take full responsibility for the content of the
published article.
Funding
This research received no external funding.
Conflict
of Interest
The authors declare that they have no competing
interests.
References
Almanea, M. (2024). Instructors’ and learners’ perspectives on
using ChatGPT in English as a foreign language courses
and its effect on academic integrity. Computer
Assisted Language Learning, 39(3),
512–527. https://doi.org/10.1080/09588221.2024.2410158
Bieliauskaitė, J. (2021). Solidarity in academia and its relationship
to academic integrity. Journal of
Academic Ethics, 19(3), 309–322. https://doi.org/10.1007/s10805-021-09420-6
Brickhill, M., Andrews, G.,
& Nieuwoudt, J. (2025). Developing student agency towards academic
integrity through an educative approach: Exploring students’ experiences and
perspectives. Journal of Academic Ethics,
23(3), 951–975. https://doi.org/10.1007/s10805-024-09567-y
Cong-Lem, N., &
Thuy-Dung, N. T. (2026). Towards a framework for understanding and assessing
critical thinking in generative AI-enhanced learning environments: A scoping
review. European Journal of Psychology of
Education, 41(2), 41. https://doi.org/10.1007/s10212-026-01089-y
Cong‐Lem, N., Van Bui, C. N.,
Nguyen, N. P. N., & Huynh, M. Q. (2025). Bridging or breaking? A systematic
review of how generative AI shapes equity in foreign language education. International Journal of Applied
Linguistics, 1–12. https://doi.org/10.1111/ijal.70052
Cummings, R. E., Monroe, S. M., & Watkins, M. (2024). Generative AI in first-year writing: An early analysis of affordances, limitations, and a framework for the future. Computers and Composition, 71, 102827. https://doi.org/10.1016/j.compcom.2024.102827
Duong, T. N. N., Hercz, M.,
Nguyen, D. K., Phan, T. T. V., Nguyen, T. C., Nguyen, V. T., & Pham, L. B.
N. (2025). Escaping the blame spiral: Student-centered teaching reform in
Vietnam. Studies in Higher Education, 50(11), 2631–2645. https://doi.org/10.1080/03075079.2024.2444390
Dong, Y. (2023).
Revolutionizing academic English writing through AI-powered pedagogy: Practical
exploration of teaching process and assessment. Journal of Higher Education Research, 4(2), 52–57. https://doi.org/10.32629/jher.v4i2.1188
El-Garawany,
M. S. M. (2024). The effects of a QuillBot-based
intervention on English language majors’ EFL writing performance, apprehension,
and self-efficacy. Language Teaching Research Quarterly, 43, 167–189. https://doi.org/10.32038/ltrq.2024.43.10
Gayed, J. M., Carlon, M. K.
J., Oriola, A. M., & Cross, J. S. (2022). Exploring an AI-based writing assistant’s
impact on English language learners. Computers
and Education: Artificial Intelligence, 3, 100055. https://doi.org/10.1016/j.caeai.2022.100055
Han, J., & Li, M. (2024).
Exploring ChatGPT-supported teacher feedback in the EFL context. System, 126, 103502. https://doi.org/10.1016/j.system.2024.103502
Hyland, K. (2026). AI feedback
to L2 writers. International Journal of
TESOL Studies, 8(1), 146–156. https://doi.org/10.58304/ijts.260420
Kalantzis, M., & Cope, B.
(2025). Literacy in the time of artificial intelligence. Reading Research Quarterly, 60(1), e591. https://doi.org/10.1002/rrq.591
Liu, M., Zhang, L. J., & Biebricher, C. (2024). Investigating students’ cognitive
processes in generative AI-assisted digital multimodal composing and
traditional writing. Computers &
Education, 211, 104977. https://doi.org/10.1016/j.compedu.2023.104977
Malik, A. R., Pratiwi, Y., Andajani, K., Numertayasa, I. W.,
Suharti, S., Darwis, A., & Marzuki. (2023). Exploring artificial
intelligence in academic essay: Higher education student’s perspective. International Journal of Educational
Research Open, 5, 100296. https://doi.org/10.1016/j.ijedro.2023.100296
Mohammadkarimi, E. (2023). Teachers’ reflections on academic
dishonesty in EFL students’ writings in the era of artificial intelligence. Journal of Applied Learning and Teaching, 6(2),
Article 2. https://doi.org/10.37074/jalt.2023.6.2.10
Nguyen, H., Thanh, V. D. T.,
& Dinh, V. P. (2025). Bringing AI into teaching: Understanding Vietnamese
teachers’ perspectives and pedagogical challenges. International Journal of Educational Methodology, 11(3), 335–347. https://doi.org/10.12973/ijem.11.3.335
Ofem, U. J., Owan, V. J.,
Iyam, M. A., Udeh, M. I., Anake, P. M., & Ovat, S. V. (2025). Students’ perceptions, attitudes and utilisation of ChatGPT for academic dishonesty: Multigroup
analyses via PLS‒SEM. Education and
Information Technologies, 30(1), 159–187. https://doi.org/10.1007/s10639-024-12850-5
Pratiwi, H., Suherman, Hasruddin, & Ridha, M. (2025). Between shortcut and
ethics: Navigating the use of artificial intelligence in academic writing among
Indonesian doctoral students. European
Journal of Education, 60(2), 1–16. https://doi.org/10.1111/ejed.70083
Ramoni, D., Sgura, C., Liberale, L., Montecucco, F., Ioannidis, J. P. A., & Carbone, F.
(2024). Artificial intelligence in scientific medical writing: Legitimate and
deceptive uses and ethical concerns. European
Journal of Internal Medicine, 127, 31–35. https://doi.org/10.1016/j.ejim.2024.07.012
Rowland, D. R. (2023). Two
frameworks to guide discussions around levels of acceptable use of generative
AI in student academic research and writing. Journal of Academic Language and Learning, 17(1), T31–T69.
Seo, J.-Y. (2024). Exploring
the educational potential of ChatGPT: AI-assisted narrative writing for EFL
college students. Language Teaching
Research Quarterly, 43, 1–21. https://doi.org/10.32038/ltrq.2024.43.01
Strobl, C., Ailhaud, E., Benetos, K., Devitt, A., Kruse, O., Proske, A., & Rapp,
C. (2019). Digital support for academic writing: A review of technologies and
pedagogies. Computers & Education,
131, 33–48. https://doi.org/10.1016/j.compedu.2018.12.005
Subaveerapandiyan, A., Kalbande, D., &
Ahmad, N. (2025). Perceptions of effectiveness and ethical use of AI tools in
academic writing: A study among PhD scholars in India. Information Development, 02666669251314840. https://doi.org/10.1177/02666669251314840
Teng, M. F. (2024). A systematic
review of ChatGPT for English as a foreign language writing: Opportunities, challenges,
and recommendations. International Journal of TESOL Studies, 6(3), 36–57.
https://doi.org/10.58304/ijts.20240304
Teng, M. F. (2025).
Understanding EFL student writers’ metacognitive awareness in utilizing
ChatGPT. System, 135, 1–17. https://doi.org/10.1016/j.system.2025.103848
Teng, M. F., & Shen, X.
(2026). Digital applied linguistics:
What is it and where are we in the GenAI era? Digital Applied Linguistics, 4, 1–29. https://doi.org/10.29140/dal.v4.103594
Teubner, T., Flath, C. M.,
Weinhardt, C., van der Aalst, W., & Hinz, O. (2023). Welcome to the era of
ChatGPT et al.: The prospects of large language models. Business & Information Systems Engineering, 65(2), 95–101. https://doi.org/10.1007/s12599-023-00795-x
Truong, T. D., Dinh, H. T.,
Ha, X. V., Pham, K. T., Nguyen, T. T., & Nguyen, D. V. (2026). Vietnamese
primary teachers’ autonomy in the context of educational reform. European Journal of Education, 61(2),
e70587. https://doi.org/10.1111/ejed.70587
Utami, S. P. T., Andayani, Winarni, R., & Sumarwati. (2023). Utilization
of artificial intelligence technology in an academic writing class: How do
Indonesian students perceive?
Contemporary Educational Technology, 15(4), ep450. https://doi.org/10.30935/cedtech/13419
Wang, C. (2025). Exploring
students’ generative AI-assisted writing processes: Perceptions and experiences
from native and nonnative English speakers. Technology, Knowledge and
Learning, 30(3), 1825–1846. https://doi.org/10.1007/s10758-024-09744-3
Warschauer, M., Tseng, W.,
Yim, S., Webster, T., Jacob, S., Du, Q., & Tate, T. (2023). The affordances
and contradictions of AI-generated text for writers of English as a second or
foreign language. Journal of Second
Language Writing, 62, 101071. https://doi.org/10.1016/j.jslw.2023.101071
Xiao, F., Zhu, S., & Xin,
W. (2026). Exploring the landscape of generative AI (ChatGPT)-powered writing
instruction in English as a foreign language education: A scoping review. ECNU Review of Education, 9(1),
20965311241310881. https://doi.org/10.1177/20965311241310881
Xie, Q., & Lei, Y.
(2022). Diagnostic assessment of L2 academic writing product, process and
self-regulatory strategy use with a comparative dimension. Language Assessment Quarterly, 19(3), 231–263. https://doi.org/10.1080/15434303.2021.1903470
Yoon, H.-J., & Tabari, M.
A. (2023). Authorial voice in source-based and opinion-based argumentative
writing: Patterns of voice across task types and proficiency levels. Journal of English for Academic Purposes, 62,
101228. https://doi.org/10.1016/j.jeap.2023.101228
Zhai, C., Wibowo, S., &
Li, L. D. (2024). The effects of over-reliance on AI dialogue systems on
students’ cognitive abilities: A systematic review. Smart Learning Environments, 11(1), 28. https://doi.org/10.1186/s40561-024-00316-7
Zhao, D. (2025). The impact
of AI-enhanced natural language processing tools on writing proficiency: An
analysis of language precision, content summarization, and creative writing
facilitation. Education and Information
Technologies, 30(6), 8055–8086. https://doi.org/10.1007/s10639-024-13145-5
Zhao, X. (2023). Leveraging
artificial intelligence (AI) technology for English writing: Introducing wordtune as a digital writing assistant for EFL writers. RELC Journal, 54(3), 890–894. https://doi.org/10.1177/00336882221094089
Zimmerman, B. J. (2000).
Chapter 2—Attaining self-regulation: A social cognitive perspective. In M. Boekaerts, P. R. Pintrich, &
M. Zeidner (Eds.), Handbook of self-regulation (pp. 13–39). Academic
Press. https://doi.org/10.1016/B978-012109890-2/50031-7
Author biodata
Tran Thuy Khanh-Quynh is a lecturer at
Foreign Trade University (Ho Chi Minh City branch) where she teaches General
English and English for Specific Purposes (ESP) courses. Her research interests
include academic writing, the integration of artificial intelligence in language
education, and teacher professional development.
E-mail: tranthuykhanhquynh.hcmc@ftu.edu.vn
ORCID: 0000-0003-1534-6955
Dr. Ngo Cong-Lem earned his PhD
from the Faculty of Education at Monash University, Australia. He is currently
a Lecturer in the Faculty of Foreign Languages at Dalat
University, Vietnam. His research focuses on generative AI, critical thinking,
emotions, and professional learning.
E-mail:
lemnc@dlu.edu.vn
ORCID:
0009-0005-3299-477X