Teachers’ Perceptions and Students’ Strategies in Using AI-Mediated Informal Digital Learning for Career ESL Writing

Abstract

This study aims to explore teachers’ perceptions and students’ strategies when integrating AI-mediated informal digital learning of English tools (AI-IDLE) into career ESL writing instruction. This case study involved six university instructors and over 300 students in an English writing course. Although AI-IDLE has broadened English access beyond classrooms, existing research on writing skills often neglects students’ diverse strategies that correspond to their professional aspirations, as well as teachers’ perceptions. The data included a demographic questionnaire, think-aloud protocols for real-time assessment of cognitive processes during the task, and semi-structured interviews for teachers’ validation. Findings reveal three major student strategies: (1) explicit genre understanding, (2) student-driven selection of digital multimodal tools—such as Grammarly, ChatGPT, Canva with Magic Write, and Invideo—to integrate text with images, sound, and layout for improved rhetorical accessibility, and (3) alignment with students’ post-graduation career needs. Students’ work with these AI tools demonstrated that when they created projects aligned with professional identities and future job needs, they became more aware of how to improve their writing; however, the teachers expressed hopes and doubts about the tools’ effectiveness and authenticity of the students’ work. Suggestions to use AI-IDLE to improve writing were provided.

1. Introduction

The advent of AI-driven technologies marks a paradigm shift in English language education, particularly in how teachers help English-as-a-foreign-language (EFL) students access, interact with, and personalize their learning journeys. Within this evolving landscape, a case in point, AI-mediated informal digital learning of English (AI-IDLE) has transformed the concept of informal digital learning (IDLE) in the traditional sense, which consists of “all forms of intentional or tacit learning in which students engage with digital tools either collectively or individually without direct reliance on a teacher or externally organized curriculum” (Livingstone, 2008). IDLE is thus characterized based on students’ personal interests and preferences, manifested through diverse patterns of tool use and purposeful interaction with digital resources (Dressman, 2023; Lee & Lee, 2021; Soyoof et al., 2023; Zhao & Nazir, 2022). In EFL settings, IDLE helps integrate language practice into students’ daily lives while also complementing classroom instruction, enhancing their overall English proficiency and learning experiences, and supporting curricular and pedagogical innovation (Lee & Dressman, 2018; Lee & Lee, 2021; G. Liu & Darvin, 2023; G. L. Liu et al., 2024; G. Liu et al., 2023; Zhang & Liu, 2022).

With the added integration of artificial intelligence, these AI-mediated tools, such as intelligent grammar checkers, conversational agents, and composition platforms, now mimic the support of a real-life tutor by enhancing learner autonomy and offering personalized learning pathways, meaningful real-time feedback, and simulated communication scenarios (Elimadi et al., 2024; Kavitha et al., 2025; Lakshmi et al., 2022). They can also generate texts that are contextually appropriate and recognize language varieties, cultural references, and genres (Tam, 2023), thereby potentially surpassing the capabilities of earlier informal digital learning models. Moreover, AI can take the form of multimodal systems capable of processing and generating multiple types of input and output, such as text, images, audio, and video. The release of GPT-4o in May 2024, for example, marked a significant milestone, introducing the first fully integrated multimodal model made widely accessible to the public. According to Carciu and Muresan (2024), the ability to engage in multimodal composition platforms is increasingly essential for EFL students, as it enhances both their writing abilities and digital competencies, which are crucial for success in the modern job market. In this light, the advancement of AI fosters immersive, context-rich, and interactive learning experiences that closely align with real-world communication demands and future employability goals (Aditya et al., 2024; Mubarok et al., 2024). Unfortunately, the neglect of this domain is particularly striking in teacher education programs in non-Western contexts, which tend to be constrained by rigid curricula, large classes, and limited time beyond the use of pre-approved coursebooks (Nghia et al., 2023). Existing ESL writing instruction tends to prioritize academic essays or test preparation, overlooking the practical literacies that students need for workplace readiness (Windsor & Hancock, 2024). Viewed holistically, these developments position AI-IDLE as a potential transformative approach that not only extends the boundaries of informal language learning but also equips EFL students with the multimodal communicative and technological skills necessary for global employability.

Notably, EFL students’ acceptance and willingness to use AI tools, both for classroom-based language learning (Lin & Yu, 2025; Zheng et al., 2024) and for informal engagement beyond class time (G. L. Liu et al., 2024), are inevitable and shaped by a range of factors, and the extent to which students engage with such tools is contingent on instructors’ perceptions and pedagogical decisions (Chun, 2019; Fernández-Batanero et al., 2021; Li & Thien, 2025; Zhi & Wang, 2024). This factor ultimately influences the diverse strategies students adopt when interacting with AI-mediated tools. Teachers’ positive perceptions toward AI integration significantly influence students’ strategies to shape their usage patterns and eagerness to explore these tools for language development purposefully and critically (Ahmadi Fatalaki et al., 2025; Lee, 2020; Pokrivcakova, 2023). Teachers can also boost EFL students’ participation in IDLE by creating and suggesting activities that build their skills and promote a sense of belonging, especially in online spaces (Zadorozhnyy & Lee, 2025). Conversely, in instructional contexts where AI tools are discouraged or viewed with skepticism, students may hesitate to adopt them or may use them in ways that lack pedagogical direction (Kalra, 2024; Zhi & Wang, 2024).

What must be noted is that despite growing interest in AI applications in EFL settings, current research has predominantly focused on students’ use of traditional, non-AI-powered IDLE for vocabulary and reading (Gowasa et al., 2019; Zhao & Nazir, 2022), or on early AI-IDLE interventions targeting oral communication skills (Guan et al., 2024; G. L. Liu et al., 2024). However, little attention has been paid to how EFL students, guided by teachers’ perceptions and instructional support, leverage AI-mediated informal digital learning for multimodal technical and academic writing, a skill increasingly essential for professional communication and employability. This imbalance, where current research remains disproportionately focused on vocabulary acquisition and oral communication outcomes (e.g., Guan et al., 2024; G. L. Liu et al., 2024), while career-oriented writing and multimodal professional genres receive far less attention, is particularly significant. Meanwhile, today’s EFL learners must increasingly produce resumes, cover letters, portfolios, and other informational texts that integrate written content with visuals and digital design to meet global employability demands (Aditya et al., 2024; Nghia et al., 2023). This gap is particularly significant as multimodal career writing becomes a key component of 21st-century language education. In other words, career ESL writing denotes a family of informational and professional genres central to employability and early-career teacher preparation, including résumés/CVs, cover letters, teaching statements, lesson plans, reflective statements, email/letter writing, and short informational reports.

Therefore, this qualitative study is one of the first studies seeking to address the gap of limited research on how EFL students leverage AI-mediated informal digital learning for multimodal academic and career-oriented writing, particularly when guided by teachers’ perceptions and pedagogical support by examining both teachers’ perceptions and students’ strategies in utilizing AI-mediated informal digital learning for career ESL writing. The unique contributions of this study are to foreground multimodal writing as a critical skill in AI-IDLE research, to illuminate the role of teachers’ attitudes in shaping students’ engagement strategies, and to propose pedagogically informed pathways for integrating AI tools into informal learning practices in ways that enhance employability and equitable access. In other words, from the perspectives of research conceptualization, this study seeks to elucidate the intersection of instructional perceptions and student strategies in the utilization of AI-IDLE, aiming to provide practical insights into the barriers that impede student engagement with this technology while promoting more effective, equitable, and employment-relevant language learning in the era of artificial intelligence.

2. Literature Review

2.1. Teachers’ Perceptions Towards AI Integration in Informal Digital Learning of English

Over the past two decades, informal digital learning of English (IDLE) has gained recognition as a critical complement to formal classroom instruction, offering students flexible, self-directed opportunities to develop linguistic, communicative, and global competencies (Dressman, 2023; Lee, 2019; Soyoof et al., 2023). In IDLE, it is students—not language teachers—who select the resources they use to engage in authentic, multimodal communication (e.g., mobile apps, web apps, computer software, etc.). This learner-driven approach supports the development of vocabulary, listening skills, and intercultural understanding by allowing students to tailor their learning to real-world contexts and personal interests (Lee & Dressman, 2018). Several key factors influence the success of IDLE, with students’ affective states—such as their psychological responsiveness to teacher perceptions, grit, self-confidence, and motivation—playing a crucial role in shaping and sustaining productive engagement (Lee & Drajati, 2019; Lee et al., 2024; G. L. Liu & Wang, 2024; Rezai et al., 2024). Regarding teacher perceptions, despite its potential, traditional IDLE practices are claimed by teachers to be limited in their capacity to provide individualized feedback, scaffolded learning trajectories, or tailored language models to meet the diverse needs of students, particularly in complex academic or career-oriented writing tasks. As a result, students may engage in repetitive or superficial language activities that lack meaningful interaction, critical thinking, or genre-specific guidance (G. L. Liu & Wang, 2024; Tazhenova et al., 2024).

The recent integration of artificial intelligence into Informal Digital Learning of English (IDLE)—increasingly referred to as AI-IDLE—offers new possibilities for enhancing how English as a Second Language (ESL) learners develop professional writing skills outside the classroom. By leveraging interactive tools such as intelligent chatbots, grammar-enhancement platforms, and multimodal composition generators, AI-IDLE provides adaptive, real-time support that promotes learner autonomy, strategic revision, and metacognitive engagement. These tools are particularly valuable for students engaged in career ESL writing, which involves producing clear, structured, and contextually appropriate texts for professional purposes. In the context of this study, such tasks include developing teaching portfolios, a key form of informational writing that requires factual, purpose-driven communication tailored to academic or workplace standards. Informational texts—common in resumes, cover letters, reports, and instructional materials—are nonfictional genres designed to convey content clearly and efficiently to a target audience (Boyer, 2017). As AI-IDLE expands access to genre-specific input and supports iterative drafting processes, it signals a shift in how informal digital environments can be harnessed to build sophisticated writing skills aligned with real-world career demands.

In such context, similarly as IDLE, teachers’ perceptions toward AI-IDLE influence how students perceive and use these tools: positive perceptions lead to guided, reflective usage, whereas skepticism or lack of training may result in unstructured or even counterproductive application (Ahmadi Fatalaki et al., 2025; D. R. Cotton et al., 2024; Kalra, 2024). In other words, rooted in Self-Determination Theory (SDT, Ryan & Deci, 2000), learners’ motivation to engage in informal digital learning of English (IDLE) and AI-mediated tools (AI-IDLE) is shaped by the fulfillment of three basic psychological needs: autonomy, competence, and relatedness.

Among these, relatedness, or students’ sense of connection with teachers and peers, plays a particularly salient role in bridging formal instruction and informal exploration. When teachers demonstrate openness, support, and familiarity with AI tools, they help cultivate students’ perceived competence and willingness to explore these tools beyond the classroom (Zadorozhnyy & Lee, 2025; Zhang & Liu, 2022). Thus, the social and emotional climate created by teachers becomes central to students’ engagement with technology-mediated language learning. For instance, when it comes to productive strategies, such as generating texts, interacting with AI writing assistants, or composing multimodal content, students are also more likely to implement these advanced, output-oriented strategies when they also feel a strong sense of relatedness or social connection with the teachers in their formal classroom environments. This sense of belonging appears to be essential in motivating students to extend their classroom learning into informal spaces by applying and experimenting with productive language tasks. In digital settings, relatedness with teacher’s perceptions, such as encouragement, acts as a key moderating psychological factor, amplifying the link between perceived competence and both receptive and productive IDLE engagement (G. L. Liu & Wang, 2024). By focusing on how teachers’ positive or skeptical views of AI influence students’ engagement with career-oriented writing, this study operationalizes SDT’s constructs as both an explanatory and analytical framework, thereby moving beyond surface-level references to motivation and placing psychological needs at the center of interpretation.

2.2. Artificial Intelligence-Powered Tools in Informal Digital Learning of English (AI-IDLE): Students’ Usage Patterns in Writing

AI-IDLE allows students to experiment with language production via writing beyond static input-output models, making language learning more interactive and strategic. For instance, Nazari et al. (2021) found that students using AI writing assistants improved the organization and coherence of their texts in informal learning environments, attributing this progress to iterative revisions enabled by real-time feedback and lexical suggestions. G. L. Liu et al. (2024) reported that many learners favored AI-supported proofreading tools as a primary method for enhancing their English writing skills, highlighting the perceived usefulness of automated feedback in improving grammatical accuracy, word choice, and sentence structure.

In another action research study, L2 English students used Summary Writing-PAL (SW-PAL), an AI-driven writing strategy tutor that Chew et al. (2019) developed to support learners in improving their summarization skills through guided strategy activation. Grounded in the activation design principle, which emphasizes prompting learners to recall prior knowledge and apply cognitive strategies during learning, SW-PAL encouraged students to engage in pre-writing planning, identify main ideas, and structure content coherently before composing. Rather than offering corrective feedback alone, the system promoted metacognitive engagement by guiding students to reflect on their writing choices and monitor their strategic decisions. Through these interactions, learners gradually internalized key academic writing strategies, including paraphrasing and organizing hierarchical information, and began to transfer these skills to other academic tasks. SW-PAL illustrates how AI tools designed with pedagogical principles can serve not just as automated assistants, but as catalysts for strategic, self-regulated learning in informal digital environments. Students’ writing behaviors also included using ChatGPT4 to generate model texts, revise sentence structures, and brainstorm topic ideas—strategies that emphasized efficiency and linguistic scaffolding. Learners reported that the tool facilitated smoother drafting processes by automating certain aspects of composition, such as lexical selection and syntactic organization. Many engaged in iterative interactions, feeding revised prompts and incorporating AI suggestions selectively, demonstrating a recursive writing process aligned with process-oriented pedagogies.

Furthermore, emerging studies have begun to document how AI-IDLE enables students to experiment with multimodal and context-specific writing practices. G. L. Liu et al. (2024), in a study grounded in sociocognitive writing models, showed that learners developed greater metacognitive awareness through leisure-related digital practices such as online gaming, YouTube video watching, or fan fiction writing. That being said, although emerging studies have begun to map out the benefits and challenges of AI-IDLE in multimodal writing, much of the literature still focuses on its utility for vocabulary learning or general writing enhancement. Less attention has been paid to how learners use AI-IDLE for specific academic genres, such as career-oriented writing, or how it shapes their ability to adapt content for real-world multimodal communicative purposes. Furthermore, at the same time, scholars caution that the use of AI in informal learning raises ethical concerns, particularly around data privacy and AI bias. Students’ personal information may be collected or stored without transparent safeguards, while biased training data could inadvertently reinforce stereotypes or inequitable language practices (Dai et al., 2025). Addressing these ethical issues is essential to ensure that AI-IDLE fosters not only effective but also responsible and inclusive learning environments.

3. Research Questions

To fulfill the purpose of the study, the study sought to answer the following research questions:

• What are teachers’ perceptions regarding the integration of AI-mediated informal digital learning for career ESL writing?
• What strategies do ESL students employ when using AI-mediated informal digital learning tools to support their career ESL writing?

4. Methods

4.1. Pedagogical Setting & Participants

The study involved six teacher participants and their respective classes, totaling approximately 300 students (with 50 students per class) from a teacher education university in Vietnam. All students were enrolled in academic programs where English was used as the medium of instruction. Participating teachers and their students were selected through purposive sampling at a conference for public universities in the North of Vietnam about AI implementations in English teaching to align with the study’s objectives. Based on the placement exams administered at the beginning of the academic year, all student participants demonstrated an intermediate level of English proficiency. Students were not individually selected but included as intact class groups to reflect authentic learning contexts.

In line with Self-Determination Theory (Ryan & Deci, 2000), the study design specifically sought to examine how teachers’ perceptions influenced students’ motivation across three constructs: autonomy, competence, and relatedness. Students were ensured to have sufficient language ability to exercise autonomy in tool choice, reflect on their competence in AI-supported writing tasks, and articulate the ways in which teacher encouragement or skepticism shaped their sense of relatedness in extending classroom learning into informal spaces. Thus, by focusing on students at the intermediate proficiency level, the study controlled for extreme variation in English skills, strengthening the reliability of cross-class comparisons.

Additionally, the student sample represented a diverse population within the teacher education program. Overall, the sample comprised 64% female and 36% male students. The age distribution reflected typical undergraduate enrollment patterns, with 60% of students aged 18–20 (Year 1) and 40% aged 21–22 (Year 2). Students represented a range of academic tracks within teacher education, covering multiple tracks, most in English education, followed by primary education, educational leadership, biology, chemistry, and mathematics teacher education. To further contextualize the participants, a demographic questionnaire was administered at the outset of the study to capture relevant background information such as academic major, familiarity with technology integration, and prior exposure to AI-assisted tools.

In this study, as for teachers, all teacher participants were female, while the student participants represented a mix of genders. Overall, the age distribution reflected typical undergraduate enrollment: 46% of students were between 18 and 20 years old (Year 1), and 38% were aged 21–22 (Year 2). To contextualize the participants’ backgrounds, a demographic questionnaire was administered at the outset of the study, capturing relevant information such as teaching experience, familiarity with technology integration, and institutional context (see

Table 1. Teachers’ demographic information.

Name	Age	Years of Teaching Experience	Self-Reported Experience with Technology Integration in Formal Teaching
Ms. Minh	34	10	High—Regularly integrates tools like ChatGPT, Grammarly, and Canva into multimodal writing instruction.
Ms. Hoa	38	14	Moderate—Previously used Google Drive and presentation tools; recently adopted AI tools for higher engagement.
Ms. Ha	41	18	High—Uses AI extensively to scaffold writing development outside class; encourages exploration of linguistic variety.
Ms. Binh	33	9	Moderate—Promotes AI use for real-world writing tasks; values first-language integration for voice preservation.
Ms. Thi	40	17	Low—Hesitant about new technologies; concerned about overreliance and authenticity in student writing.
Ms. Huong	36	12	Low—Cautious about AI’s effects on motivation; notes distractions and unequal access as key barriers.

).

Instructors were selected through purposive sampling based on their (a) teaching of core English-medium courses in the teacher education program, (b) willingness to integrate digital tools into their teaching practices, and (c) prior experience of at least three years in tertiary-level instruction. Students were drawn from these instructors’ intact classes. The student selection criterion was that they were enrolled in English-medium programs and had completed institutional placement exams at the beginning of the academic year, which confirmed an intermediate level of English proficiency (equivalent to B1–B2 on the CEFR). This ensured that students possessed sufficient proficiency to engage meaningfully with AI-mediated informal digital learning while still demonstrating a need for ongoing academic English development. The number of six instructors (with pseudo names) below was chosen because it balanced depth and breadth of analysis. This sample size allowed for meaningful cross-case comparison while remaining feasible for in-depth qualitative analysis of classroom practices, interviews, and student outcomes. With approximately 50 students per class (about 300 in total), this design also ensured a large and diverse student base while keeping the scope of data collection manageable. Prior studies on teacher perceptions in IDLE/AI-IDLE contexts have similarly used small to medium samples of instructors to enable detailed exploration of pedagogical perspectives (e.g., Lee & Drajati, 2019; Soyoof et al., 2023).

4.2. Design of the Study & Data Collection

This study employed a qualitative descriptive case study design to investigate the strategies employed by Vietnamese undergraduate EFL students and the perceptions of their instructors regarding the integration of AI-mediated informal digital learning of English (AI-IDLE) into career writing. The case study approach was chosen for its capacity to provide rich, contextualized insights into how learners and educators engage with real-world writing tasks enhanced by emerging digital tools (Bovermann et al., 2018).

Specifically, the study was situated around a teaching portfolio assignment, which served as a unifying home task across all classes. As pre-service K–12 teachers in various disciplines, students were asked to complete their portfolios in English to reflect on their future professional roles, and were given autonomy to use AI-powered tools during their informal digital learning to assist with writing and design.

To capture the cognitive processes and decision-making strategies used by students in this authentic task, think-aloud protocols (TAPs) served as the primary method of data collection. TAPs were selected for their ability to provide a real-time window into learners’ strategic thinking—offering more accurate and detailed representations than retrospective self-reports (D. Cotton & Gresty, 2006). Prior to the task, student participants were briefed on the TAP procedure, including a demonstration of how to verbalize their thoughts continuously and explicitly while composing. This orientation emphasized that all thoughts—whether linguistic, procedural, or reflective—should be spoken aloud, and students practiced the process before recording their final TAP.

Students then independently completed their teaching portfolio drafts using AI-IDLE tools of their choice at home (e.g., ChatGPT, Grammarly, Canva with Magic Write, Invideo) while verbalizing their thoughts, recording them, and submitting them to the researchers. They were instructed to conduct these sessions outside the classroom to mirror their natural informal learning environment. All participants recorded their TAP sessions on their personal devices, and sessions ranged from 20 to 35 min. Finally, they were asked to complete a questionnaire on the factors that affect their use of AI-IDLE strategies.

Following the TAP sessions, semi-structured interviews were conducted with individual students’ teachers to reflect, clarify, and expand on the cognitive data captured in the TAPs. These interviews explored teachers’ perceptions on the use of specific AI tools in relation to the writing task—whether they support or oppose to the students’ usage patterns with AI. These interviews also sought to understand instructors’ perceptions toward AI-IDLE integration and its perceived pedagogical value, authenticity, and challenges in supporting students’ professional writing development.

4.3. Data Analysis

Transcripts from TAPs and teachers’ interviews were read and re-read multiple times to immerse the researcher in the content and to develop an initial sense of patterns related to students’ strategy use and teachers’ perceptions toward AI-IDLE in career ESL writing. Then, a systematic and multi-stage coding approach was then employed to ensure analytical depth and consistency (McMillan, 2009). To enhance trustworthiness, coding was conducted in two cycles: (1) open coding for initial theme generation, and (2) axial coding to refine categories and explore relationships between themes. NVivo software was used to manage and organize the data systematically. Triangulation with demographic data was employed to ensure a more robust interpretation of findings.

Specifically, in the first round of coding, meaningful excerpts from the transcripts, particularly those reflecting students’ strategies, digital tool choices, and genre awareness, were identified and assigned descriptive codes (Appendix B). Descriptive codes align with SDT’s three constructs: autonomy, competence, and relatedness, with particular attention to more elaborated descriptive codes added by the researchers about how students described their autonomy in choosing AI tools, their perceived competence in managing AI-generated drafts, and their relatedness when teacher attitudes shaped their willingness to extend learning beyond the classroom. In other words, these codes altogether captured specific aspects of AI-IDLE use, writing processes, and instructional perspectives relevant to the research questions.

Coding reliability was strengthened through collaborative analysis: a second coder independently reviewed a subset of transcripts, and the two coders compared results to ensure alignment in code application. Once all data were coded, the codes were grouped into broader thematic categories that reflected recurring patterns across participants. These emergent themes were initially proposed by the lead researcher and then reviewed by the co-coder for accuracy and conceptual clarity. To ensure these themes were grounded in the data, they were checked against the original transcripts to verify coherence and representation of participants’ perspectives. This rigorous verification process helped preserve the nuance of participants’ voices in the presentation of results. Furthermore, descriptive statistics were used to analyze the questionnaire responses regarding students’ reflections on the factors influencing their AI-IDLE strategy use. These quantitative results were then triangulated with the qualitative findings to enrich the interpretation and highlight convergence or divergence across data sources.

5. Results

Research Question 1: 

Variation in teachers’ perceptions towards the use of AI Informal Digital Learning of English to enhance ESL technical career writing.

The analysis of interviews with six experienced EFL instructors reveals a nuanced landscape of perceptions, both positive and negative, toward the integration of AI-mediated informal digital learning of English (AI-IDLE) in the development of career writing for pre-service K–12 teachers. These findings elucidate how teachers’ perceptions shape students’ strategies for utilizing AI-IDLE in writing, specifically in the creation of multimodal, audience-targeted informational texts—a genre frequently neglected in conventional ESL instruction, yet crucial for effective communication in employment and teacher self-assessment.

5.1. Positive Perceptions

1. AI-IDLE as a Source for Novelty and Engagement: All teachers consistently described how AI-IDLE had helped students by providing a sense of novelty and excitement into their English language learning beyond the classroom, as it emphasizes student-centered pedagogy, diverging from exam-focused writing methodologies prevalent in conventional EFL settings. Several notable examples illustrate teachers’ understanding and recognition of the excitement that AI-IDLE generated. For instance, Ms. Minh observed that incorporating AI-IDLE tools like ChatGPT and Grammarly gave students a sense of agency in their writing:

“They’re used to memorizing model answers or model texts, but with AI, they can start fresh off with focusing on what they want to express. Because students are not limited by the class time, they can play around with the originality of their input (prompt), which, at first, can be fragmented and rudimentary. Then, ChatGPT can put those ideas altogether into a coherent flow in one language for the target audience. The output is also shaped by the language’s writing norms and rhetorical conventions, which students can learn how different it is from their native language. It’s more motivating because they are constantly a part of the process and see their own words put into action, not just following the teacher.”

This shift, from passive recipient to active communicator, mirrors the core goals of ESL career writing, where the writer’s task is to structure and adapt factual content to meet the needs of specific audiences. For pre-service teachers, this means learning to express complex ideas in clear, accessible language that resonates with their future employers and cohorts of students—a skill that teachers like Minh see as well-supported by AI-IDLE’s capacity for individualized instant feedback and text generation, ranging from sentence structure to idea delivery, clarity, and tone.

Ms. Hoa highlighted the difference in the level of engagement between AI-IDLE and traditional IDLE,

“Unlike other online tools students might use at home, AI tools offer a higher level of engagement and personalization that makes them particularly powerful for this kind of informational writing. In the past, I used technology like Google Drive to co-author students’ writing outlines and help them formulate an oral presentation about their topic, where they could integrate URL links to video or a photo to illustrate what they were saying. But the problem was that these tools didn’t give them structured models for clear writing or advice or recommendations to connect that writing to visuals. Now, AI can provide even more personalized support almost instantly, suggesting ideas and resources to blend these elements—text, images, and video—into a cohesive multimodal project. This makes students feel like AI is interested in their personal work.”

Even in challenging group interactions or disagreements, students turn to AI for scaffolded suggestions, which in turn boosts their creative thinking and problem-solving skills. Ms. Minh emphasized that encouraging students to ask questions related to task completion in addition to linguistic input results in students being more confident using AI-IDLE to enhance (not replace) their writing, “When they see how AI helps them combine images, videos, and clear English explanations and solve conflicts in groupwork, they get more confident. They persist, not giving up on the task despite the challenges they might face. It’s not just words—it’s about how they see themselves as problem solvers, sharing their ideas.”

2. AI-IDLE as a Versatile Language Tutor: Another difference is that AI-IDLE’s model texts offer students spontaneous and flexible examples that go beyond the formulaic paragraphs found in many textbooks or other non-AI digital tools where fixed texts were simply collected and archived. According to Ms. Ha, “AI can show them different ways to explain the same idea. It’s not about memorizing a perfect script—it’s about conversing with AI to explore and play around with different word choices, phrases, and tones to figure out how to communicate clearly and effectively. This is particularly valuable for outside class time, as it is impractical to provide individualized linguistic input to every student in a large class setting with mixed levels of English proficiency.” In this vein, several teachers described AI-IDLE as functioning like an individualized expert, providing real-time, scaffolded feedback tailored to each learner’s style. Ms. Ha shared, “It matched the student’s own choices in words and grammar, so it felt more like a personal tutor than just a machine.”

In addition, almost all teachers also noted how AI-IDLE can raise students’ concentration on writing by providing them with an opportunity to pay conscious attention to specific features—such as grammar, vocabulary, pronunciation, etc.—to process and internalize them. As Ms. Hoa explained, the sentences and paragraphs generated by tools like ChatGPT and Quillbot are incredibly fast and well-structured—often above students’ current writing level, yet still within their zone of proximal development. Otherwise, they will fail to understand it and attempt once more. In other words, if students do not pay attention to the linguistic elements of the text AI generates and simply copy it, they know that the instructor may question the originality of the learning product. Ms. Binh also added that students may remember their experience with AI more vividly than traditional methods, as when they are unsure how to express ideas in English, they use their first language and receive immediate translations—thereby preserving their unique voice. This process not only strengthens their acquisitions of English structures during the task but also helps them retain and recall these structures in both languages.

Half of the interviewed teachers also noted that AI-IDLE can be a powerful tool to help students learn how to write clearer, more organized paragraphs and sentences—rather than just using it to skip the writing process. Four out of six teachers agreed with Ms. Ha’s view that students will not stop writing because of AI tools; instead, she sees writing as having two important phases. In the classroom, teachers support students in building foundational language skills—like using correct grammar and organizing their thoughts effectively. That is the first phase.

Outside of class, the second phase begins. Students should be encouraged to experiment, reflect, and revise on their own—using AI-IDLE suggestions as a metacognitive scaffold to monitor their choices and refine their thinking. Ms. Ha emphasized the importance of teachers being explicit about the exact necessary strategies students can use to complete the task successfully, such as demonstrating effective prompts for engaging with ChatGPT, while avoiding limiting the apps or tools students can use. Agreeing with her, many teachers acknowledged that AI tools help students consider how their writing might be received by real-world audiences, such as future employers. Tools like ChatGPT, Grammarly, and Canva’s Magic Write provide concrete examples of how the same message can be expressed in different tones, levels of formality, or formats. By generating models tailored to specific prompts like “write a professional description of duties in a formal cover letter to a school principal” or “create a friendly teaching portfolio introduction,” AI helps students see how language changes depending on audience and purpose.

All teachers noted that career writing and employability skills are often neglected in ESL classrooms, where the focus tends to be on academic essays and test preparation. Thus, Ms. Binh added, “AI-IDLE has emerged as a powerful way to bridge this gap by providing students with models of clear, audience-aware writing that directly relates to real-world teaching tasks. I encourage my students to use AI-IDLE because I can better support them in developing the kind of communicative flexibility and professional confidence they will need as professionals—ultimately making ESL writing instruction more relevant and impactful for students’ future careers.”

5.2. Negative Perceptions

1. AI-IDLE Overreliance and Loss of Authentic Learning: Despite these positive views, many teachers voiced significant concerns about overreliance, creativity gaps, and pedagogical fit—factors that have long complicated AI-IDLE. For instance, Ms. Thi, explained, “I can be too cautious to say this, but I cannot control how students utilize these tools. Students will just copy.” However, she admitted having anxiety about new technologies, which can act as her burden, hindering her and her students’ efforts to implement innovative practices even when benefits might be recognized. Another teacher, Ms. Huong, noted that students become very distracted when using AI. According to her, excessively refined AI-generated writing produced rapidly may result in disengaged and idle students who have difficulty seeing the worth of studying. They might skip handouts and instead use these technologies to copy others’ sentences as their own. Ms. Huong expressed concern that students might lose the sense of self-discovery when every question seems to have an answer already provided by AI.

2. Equity and Access Concerns: A further concern voiced by Ms. Thi and Ms Huong is the risk that AI-IDLE might inadvertently widen existing gaps between students from different backgrounds. Those university students from urban areas or with better internet access can explore and experiment with these tools more frequently through informal learning experience, gaining familiarity with AI’s capabilities and building greater confidence in their writing. In contrast, students from rural areas or those with limited digital resources may find it much harder to participate fully, reinforcing pre-existing educational divides. Teachers might not be able to equip all students with strategies to implement AI-IDLE since the curriculum within the classroom is still heavy with expectations about linguistic knowledge transmission.

Indeed,

Table 2. Top 5 students’ reflections on how teachers’ perceptions affect their use AI-IDLE for ESL career writing.

		n	M	S.D.
1	My teacher’s positive attitude towards using AI motivates me to experiment with different prompts and explore new ways of learning.	300	4.55	1.20
2	Because my teacher sees AI as a helpful support, I feel more confident using it as a tool to revise my own writing.	300	4.10	1.05
3	My teacher encourages me to see AI suggestions as a starting point, not the final product, which helps me develop my ideas independently.	300	4.80	0.85
4	When my teacher shares their own experiences using AI, I feel encouraged to try AI-powered strategies that improve my work.	300	3.95	1.25
5	When my teacher expresses skepticism about AI writing, I become more mindful of how I use it, making sure it doesn’t replace my own thinking.	300	3.75	1.15

below provides nuanced insights into how teachers’ perceptions influence students’ use of AI during informal digital learning (IDLE). For the descriptive statistics, the assumptions for normality for the data were checked using skewness and kurtosis values, as well as the Shapiro–Wilk test, and no major violations were detected. Overall, the data suggest that students’ engagement with AI tools is strongly mediated by their teachers’ perceptions, both positive and skeptical, toward the integration of AI in the learning process.

Firstly, the highest mean score (M = 4.80, SD = 0.85) was for the statement that teachers’ explicit encouragement to use AI suggestions as a starting point, rather than as final answers, promotes greater student autonomy in idea development. This finding underscores the importance of teachers positioning AI as a tool for exploration and enhancement, rather than as a substitute for student agency. Similarly, the statement highlighting the motivating impact of teachers’ positive perceptions toward AI use also received a relatively high mean (M = 4.55, SD = 1.20) although the slightly larger standard deviation for this statement suggests that there is considerable variability among students in how strongly they are motivated by their teachers’ perceptions.

Interestingly, the data reveal a more mixed picture regarding the impact of teachers’ openness and personal experiences with AI. The statement “When my teacher shares their own experiences using AI, I feel encouraged to try AI-powered strategies that improve my work” had a mean of 3.95 (SD = 1.25), indicating a moderate but less uniform effect. This suggests that although students generally respond well to teachers’ modeling of AI use, not all students view this as a primary source of motivation or strategy adoption. The statement with the lowest mean (M = 3.75, SD = 1.15) was related to teachers’ skepticism toward AI writing. While the mean indicates a generally positive inclination—students become more mindful and cautious in using AI—this finding also highlights a degree of ambivalence among students when teachers discuss diverse AI views or express their concerns about AI use.

Research Question 2: 

Variation in students’ strategies towards the use of AI Informal Digital Learning of English to enhance ESL technical career writing.

Table 3. Students’ reflections on their top 10 strategies to use AI-IDLE for ESL career writing.

		n	M	S.D.
1	I use AI-IDLE after each class since its information is informative and tailored to my personal searches, enriching my English writing and career orientation.	300	4.85	0.95
2	Thanks to the writing suggestion from AI, I use it to reorganize my ideas in a way that feels more natural and logically consistent with the genre.	300	4.95	0.92
3	I use AI to analyze and evaluate the quality of my draft’s structure and see how ideas connect more clearly, which makes my writing more coherent and easier to follow.	300	4.80	1.00
4	I use AI to improve linguistic elements, including, but not exclusively, vocabulary, spellings, grammar, and punctuation.	300	4.75	1.05
5	I actively explore different prompts and questions in AI to explore and refine how I explain my teaching duties and future roles as an educator.	300	4.78	0.97
6	I use AI-generated outlines as a starting point, but I make sure to adapt and personalize the best outline to fit my own teaching context and voice.	300	4.92	0.94
7	I google, watch tutorials, or ask AI for guidance on how to integrate visuals, audio, and videos suggested by AI to create multimodal presentations that clearly convey my teaching ideas and personal reflections.	300	4.85	0.95
8	I review AI’s sentence and paragraph structures carefully to understand how to logically sequence information for different audiences, such as students or colleagues.	300	4.75	1.02
9	I treat AI suggestions as a learning tool, not a final product, by revising them to ensure they align with my own understanding and the specific requirements of informational texts.	300	4.80	0.98
10	I seek feedback from teachers and peers on how I use AI-IDLE outputs to ensure that my writing stays authentic, accurate, and relevant to the assigned tasks.	300	4.77	1.01

below presents descriptive statistics summarizing how students use AI-IDLE as part of their writing strategies. The mean scores, all of which are above 4.70 on a 5-point scale, highlight a strong tendency to use and leverage AI-IDLE in various aspects of their writing development. Nonetheless, the standard deviations (ranging from 0.92 to 1.05) suggest that while most students view AI-IDLE as a valuable tool for improving their career writing, there is noticeable variation in how consistently they actually apply these strategies in practice.

The highest mean score (M = 4.95, S.D. = 0.92, Item 2) was for students’ use of AI suggestions to reorganize ideas in a way that feels more natural and genre-appropriate. This suggests that rather than merely correcting surface-level errors, students appear to know to use AI-IDLE’s outputs as exemplars of rhetorical and structural coherence, a critical dimension in crafting genre specific texts that must be both accessible to a target audience. In addition, the high mean (M = 4.92, Item 6) and low standard deviation (S.D. = 0.94) for using AI-generated outlines and suggestions as starting points indicates a shared strategic approach: students value AI-IDLE’s capacity to provide a foundational structure, which they then adapt to meet specific contexts, yet they are aware that while AI can offer a draft, the final product must be shaped by their own voice and pedagogical goals. Interestingly, while most strategies scored highly, items relating to seeking teacher feedback (M = 4.77, Item 10) and reviewing AI’s linguistic structures (M = 4.75, Item 4) had slightly lower means. This may indicate that while students are enthusiastic about the autonomy and experimentation that AI-IDLE affords, they still recognize the importance of external validation, human judgment, and linguistic accuracy. This dual reliance—on both AI’s affordances and traditional teacher perceptions and feedback—reflects a hybrid learning ecosystem that blends informal digital learning with formal pedagogical oversight.

Specifically, various AI-powered tools that university EFL students used for creating multimodal informational genre texts outside of the classroom emerged mainly with three categories of linguistic tools, audio/auditorial tools, and visual tools: (1) students had an explicit genre understanding throughout their informal digital learning, (2) students selected digital multimodal tools intentionally and diversely—such as Grammarly, ChatGPT, Canva with Magic Write, and Invideo—to integrate text with images, sound, and layout for improved rhetorical accessibility, and (3) students knew how to align with their post-graduation career needs.

Students’ work with these AI tools demonstrated that when they created projects aligned with professional identities and future job needs, they became more aware of how to improve their writing. For instance, a student shared,

“I didn’t take Grammarly’s first suggestion. I knew that a factual recount represents a form of informational text that narrates real events in a chronological sequence, so I tried rewriting my teaching descriptions five or six times using different versions it suggested. Then I compared it with the guiding template of the genre my teacher gave to see if AI aligns with that structure of orientation, records of events, and re-orientation.”.

(Participant 5)

Another participant emphasized, “I used Wordtune to see at least three different phrasings for the same sentence. I asked it about the tone each word or phrase convey, whether it is too formal or casual. I kept rewriting the word in sentence context until it felt clear and natural” (Participant 8). This iterative rewriting cycle, supported by AI tools like Grammarly and Wordtune, reveals how students intentionally refined their word choices in texts to meet genre-specific expectations and audience needs, ensuring that their multimodal projects aligned with the clear, factual, and structured style typical of informational text. This process extended and reinforced their formal classroom learning of genre conventions.

Moreover, students intentionally and diversely selected digital multimodal tools to integrate text with images, sound, and layout for improved rhetorical accessibility. Notably, the ability of AI tools like Canva’s Magic Write and other text-to-image generators to create relevant visuals from textual descriptions significantly influenced students’ writing practices. Students described how seeing these AI-generated images allowed them to ensure that their text and visuals worked together to present clear and coherent information. One student commented, “After using Magic Write in Canva, I visualized each sentence as a picture. Seeing how the AI turned my words into images helped me notice where my descriptions were too vague or not matching the picture. I rewrote those parts to make them clearer. This part is time-consuming, so informal digital learning is necessary to explore creativity.” (Participant 6).

A third prominent theme in the findings revealed that students were highly attuned to aligning their IDLE with their post-graduation career needs as future K–12 teachers. AI-powered IDLE tools—such as Grammarly, ChatGPT, Canva with Magic Write, and Murf.ai—filled this gap by generating rewriting suggestions and automated voiceovers that also contain explicit precise information about the topic, or the career duties in this case, about which students inquired. These tools’ feedback encouraged students to elevate their English not just for accuracy but for communicative effectiveness, resonating with students’ desire to develop employability-oriented writing abilities that empower them to use English as a tool for global communication and professional growth.

6. Discussions

The findings of this study shed light on the complex interplay between teachers’ perceptions and students’ strategies in the use of AI-IDLE for career ESL writing. Both research questions, teachers’ perceptions and students’ strategic practices highlight how AI-IDLE is reshaping the pedagogical landscape, not by replacing classroom instruction, but by amplifying learners’ autonomy, creativity, and awareness of genre-specific writing in the digital age. These results contribute to current scholarship on digital literacy, AI integration in language learning, and the sociocognitive dimensions of ESL instruction in informal settings. The Discussion mirrors the Results (RQ1 → RQ2 → convergence), keeping interpretations anchored in the adopted SDT constructs across data sources and emphasizing comparability over novelty within subsections.

First, the study confirms that teachers’ perceptions of AI-mediated IDLE significantly influence students’ engagement with these tools, highlighting an evolving role for ESL educators in the AI era, one that shifts the focus toward fostering learner autonomy and self-regulation (cf. Carciu & Muresan, 2024; Rezai et al., 2024; Zadorozhnyy & Lee, 2025). It means that teachers no longer serve solely as transmitters of linguistic knowledge but act as facilitators of critical engagement with AI tools. Participants emphasized that their encouragement to use AI as a scaffold, not a shortcut, enabled students to become more confident in consolidating their knowledge about the genre in class as well as enhancing their writing skills for professional communication.

Nonetheless, their perceptions towards the use of AI-IDLE are complex because they range from optimism to anxiety. These varied attitudes, ranging from optimism to caution, mirrored broader concerns in the literature about digital equity and the readiness of both educators and learners to adapt to emerging tools (Fernández-Batanero et al., 2021; Rezai et al., 2024). In a general sense, the classrooms where teachers embrace AI as a tool for analytical modeling, multimodal experimentation, and reflective practice, students are more likely to treat AI outputs as dynamic drafts, starting points for further adaptation and contextualization, particularly in describing and reflecting on the informational genre with the topic regarding their future teaching duties and scenarios. In this light, AI-IDLE emerges as both a scaffold, offering structured guidance and analytical models, and an empowering resource, building students’ confidence and creativity in teaching themselves through multimodal informational texts.

Conversely, in classrooms where caution prevails because teachers often struggle to translate students’ informal technology use into effective instructional practice, students may use AI in superficial ways, treating it as a quick fix rather than a means of developing real-world teaching skills. Such a finding aligns with similar findings in several previous studies of IDLE (cf. Kalra, 2024; Lee, 2020; Pöntinen et al., 2017; Zhi & Wang, 2024). However, unlike those studies, the findings of this study also suggest that students do exhibit a notable degree of agency. Although shaped by teachers’ authoritative control and excessive concern, students demonstrate strategic and purposeful use of AI-mediated IDLE to accomplish specific components of their writing projects. Thus, the study highlights a potential tension between teachers’ traditional emphasis on authoritative knowledge—often accompanied by reluctance to adopt tools perceived as potentially harmful—and students’ actual patterns of use, which are marked by agency and purposeful engagement. This underscores the need to prepare both current and future educators not only to develop engaging instructional strategies that meaningfully integrate AI tools within informal digital learning contexts, but also to critically reflect on their own perceptions and embrace their role as agents of pedagogical change in the evolving landscape of education (Baek et al., 2024).

Second, the study reveals that students employed AI-IDLE tools in ways that reflect an emerging understanding of informational writing genres and multimodal composition related to their future career demands. Rather than relying solely on traditional linear models of learning, students are increasingly engaging in multimodal, recursive, and collaborative processes with AI tools, positioning them as integral components of participatory learning. Specifically, students themselves recognize this inherent tension—on one hand, they value the efficiency, accessibility, and linguistic support that AI-mediated IDLE provides; on the other, they are aware of its limitations and the necessity to critically engage with, revise, and personalize the generated content. This process of validation is essential to ensure that AI outputs not only meet academic standards but also align with the pedagogical and communicative demands of real-world contexts. In doing so, students begin to position themselves not merely as passive recipients of AI assistance, but as reflective practitioners who exercise agency in shaping and contextualizing their learning materials. Similar awareness has also been found in recent studies, such as G. L. Liu et al. (2024), Guan et al. (2024), Carciu and Muresan (2024), and Zadorozhnyy and Lee (2025), yet the originality of this study’s findings lies in its focus on demonstrating a unique blend of learner agency, genre awareness, and career readiness purpose.

From the conceptual lens of Self-Determination Theory, our findings highlight how teacher encouragement supported students’ autonomy (using AI suggestions as starting points rather than final answers), strengthened their sense of competence (building confidence in drafting multimodal texts, Appendix A), and fostered relatedness (creating a sense of validation and belonging when teachers modeled openness to AI use). Conversely, when teachers voiced skepticism or anxiety, students reported diminished confidence and a weaker sense of relatedness, which in some cases hindered their willingness to extend AI-supported writing beyond the classroom. These dynamics demonstrate that the fulfillment, or thwarting, of SDT’s three motivational needs helps explain variability in students’ engagement with AI-IDLE.

Third, while urban and digitally fluent students are able to experiment with and refine their writing extensively through AI-mediated IDLE, those with limited access or weaker English proficiency may struggle to fully capitalize on these tools, corroborating findings by G. Liu and Darvin (2023) and Soyoof et al. (2023). However, the data also suggest teachers observed that students with better internet access or digital literacy were more adept at using AI-IDLE effectively, suggesting that informal learning opportunities are unevenly distributed. Some teachers expressed apprehension about overreliance on AI, highlighting the need for clear guidelines.

Lastly, the study points to a persistent gap in traditional ESL instruction, which often privileges academic essay writing while overlooking the practical literacy demands such as multimodal career writing. Students’ consistently high ratings of AI-IDLE strategies suggest that they perceive these tools as bridging that gap. From a multiliteracies perspective, as explored by Carciu and Muresan (2024), their use of AI-IDLE reflects an emerging capacity to engage in meaning-making across modes, audiences, and contexts, fostering self-directed exploration of professional topics. This aligns with the growing body of research that emphasizes the importance of digital literacy, affective engagement, AI-supported multimodal composition, and importantly, employability for ESL students.

That being said, the limitation of this study also lies in the fact that beyond teacher mediation, variation in AI-IDLE uptake also could be due to individual differences in prior digital literacy and program track. For instance, students reporting higher comfort with digital tools appropriated AI-IDLE more strategically (prompt iteration, genre-aware revision), suggesting a task–technology fit mechanism. Program-track contrasts implied that genre familiarity (e.g., teaching portfolios) scaffolds tool use, whereas less familiar genres required heavier teacher mediation. Such differences as outcomes of socioculturally mediated action, where tools, community norms, and prior experiences shape autonomy and competence, rather than teacher effects alone.

7. Conclusions

This study sheds light on how the interplay between teachers’ attitudes and students’ strategies critically shapes the use of AI-mediated informal digital learning of English (AI-IDLE) in career ESL writing contexts. Findings reveal that while students increasingly experiment with AI tools to meet real-world communicative demands, their ability to do so meaningfully is often influenced by the degree of support—or resistance—shown by their instructors. The study also contributes to a growing body of research advocating for context-sensitive, reflective pedagogy—one that equips future educators with not only AI literacy, but also the critical awareness needed to purposefully integrate AI into their instructional design. AI-mediated informal digital learning should not be viewed as a replacement for robust writing instruction, but rather as a valuable complement to it in informal learning settings.

The challenge ahead lies in promoting equitable access, fostering responsible AI use, and preparing teachers to support this evolving landscape of employability-focused communication. As a limitation, the study draws from a relatively small sample in a specific educational context, which may affect the generalizability of its findings. In addition, this study targets analytic generalization where findings are most transferable to EFL programs with comparable access to digital tools, portfolio-style assessment, and similar informational or professional writing genres; they are more tentative for highly technical disciplines or contexts with limited connectivity or distinct assessment norms. Thus, future research should explore diverse learning environments and longitudinal impacts of AI-IDLE integration on both teaching practices and student outcomes.