Flipped design thinking with generative AI for digital literacy in ODL Open Access

Digital Literacy has become a foundational competency for students in open and distance learning (ODL) systems, particularly within Asian higher education contexts undergoing rapid digital expansion. As learners increasingly interact with digital platforms, devices, and online resources, universities must ensure that students are equipped with the evaluative, productive, and communicative capacities needed to participate effectively in technology-mediated learning (Bahri et al., 2024; Olney et al., 2021). This challenge is especially evident in large-scale ODL institutions such as Universitas Terbuka (UT), where diverse and geographically dispersed learners present varying levels of digital readiness, engagement patterns, and technological access (Berg, 2020; Budiman and Syafrony, 2023; Nayak et al., 2020). While online tutorial systems enable flexible access to course materials, they do not always provide structured opportunities for students to develop higher-order digital practices beyond basic information retrieval and document preparation.

Conventional tutorial approaches in ODL environments tend to promote functional rather than transformative Digital Literacy. These approaches commonly emphasise searching for online materials, navigating the learning management system, or producing linear written work (Mailizar et al., 2022; Rusli et al., 2023; Suwardika et al., 2024). Although these competencies remain essential, research suggests that they are insufficient for cultivating advanced skills such as source evaluation, multimodal synthesis, reflective content creation, and digital communication – skills required by contemporary academic and professional contexts (Bahri et al., 2024; Raymundo, 2020; Tian et al., 2023). Consequently, there is a need for pedagogical models that intentionally support students in developing integrated Digital Literacy capabilities that extend beyond operational ICT skills.

Flipped learning offers a promising foundation for addressing this need. By shifting content acquisition to out-of-class spaces and reserving synchronous time for active, inquiry-led learning, flipped classrooms encourage autonomy, collaboration, and deep engagement with digital materials (Gu et al., 2022; Huang, 2023). Flipped approaches can be particularly powerful in ODL environments, where structured activities are essential for reducing transactional distance and supporting learner agency. When paired with design thinking, flipped learning can be organised around iterative cycles of empathy, ideation, prototyping, and testing, encouraging students to engage meaningfully with digital information and to generate purposeful digital artefacts (Chunpungsuk et al., 2021; Jia et al., 2023; Sriwisathiyakun, 2023). Design thinking thus provides a flexible, student-centred framework for cultivating evaluative, creative, and reflective digital practices.

The rapid development of generative artificial intelligence (AI) has added further pedagogical possibilities. In this study, the generative AI tool used in the FCDT-AI condition was ChatGPT. AI tools can provide adaptive scaffolding, enhance ideational fluency, support multimodal production, and guide learners through complex digital tasks, thereby expanding their capacity to analyse, evaluate, and create digital artefacts (Joseph et al., 2024; Ng et al., 2024b; Zhang and Zhang, 2024). When incorporated into flipped and design-thinking-based learning environments, AI can reduce cognitive load, offer personalised guidance, extend access to examples and feedback, and support diverse learners who may otherwise struggle to engage in higher-order digital tasks (Engeness et al., 2025; Laupichler, 2022). These advantages are particularly relevant in ODL contexts, where students work autonomously and where gaps in preparedness may constrain participation in complex forms of digital inquiry. At the same time, the growing presence of AI in education underscores the importance of ethical, transparent, and responsible pedagogical design, which aligns with current discourse on academic integrity and AI literacy in higher education (Nguyen et al., 2023).

Despite increasing interest in AI-supported learning, empirical work remains limited on how generative AI can be systematically integrated into flipped, design-thinking-based instruction to strengthen Digital Literacy in ODL settings. Existing research tends to examine isolated components of digital literacy development, such as information retrieval or media production, or focuses solely on AI literacy rather than on broader digital capability (Lim et al., 2025; Ng et al., 2024b). Few studies compare different levels of scaffolding across multiple instructional models, and almost none have evaluated AI-enhanced flipped models in Asian open university contexts, where instructional scale, learner heterogeneity, and infrastructure constraints differ substantially from campus-based environments. As a result, there is insufficient empirical evidence on how pedagogical conditions that vary in structure, design thinking integration, and AI support influence the development of Digital Literacy among distance learners. This gap is particularly pressing for institutions such as UT, where large enrolments and diverse learner backgrounds necessitate pedagogical models that are both robust and scalable.

To address this gap, the present study evaluates the effects of three instructional conditions – Conventional Tutorial, Flipped Classroom Design Thinking (FCDT), and an AI-augmented FCDT-AI model – on students' Digital Literacy in an undergraduate Academic Writing Techniques course at UT. Using a within-subjects repeated-measures design, all 26 participants experienced each instructional condition in counterbalanced order via TUWEB, the institutional learning management system. Digital Literacy was assessed through a multidimensional performance-based questionnaire aligned with contemporary frameworks, and qualitative reflection logs were analysed thematically to illuminate the mechanisms underlying observed quantitative differences. By combining statistical analysis with qualitative insights, the study offers a comprehensive examination of how structured scaffolding, design-thinking processes, and AI-enhanced support shape learners' Digital Literacy in an ODL environment.

This research contributes to digital pedagogy in two key ways. First, it provides empirical evidence comparing the relative effects of conventional, flipped, and AI-supported flipped instructional models on Digital Literacy development, thereby offering clarity on the pedagogical value of integrating design thinking and AI into ODL tutorial systems. Second, it introduces a theoretically grounded and scalable framework – FCDT-AI – that reflects emerging priorities among Asian open universities for ethical and effective integration of AI into teaching and learning. By examining how students progress from basic information access to more advanced, multimodal, and reflective digital practices across these conditions, the study addresses a critical need within the region's digital transformation agenda and offers a practical model for strengthening Digital Literacy among diverse, geographically dispersed learners.

This study employed a within-subjects repeated-measures design to examine the effects of three instructional conditions – Conventional Tutorial, Flipped Classroom Design Thinking (FCDT), and an AI-supported FCDT-AI model – on students' Digital Literacy. Specifically, the AI-supported intervention was implemented using ChatGPT as the generative AI platform. A repeated-measures approach was selected to control for inter-individual variability and to enable each participant to experience all three conditions, thereby increasing statistical power and reducing error variance. A mixed-methods strategy was adopted (Creswell and Clark, 2018), integrating quantitative performance data with qualitative reflection logs to provide a holistic account of Digital Literacy development. Quantitative data were analysed using repeated-measures ANOVA, while qualitative data were examined through reflexive thematic analysis.

The study involved 26 undergraduate students enrolled in the Academic Writing Techniques course at Universitas Terbuka (UT) Denpasar. Participants reflected UT's typical ODL learner profile: adults balancing study with work, geographically dispersed locations, and varied levels of digital readiness. Participation was voluntary, no incentives were offered, and students could withdraw at any time without academic penalty. Although demographic characteristics such as age, gender, and occupational background were not collected in detail, all participants provided informed consent, and inclusion criteria required active enrolment in the course and ability to access TUWEB, the university's learning management system.

All instructional activities were delivered asynchronously and synchronously through TUWEB. Participants completed the three instructional conditions in counterbalanced order to minimise order and carryover effects. Each condition spanned one tutorial cycle.

The conditions differed in the level of scaffolding, the integration of design thinking, and the use of generative AI. Table 1 summarises the structural and pedagogical characteristics of the conditions. The Conventional Tutorial reflected UT's standard tutorial model. The FCDT condition incorporated structured design-thinking cycles, supported by Learning Activity Sheets (Lembar Kerja Mahasiswa/LKM). The FCDT-AI condition followed the same structure but integrated AI-mediated scaffolding in ideation, prototyping, and evaluation phases. In this condition, ChatGPT was used as the generative AI platform to support students' ideation, prototyping, feedback interpretation, and evaluation activities. Its use was tutor-guided, prompt-supported, and bounded by ethical instructions to ensure responsible and transparent academic use.

Digital Literacy was operationalised using a comprehensive multidimensional framework aligned with contemporary definitions (Burton et al., 2015; Press et al., 2022; Reid et al., 2023) and defined as the capability to locate, understand, evaluate, produce, and share digital information through devices such as smartphones, tablets, laptops, and desktop computers. The construct comprised six integrated domains adapted from Covello (2010), including Information Literacy (locating, analysing, synthesising, and evaluating digital sources while applying citation ethics), Computer Literacy (operating digital technologies for academic tasks), Media Literacy (accessing, interpreting, and communicating information across digital media), Communication Literacy (interacting and collaborating through the LMS, Microsoft Teams, and other digital platforms), Visual Literacy (reading, interpreting, and producing visual or digital representations), and Technology Literacy (using digital tools to enhance learning, productivity, and performance). The questionnaire employed a five-point Likert scale (1 = strongly disagree to 5 = strongly agree) and was administered after each instructional condition. Content validity was established through expert review by ODL and digital literacy specialists, and internal consistency for the overall scale demonstrated acceptable reliability (Cronbach's α > 0.80).

Following each condition, students completed individual reflection logs responding to open-ended prompts regarding their understanding of TUWEB activities and their perceived development of Digital Literacy. Logs typically ranged from 150 to 250 words and served as qualitative data for triangulating the quantitative findings. These reflection logs were intended to capture students' self-reported perceptions and experiences of Digital Literacy development rather than objectively observed behavioural outcomes.

The study was conducted over three consecutive tutorial cycles, each lasting two weeks. All participants were assigned randomly into counterbalanced groups to complete the three instructional conditions in different sequences, ensuring that no single condition systematically preceded or followed another.

At the end of each cycle, students.

1. completed the Digital Literacy questionnaire; and

2. submitted a written reflection log via TUWEB.

The procedure ensured that data collection for each condition occurred independently and that exposure to one condition did not influence performance in subsequent conditions. A procedural flowchart is provided in Figure 1.

Quantitative analyses were conducted in JASP 0.19.3 following guidelines for repeated-measures designs (Goss-Sampson, 2022). Assumptions were checked using Q–Q plots (normality) and Mauchly's test (sphericity). Because sphericity was violated, the Huynh–Feldt correction was applied. A repeated-measures ANOVA assessed differences in Digital Literacy across the three conditions, with partial η² reported as the effect-size estimator. Holm-adjusted post-hoc comparisons were used to identify pairwise differences. Raincloud plots visualised the distribution and variability of scores.

Reflection logs were analysed using reflexive thematic analysis (Braun and Clarke, 2006). The analysis followed six phases: familiarisation, initial coding, theme development, theme review, theme definition, and reporting. Coding was primarily inductive, focussing on students' descriptions of their Digital Literacy practices. Reflexive memos were used to document analytic decisions and enhance transparency. The qualitative analysis was used to interpret perceived mechanisms and reported practices underlying the quantitative patterns, rather than to establish directly observed behavioural outcomes.

Digital Literacy scores were collected under three instructional conditions delivered via TUWEB: Conventional Tutorial, Flipped Classroom Design Thinking (FCDT), and the AI-supported FCDT-AI model. Visual inspection of the Raincloud plots (Figure 2) indicated a clear upward progression in both central tendency and score distribution across the conditions. The Conventional condition showed the lowest and narrowest distribution, the FCDT condition displayed higher and more varied scores, and the FCDT-AI condition produced the highest concentration of scores in the upper range. These patterns provided an initial indication of meaningful differences across the instructional models.

A Q–Q plot of residuals (Figure 3) demonstrated acceptable alignment with the theoretical normal distribution, with only slight deviations in the extreme tails, supporting the suitability of the repeated-measures ANOVA. Mauchly's test indicated a violation of sphericity, W = 0.777, χ²(2) = 6.07, p = 0.048, and therefore the Huynh–Feldt correction (ε = 0.867) was applied to adjust degrees of freedom (Table 2).

The Huynh–Feldt–adjusted repeated-measures ANOVA revealed a highly significant effect of instructional condition on Digital Literacy, F(1.734, 43.338) = 78.707, p < 0.001, partial η² = 0.759 (Table 3). This effect size indicates that approximately 76% of the variance in Digital Literacy scores can be attributed to differences across the three instructional designs – a very large effect according to conventional benchmarks.

To explore which specific contrasts drove the overall effect, Holm-adjusted pairwise comparisons were conducted (Table 4). Holm-adjusted post-hoc comparisons between instructional conditions). All pairwise comparisons were statistically significant at p < 0.001.

Students obtained significantly higher Digital Literacy scores in the FCDT condition than in the Conventional tutorial (mean difference, MD = −1.709, t(25) = −4.601, p < 0.001), with a small but reliable effect size (Cohen's d = −0.166). This suggests that the introduction of design-thinking-oriented flipped activities, supported by the UT module and LKM, provides a meaningful though modest enhancement over conventional tutorial practices.

The largest contrast emerged between the FCDT-AI and Conventional conditions (MD = −6.182, t(25) = −10.668, p < 0.001, d = −0.601), reflecting a large effect. Students in the AI-supported flipped condition not only outperformed those in the Conventional tutorial but did so by a considerable margin, indicating that AI-mediated scaffolding markedly strengthens students' ability to locate, evaluate, and use digital information.

Finally, the FCDT-AI condition also produced significantly higher scores than the FCDT condition without AI (MD = −4.473, t(25) = −8.126, p < 0.001, d = −0.435), with a medium effect size. This shows that, beyond the benefits attributable to flipped design thinking and LKM, the integration of AI tools yields an additional and practically meaningful gain in Digital Literacy.

Taken together, the quantitative analyses reveal a clear performance gradient: Conventional < FCDT < FCDT-AI, both in terms of mean scores and the distribution of outcomes.

To deepen the interpretation of the quantitative differences among the three instructional conditions (Conventional, FCDT, and FCDT-AI), student reflection logs were analysed thematically with a focus on Digital Literacy development. Responses from all students in each cycle were examined. Three progressively complex profiles of digital engagement emerged across the three conditions. Illustrative quotations are presented with respondent codes.

Student reflections after the Conventional tutorial predominantly described Digital Literacy as basic digital functioning: accessing online materials, searching for information, reading journal articles, and using simple word-processing tools. The majority of students perceived improvement mainly in accessing PDFs, reading articles, or typing assignments.

Students rarely reported deeper digital engagement. Digital tools were used mainly for: downloading and reading articles, and typing in Microsoft Word, uploading assignments through TUWEB. The Conventional tutorial fostered foundational Digital Literacy abilities – information access, basic navigation, and simple academic use of digital tools.

Reflections from the FCDT condition showed a clear qualitative shift. Students not only accessed digital information but also evaluated and used it to complete structured design-thinking tasks. Many described deeper engagement with the UT module, LKM activities, and more systematic search processes.

Students also reported early forms of digital production, such as: designing infographics, producing multimedia, organising digital references, and participating in collaborative online discussions. Thus, the FCDT condition fostered intermediate-level Digital Literacy – critical evaluation of sources, structured digital searching, content creation, and purposeful use of academic technologies.

The highest level of Digital Literacy appeared in the FCDT-AI reflections. Students consistently described complex, multimodal digital practices that integrated AI-supported thinking, multiple digital platforms, and reflective analysis.

Students showed mastery in: digital information retrieval using multiple databases (DOAJ, Scopus, Google Scholar, Garuda), digital evaluation and credibility filtering, multimodal creation (infographics, multimedia presentations, digital-native redesigns), collaborative digital communication (forums, feedback exchange), AI-supported idea generation and search strategies, and file formatting, uploading, and digital workflow management. These reflections demonstrate advanced Digital Literacy – critical digital judgement, creative and multimodal production, collaborative digital communication, and strategic integration of AI tools. This aligns strongly with the superior Digital Literacy performance observed quantitatively in the FCDT-AI condition.

The convergent quantitative and qualitative evidence paints a coherent picture of how different instructional designs shaped students' Digital Literacy. Quantitatively, the repeated-measures ANOVA and Holm-adjusted comparisons demonstrate a strong and statistically significant gradient in performance, with the FCDT-AI condition outperforming both FCDT and the Conventional tutorial. The Raincloud plots (Figure 1) reinforce this by showing a systematic upward shift in the distribution of Digital Literacy scores across the three conditions, while the Q–Q plot (Figure 2) confirms that these differences are not artefacts of distributional violations.

Qualitatively, the reflection logs explain why these differences emerged. In the Conventional tutorial, students mainly engaged in basic digital tasks – searching, reading, and typing – reflecting the lowest mean scores. The FCDT condition, with its structured out-class module use and in-class LKM scaffolding, encouraged more purposeful digital searching, evaluation of source credibility, and initial content creation, consistent with the intermediate performance level. The FCDT-AI condition added an additional layer of AI-supported guidance and multimodal production tasks, leading students to design innovative ways of accessing, visualising, and communicating information digitally. This deep, reflective engagement with technology aligns with the largest gains in Digital Literacy and the large effect size (partial η² = 0.759).

The joint interpretation indicates that the more the instructional design foregrounded structured scaffolding, design thinking, and AI-mediated digital inquiry, the more sophisticated and powerful students' Digital Literacy became. These findings provide strong empirical support for the pedagogical value of AI-enhanced flipped classroom models, particularly when paired with well-designed modules and LKMs in open and distance higher education contexts.

This study investigated the extent to which three instructional conditions – Conventional tutorial, FCDT, and FCDT-AI – enhanced working students' Digital Literacy in an ODL environment. The quantitative and qualitative findings consistently show that Digital Literacy follows a progressive trajectory across the three conditions, with the FCDT-AI model producing the greatest gains. This discussion interprets these findings in relation to the existing literature on flipped classrooms, design thinking, digital literacy, and responsible AI integration in higher education.

The significant differences revealed by the repeated-measures ANOVA, combined with the thematic patterns emerging from student reflections, confirm a cumulative increase in digital literacy competencies. The qualitative data provide explanatory depth, showing how learners shifted from basic digital access (Conventional), to structured evaluation and early content creation (FCDT), and finally to advanced multimodal digital practices (FCDT-AI). This convergence mirrors Creswell and Clark (2018) argument that mixed-methods designs allow researchers to trace how and why learning outcomes emerge across interventions. The alignment between numerical gains and increasingly sophisticated digital practices strengthens the internal validity of the findings.

The lowest Digital Literacy scores observed in the Conventional tutorial are consistent with students' self-reported engagement in primarily functional digital tasks – searching for journal articles, reading learning materials, and typing assignments – activities which prior research has identified as foundational but insufficient for developing higher-order digital competencies (Hewindati et al., 2023; Laba Laksana et al., 2025). Although these are essential competencies, they do not inherently promote the more advanced evaluative, creative, and communicative domains of Digital Literacy, which require critical engagement with information, synthesis across media, and purposeful digital production (Amin and Mirza, 2020; Tian et al., 2023; Xiao et al., 2024). The limited scaffolding within this condition likely contributed to the narrow distribution of scores, reflecting restricted opportunities to extend digital practice beyond basic access, a pattern in line with studies showing that unstructured exposure to digital tools without explicit guidance tends to yield confined improvements in digital literacy (Qamar et al., 2024; Tinmaz et al., 2023).

In the FCDT condition, Digital Literacy improved both quantitatively and qualitatively. The structured design-thinking phases – empathy, ideation, prototyping, and testing – encouraged students to explore digital sources purposefully and to evaluate their credibility, mirroring evidence that design-thinking processes foster analytic, evaluative, and reflective engagement with digital resources (Qamar et al., 2024; Tinmaz et al., 2023). Moreover, the use of Learning Activity Sheets (Lembar Kerja Mahasiswa/LKM) provided systematic scaffolds that guided students in producing digital artefacts such as presentations and infographics, consistent with research showing that well-structured worksheets and task sequences support the integration of information, media, and visual literacy skills in digital environments (Hewindati et al., 2023). These activities required learners to combine information, media, and visual literacy skills, thereby aligning with intermediate to advanced levels in multidimensional digital literacy frameworks that emphasise multi-domain competence rather than isolated technical skills (Amin and Mirza, 2020).

The largest improvements occurred in the FCDT-AI condition, where the integration of generative AI supported higher-order digital engagement. Students reported leveraging AI tools to expand ideational fluency, identify conceptual gaps, refine prototypes, and visualise information innovatively, echoing studies that associate AI-assisted learning with enhanced creativity, deeper engagement, self-efficacy, and more sophisticated digital artefact production (Anderson et al., 2025; Suwardika et al., 2026; Zhou and Peng, 2025). These practices reflect an advanced form of Digital Literacy that integrates critical judgement with creative and strategic use of digital tools, in line with multidimensional accounts of AI-related literacy that link cognitive, behavioural, and ethical engagement with complex digital tasks (Ng et al., 2024a). The minimal overlap in score distributions between the Conventional and FCDT-AI conditions, alongside the large effect size (partial η² = 0.759), indicates that AI-enabled scaffolding did more than augment existing capacities; it transformed learners' digital behaviours and broadened the range of achievable performance, a pattern consistent with evidence that AI-driven supports can shift learners from basic operational skills towards higher-order, integrated digital literacy (Suwardika et al., 2024; Tian et al., 2023; Zhou and Peng, 2025).

The findings are particularly relevant to ODL institutions such as Universitas Terbuka (UT), where learners are diverse, geographically dispersed, and often vary widely in digital readiness, a pattern widely documented in Asian ODL settings that require structured instructional support to reduce cognitive load and transactional distance (Goria and Konstantinidis, 2023; Lakmali et al., 2021; Zuhairi et al., 2020b). The progressive improvements observed across the three conditions highlight the critical role of structured scaffolding in supporting digital development among distance learners, consistent with evidence that ODL students benefit from explicit guidance and staged pedagogical support to mitigate disparities in readiness and engagement (Ahmed et al., 2022; Zuhairi et al., 2020a). The FCDT and FCDT-AI models, by aligning design-thinking processes with online learning activities, provided a coherent pedagogical mechanism for advancing Digital Literacy without increasing tutor workload excessively, reflecting findings that design-thinking activities combined with technology-enabled scaffolds can enhance autonomy and sustained engagement while avoiding undue instructional burden in large-scale ODL contexts (Engeness et al., 2025; Xue et al., 2025).

Moreover, the incorporation of AI-enabled scaffolds resonates strongly with AAOU's strategic priorities concerning digital transformation and equitable access to quality education across Asia, particularly in relation to the need for scalable and inclusive technological interventions in open universities (Salgado Granda et al., 2024; Suwardika et al., 2026; Xiao et al., 2024). By reducing cognitive load and supporting learners' decision-making, AI-mediated support mechanisms can enhance autonomy, reduce transactional distance, and promote active participation in ODL environments, echoing studies demonstrating that AI tools facilitate self-regulation, adaptive planning, and engagement in online learning ecosystems (Engeness et al., 2025; Sardi et al., 2025; Xue et al., 2025). These findings thus contribute to ongoing debates about how ODL institutions can responsibly integrate AI to enrich learning experiences while maintaining pedagogical integrity, aligning with contemporary ethical frameworks emphasising transparency, learner well-being, and principled AI deployment in education (Anderson et al., 2025; Ng et al., 2024a; Nguyen et al., 2023).

ODL tutors need explicit training in flipped pedagogies, digital content creation, and ethical AI facilitation. Without adequate tutor preparation, flipped and AI-supported models risk reverting to superficial use of technology rather than meaningful integration (Das and Bhattacharyya, 2023; Itasanmi et al., 2025; Uçar et al., 2024). Curriculum designers should embed Digital Literacy indicators within learning outcomes, assessments, and tutorial workflows. The FCDT-AI framework provides a feasible template: guided pre-class exploration, collaborative problem-solving, multimodal production, and reflective, ethically informed use of AI. Such alignment ensures that digital literacy development is intentional rather than incidental.

Implications for curriculum and tutor development also emerge clearly from the developmental progression observed across the three instructional conditions. The movement from basic digital access and retrieval in the Conventional tutorial, to structured evaluation and emerging digital production in the FCDT condition, and finally to advanced, multimodal, and AI-mediated digital engagement in the FCDT-AI condition indicates that Digital Literacy should be sequenced developmentally within the curriculum rather than treated as an incidental or generic competence. In ODL contexts, this means designing curricula that guide students from foundational skills, such as locating and organising information, towards higher-order capabilities involving source evaluation, multimodal composition, collaborative digital communication, and reflective use of AI-supported tools, in line with evidence that digital and information literacy are most effective when embedded progressively within programme design rather than addressed as isolated technical skills (Amin and Mirza, 2020). Such sequencing would allow learning outcomes, tasks, and assessment criteria to reflect increasingly complex forms of digital practice across tutorial cycles or course stages, while accommodating differences in learners' prior experience and disciplinary background (Das and Bhattacharyya, 2023). Effective implementation, however, requires institutional support beyond individual tutor effort, including professional development, exemplar materials, shared rubrics, ethical AI-use guidelines, and ongoing pedagogical and technical support (Ferdousi et al., 2022). Without such infrastructure, the transformative potential of the FCDT-AI approach may remain uneven and overly dependent on individual tutor capability (Sembiring and Rahayu, 2020).

The integration of generative AI requires careful attention to academic integrity, algorithmic bias, transparency, and responsible use. International guidelines emphasise that students must be taught to question the accuracy of AI outputs, recognise limitations, and maintain originality (UNESCO, 2023). In this study, AI-supported tasks were structured to promote critical evaluation rather than passive adoption of AI-generated content. This is consistent with calls for the development of AI literacy as a core component of digital literacy in higher education (Ng et al., 2024a; Wang et al., 2025).

Several limitations must be acknowledged. First, the sample was restricted to one programme in one institution, limiting generalisability. Replication in other disciplines, age groups, and institutions is needed. Second, qualitative reflections may be influenced by self-report bias. Although the repeated-measures design increased statistical efficiency for this modest sample by allowing each participant to serve as their own control, the absence of demographic variables such as age, gender, employment status, and prior digital experience limits interpretation of whether the observed effects operated similarly across different student subgroups. Complementary sources such as learning analytics or artefact analysis would strengthen triangulation (Smith and Firth, 2011). Third, the AI component was intentionally constrained to uphold ethical safety; future studies could systematically vary AI affordances to examine their differential effects on digital literacy development.

The study demonstrates that Digital Literacy among working ODL students can be improved meaningfully through well-designed flipped instruction grounded in design thinking and enriched by guided AI use. The FCDT-AI model fostered advanced multimodal, evaluative, and creative digital practices that align with global expectations for 21st-century academic and professional competencies. The findings contribute to growing scholarship on AI-supported learning and offer a practical, scalable model for ODL institutions seeking to enhance Digital Literacy in flexible learning environments.

The authors gratefully acknowledge the financial support provided by LPPM Universitas Terbuka for this research and the subsequent publication of this article.