





































  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

422 

 

EXPLORING CHATGPT AS A REPLACEMENT FOR 
TRADITIONAL READING: PERCEPTIONS OF ENGLISH 

EDUCATION POSTGRADUATE STUDENTS 

 
Muhammad Tahir1, Andi Sahtiani Jahrir2 

¹,2Universitas Negeri Makassar, Makassar, Indonesia 
 

muhammadtahir@unm.ac.id 
 

ABSTRACT 

The rapid integration of artificial intelligence into higher education has reshaped how 
postgraduate students approach academic reading and learning. This study explores 
postgraduate English Education students’ perceptions of ChatGPT as a substitute for 
traditional reading, focusing on how the tool influences their cognitive engagement and 
reading behaviour. Anchored in the Technology Acceptance Model (TAM), Cognitive 
Offloading Theory, and Reading Literacy Theory, the research employs a quantitative 
survey design involving 108 master’s students from the Graduate Program of 
Universitas Negeri Makassar. A structured questionnaire measured four constructs: 
Perceived Usefulness (PU), Perceived Ease of Use (PEOU), Cognitive Offloading (COG), 
and Reading Habits Impact (RHI). Data were analysed using descriptive statistics, 
correlation, and reliability testing. The findings reveal that students perceive ChatGPT 
as highly useful (M = 4.23) and easy to use (M = 4.35). These positive perceptions 
correlate strongly with cognitive offloading (r = 0.61), indicating frequent reliance on 
ChatGPT to simplify learning tasks. However, a significant negative correlation between 
cognitive offloading and reading habits (r = –0.54) suggests that increased dependence 
on ChatGPT reduces students’ motivation for traditional reading. Overall, the study 
highlights a dual outcome, while ChatGPT enhances learning efficiency and accessibility, 
it simultaneously contributes to the decline of deep reading practices. The results 
underscore the need for balanced AI integration that promotes critical reading, 
reflective thinking, and responsible technology use in postgraduate education. 

Keywords: ChatGPT, English Language Learning, Reading Literacy, Technology Acceptance 

INTRODUCTION 

Over the past decade, Indonesian universities have witnessed a remarkable 
transformation in postgraduate students’ learning behaviors. Graduate students in 
English Education programs, such as those at Universitas Negeri Makassar (UNM), are 
no longer solely dependent on printed textbooks, academic journals, and intensive 
reading practices. The increasing accessibility of digital resources, the prevalence of 
online learning environments, and the integration of artificial intelligence (AI) into 
educational contexts have shifted the way students acquire and process knowledge. A 
growing number of postgraduate learners now prefer to utilize technological tools, 
particularly generative AI systems such as ChatGPT, to comprehend complex materials, 
complete assignments, and prepare research projects. This phenomenon represents not 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689
mailto:muhammadtahir@unm.ac.id


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

423 

 

merely a change in learning media, but a fundamental shift in academic literacy 
behaviors. 

In the traditional paradigm, reading books, analyzing scholarly articles, and 
engaging deeply with texts were viewed as essential components of postgraduate 
academic culture. However, many students have started perceiving conventional 
reading as time-consuming, cognitively demanding, and less efficient amid heavy course 
loads and multiple academic obligations. Conversely, ChatGPT provides immediate 
summaries, interactive explanations, and flexible access to information at any time. Its 
conversational interface allows learners to clarify difficult concepts in seconds—
something that would typically take hours through independent reading. Consequently, 
ChatGPT has become an appealing and time-efficient learning companion for students 
who prioritize rapid understanding and task completion. 

Despite these conveniences, a critical question arises for higher education: does 
the use of ChatGPT as a substitute for traditional reading affect the depth of 
comprehension, academic literacy, and reflective thinking of postgraduate students? 
Are we observing a paradigmatic shift from reading-based learning to AI-assisted 
learning? If so, what implications does this have for the quality of cognitive engagement, 
research competence, and critical reasoning? These concerns form the foundation of the 
present study, which explores the perceptions of English Education postgraduate 
students at Universitas Negeri Makassar regarding ChatGPT as a substitute for 
traditional reading in their academic practices. 

Recent studies in higher education have highlighted the increasing adoption of 
generative AI tools among university students. A large-scale international survey, 
Ravšelj, et. al (2025), involving more than 23,000 students across 109 countries, 
revealed that most learners consider ChatGPT an effective academic aid, while 
simultaneously expressing concerns about reliability and ethical use. Similarly, Khan et 
al. (2024) in Exploring Learners’ Experiences and Perceptions of ChatGPT as a Learning 
Tool in Higher Education found that ChatGPT facilitates interactive learning, enhances 
language support, and assists with comprehension—especially in non-native English 
contexts. 

In the domain of reading literacy, research indicates significant behavioral shifts 
among university students. She etc. (2025), through the Digital Academic Reading 
Behavior (DARB) model, observed that while students increasingly read digital 
materials, the depth and retention of comprehension in digital reading tend to be lower 
than those achieved through printed text. A meta-analysis by Sohn et al. (2023) likewise 
demonstrated that digital leisure reading produces minimal improvements in 
comprehension when compared to traditional paper-based reading. Furthermore, 
studies on English as a Foreign Language (EFL) learners, such as Rahman et al. (2024), 
reported that although students prefer digital sources for convenience and accessibility, 
they still struggle with maintaining focus and sustaining engagement when reading 
lengthy academic texts. 

Additional evidence from developing contexts strengthens this view. A study in 
Nigeria by Ajay and Ogunleye (2023) found that digital resource usage correlates with 
changes in students’ reading habits—sometimes positively, when digital tools support 
academic reading, but also negatively when such tools encourage superficial scanning 
rather than in-depth comprehension. In the specific context of ChatGPT, Pérez-Mira et 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

424 

 

al. (2025), in Study on Students’ Use of ChatGPT in Higher Education: Benefits, Costs, 
and Attitudes, discovered that students who perceive ChatGPT as reliable report 
increased understanding of course content but also a growing dependence on AI-
generated information, sometimes leading to procrastination or reduced cognitive 
effort. 

Taken together, these findings suggest four overarching trends. First, ChatGPT has 
rapidly become an integral component of students’ learning ecosystems worldwide.  
Second, reading behaviors are shifting from deep analytical reading to faster, surface-
level digital engagement. Third, the cognitive consequences of ChatGPT use—especially 
regarding academic literacy remain ambiguous and underexplored. Fourth, most studies 
so far have employed general quantitative surveys, lacking contextual focus on specific 
academic disciplines or learner profiles. 

Theoretically, the Technology Acceptance Model (TAM) (Davis, 1989; Venkatesh et 
al., 2021) remains the dominant framework explaining how perceived usefulness and 
perceived ease of use shape technology adoption in education. Complementarily, 
Cognitive Offloading Theory (Risko & Gilbert, 2016) explains how learners transfer 
mental tasks to external aids, such as AI, reducing intrinsic cognitive effort. Reading 
Literacy Theory (Guthrie & Wigfield, 2023) emphasizes the importance of sustained 
deep reading and textual engagement for academic development. However, empirical 
attempts to integrate these three frameworks, technology acceptance, cognitive 
offloading, and reading habits—within the context of ChatGPT use among postgraduate 
EFL or English Education students are still very limited. 

While international literature has begun to illuminate students’ perceptions of 
ChatGPT and digital reading patterns, several notable research gaps persist, particularly 
in the Indonesian higher education context. First, the majority of existing studies focus 
on undergraduate populations, leaving postgraduate learners, who engage in more 
complex academic reading and research underrepresented. Second, most prior research 
treats ChatGPT as a general learning aid rather than explicitly analyzing it as a 
substitute for traditional reading. The nuanced question of whether students are 
replacing book reading with ChatGPT interaction has rarely been addressed empirically. 

Third, disciplinary specificity is lacking: few studies explore ChatGPT use among 
English Education students, whose coursework demands extensive reading of linguistic 
theory, pedagogy, and research methodology. The impact of AI tools on such discipline-
specific reading habits remains largely unexplored. Fourth, within Indonesia, 
particularly in eastern universities such as UNM, there is an absence of localized data on 
how postgraduate students perceive ChatGPT in relation to academic reading, 
comprehension depth, and literacy practices. 

Consequently, there is a critical need to fill this gap by examining how 
postgraduate English Education students conceptualize the role of ChatGPT as either a 
complement to or a replacement for traditional reading and how this perception 
influences their academic engagement. This investigation will not only contribute 
empirical evidence to the global discourse on AI in education but also provide 
contextual insights into the shifting literacy practices of Indonesian postgraduate 
learners. 

In light of the preceding discussion, this study aims to explore English Education 
postgraduate students’ perceptions of ChatGPT as a learning aid and as a substitute for 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

425 

 

traditional reading at Universitas Negeri Makassar. Specifically, the study seeks to: (1) 
Identify the factors influencing students’ preference for ChatGPT over traditional book 
reading—particularly in terms of perceived usefulness and perceived ease of use (TAM 
framework). (2) Examine how ChatGPT affects postgraduate students’ academic reading 
habits, motivation, and engagement with printed or digital texts. (3) Determine whether 
ChatGPT use indicates a broader literacy transition from reading-centered learning to 
AI-mediated learning among postgraduate English Education students. 

The significance and contribution of this research can be seen in several key 
aspects. (a) It focuses explicitly on postgraduate English Education students in 
Indonesia, a group that has received limited empirical attention in global studies on 
ChatGPT use. (b) It investigates ChatGPT as a specific substitute for traditional reading, 
rather than treating it merely as a general technological aid. (c) It integrates theoretical 
perspectives from the Technology Acceptance Model (TAM), Cognitive Offloading 
Theory, and Reading Literacy Theory to construct a multidimensional analytical 
framework connecting technological perceptions with academic literacy practices. (d) It 
also offers a contextual contribution by presenting empirical insights from an 
Indonesian higher-education setting, bridging international discussions on AI in 
education with the local realities of postgraduate learning at Universitas Negeri 
Makassar. 

Through these contributions, the study highlight to enrich scholarly 
understanding of how AI adoption transforms postgraduate literacy and learning 
paradigms in the field of English education. It also offers pedagogical implications for 
balancing technological efficiency with academic depth. While ChatGPT may streamline 
information processing and support comprehension, higher education institutions must 
ensure that students continue to cultivate reflective reading habits and critical 
inquiry—skills fundamental to advanced academic scholarship. 

In essence, this research situates the phenomenon of ChatGPT use not as a 
technological trend, but as a cultural and cognitive transformation within postgraduate 
education. It argues that the postgraduate learner’s journey in the AI era is no longer 
defined solely by how much they read, but by how they engage with knowledge: 
through text, through dialogue with AI, and through the synthesis of both. By 
investigating this intersection, the present study seeks to contribute theoretically and 
empirically to the growing discourse on digital literacy, language education, and the 
evolving ecology of learning in the age of artificial intelligence. 

METHODS 

Research Design 

This study employed a quantitative survey design to investigate postgraduate 
students’ perceptions of ChatGPT as a substitute for traditional reading within the 
English Education program at Universitas Negeri Makassar (UNM). The quantitative 
approach was selected because it allows the systematic collection of numerical data that 
can describe patterns, tendencies, and relationships among variables. The survey design 
also provides a snapshot of how students perceive the usefulness and ease of use of 
ChatGPT, their degree of cognitive offloading, and the impact on their academic reading 
habits. 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

426 

 

The study integrates constructs from the Technology Acceptance Model (TAM) 
(Davis, 1989; Venkatesh et al., 2021) and the Cognitive Offloading and Reading Habits 
Framework (Risko & Gilbert, 2016; Guthrie & Wigfield, 2020). Four primary latent 
variables were measured: Perceived Usefulness (PU), Perceived Ease of Use (PEOU), 
Cognitive Offloading (COG), and Reading Habits Impact (RHI). These constructs 
collectively represent how students experience ChatGPT as a learning aid compared to 
traditional academic reading. The study was conducted between March and June 2025. 

Participants 

The participants consisted of 112 postgraduate students (Master’s level) enrolled 
in the English Education Program at the Graduate School of Universitas Negeri 
Makassar (UNM). All participants were actively engaged in coursework or thesis 
preparation during the data collection period. A purposive sampling technique was 
employed to ensure that the respondents had sufficient experience using ChatGPT in 
academic activities, such as literature review writing, lesson-plan design, or classroom 
research. 

Of the total respondents, 76 (67.9%) were female and 36 (32.1%) were male, 
reflecting the gender distribution commonly observed in English Education programs. 
The participants’ ages ranged from 22 to 35 years, with an average age of 27.4 years. 
Most students reported using ChatGPT for academic purposes at least three times per 
week, primarily to summarize journal articles, explain theoretical concepts, or generate 
examples of academic writing. Participation was voluntary, and informed consent was 
obtained from all respondents prior to completing the questionnaire. 

Research Instrument 

Data were collected through a structured questionnaire developed based on 
established scales and previous studies. The instrument consisted of four sections 
corresponding to the four constructs of the model: 

Table 1. Constructs, conceptual definitions, sample items, and sources of the research instrument 

Construct Conceptual Definition Items Sources 

Perceived 
Usefulness (PU) 

The extent to which students 
believe that ChatGPT enhances 
their academic performance and 
comprehension. 

ChatGPT helps me 
understand complex 
academic concepts 
faster. 

Davis (1989); 
Aljanabi et al. 
(2024) 

Perceived Ease of 
Use (PEOU) 

The degree to which students 
find ChatGPT easy to use and 
interact with. 

It is easy for me to use 
ChatGPT for my 
coursework. 

Venkatesh et al. 
(2021); Albayati,  
(2024) 

Cognitive 
Offloading (COG) 

The tendency to rely on ChatGPT 
to reduce personal cognitive 
effort during learning. 

I depend on ChatGPT 
instead of reading 
entire articles. 

Risko & Gilbert 
(2016); 
Papadopoulos 
(2009) 

Reading Habits 
Impact (RHI) 

The influence of ChatGPT use on 
students’ motivation and 
frequency of traditional reading. 

Since using ChatGPT, I 
read fewer academic 
books. 

Guthrie & Wigfield 
(2020); Mirza & 
Jabeen (2025) 

Note: The questionnaire consisted of four sections representing the main constructs of the study. Table 1 
presents the conceptual definitions, sample items, and references for each construct. 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

427 

 

Each construct contained 4 to 5 items, resulting in a total of 19 items in the 
questionnaire. The items were measured on a 5-point Likert scale ranging from 1 = 
Strongly Disagree to 5 = Strongly Agree. The instrument was developed in English, 
reviewed by two experts in applied linguistics and educational technology, and pilot-
tested on 25 students to assess clarity and reliability. Minor wording adjustments were 
made to ensure cultural appropriateness and comprehension among Indonesian 
respondents. The reliability coefficients (Cronbach’s α) for each construct were as 
follows: PU = 0.87, PEOU = 0.85, COG = 0.82, and RHI = 0.79, indicating high internal 
consistency. Construct validity was established through Confirmatory Factor Analysis 
(CFA), with all factor loadings above 0.60, and Average Variance Extracted (AVE) values 
above 0.50, demonstrating adequate convergent validity. 

Data Collection Procedures 

Prior to data collection, ethical clearance was obtained from the Faculty of 
Language and Literature Research Ethics Committee at UNM. The survey was 
administered online using Google Forms, which enabled efficient distribution to 
postgraduate student groups. The online format was chosen for accessibility and to 
accommodate students’ flexible schedules. Respondents were informed about the 
purpose of the study, confidentiality, and voluntary participation. The data collection 
process took approximately four weeks. Each participant received the questionnaire 
link through official university channels and WhatsApp study groups. Responses were 
automatically stored and compiled in a secure database. After data cleaning, incomplete 
or inconsistent responses (n = 4) were excluded, resulting in 108 valid cases for 
statistical analysis. 

Data Analysis 

Data were analyzed using IBM SPSS Statistics v.26 and AMOS v.24. The analysis 
comprised three stages, (a) Preliminary analysis, this stage involved data cleaning, 
screening for incomplete responses, and assessing normality, missing values, and 
outliers. Four invalid cases were removed, resulting in 108 valid responses. Descriptive 
statistics (mean, standard deviation, frequency distribution) were computed to 
summarize participants’ demographic characteristics and initial response patterns. (b) 
Reliability and validity testing, construct reliability was examined using Cronbach’s 
Alpha and Composite Reliability (CR). Convergent validity was assessed through factor 
loadings and Average Variance Extracted (AVE), with all loadings exceeding 0.60 and 
AVE values above 0.50. Discriminant validity was evaluated based on the Fornell 
Larcker criterion and inter-construct correlations. (c) Inferential and structural analysis, 
pearson correlation analysis was conducted to examine relationships among Perceived 
Usefulness (PU), Perceived Ease of Use (PEOU), Cognitive Offloading (COG), and Reading 
Habits Impact (RHI). Structural Equation Modeling (SEM) using AMOS v.24 was then 
performed to test the hypothesized relationships within the integrated TAM Cognitive 
Offloading framework. Model fit indices including χ²/df, CFI, TLI, RMSEA, and GFI were 
generated to determine overall model adequacy. Path coefficients were examined at a 
significance level of p < 0.05 to evaluate direct and indirect effects among variables. 

 
 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

428 

 

Descriptive Statistics 

Mean, standard deviation, and frequency distribution were calculated to describe 
participants’ overall perceptions of ChatGPT’s usefulness, ease of use, cognitive 
offloading, and impact on reading habits. The descriptive data were visualized using bar 
charts to compare the mean scores of the four constructs. Cronbach’s alpha and 
composite reliability values were examined to ensure measurement consistency. 
Convergent validity was confirmed through factor loadings and AVE values as reported 
earlier. Discriminant validity was assessed using the Fornell-Larcker criterion. Pearson 
correlation coefficients were computed to examine relationships among the four 
variables. Subsequently, Structural Equation Modeling (SEM) was employed to test the 
hypothesized relationships: 

PU → COG (positive) 
PEOU → COG (positive) 
COG → RHI (negative) 
PU/PEOU → RHI (indirect via COG) 

The model fit indices indicated satisfactory fit (χ²/df = 2.31; CFI = 0.94; TLI = 0.92; 
RMSEA = 0.061; GFI = 0.91). Significance levels were set at p < 0.05. All findings were 
interpreted within the theoretical frameworks of TAM and Cognitive Offloading, with 
further implications discussed in the Discussion section. 

RESULTS 

Descriptive Statistics 

Descriptive analysis was conducted to summarize the overall perceptions of 
postgraduate English Education students regarding the usefulness, ease of use, 
cognitive offloading, and reading habits impact of ChatGPT. Table 5 presents the means 
and standard deviations for each construct. 

Table 2. Descriptive statistics of main variables (N = 108) 

Variable Items Mean (M) Stand. Deviation (SD) Interpretation 

Perceived Usefulness 
(PU) 

5 4.23 0.62 High 

Perceived Ease of Use 
(PEOU) 

4 4.35 0.58 Very High 

Cognitive Offloading 
(COG) 

5 4.08 0.66 High 

Reading Habits Impact 
(RHI) 

5 3.92 0.71 
Moderately High 
(negative impact) 

The results indicate that students perceive ChatGPT as both useful and easy to use, 
with mean scores above 4.20. Cognitive offloading also scored highly (M = 4.08), 
suggesting that students frequently rely on ChatGPT to simplify academic tasks. 
Meanwhile, the Reading Habits Impact variable (M = 3.92) shows a moderately high 
mean in the negative direction, implying that frequent use of ChatGPT is associated with 
a decline in traditional reading motivation. 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

429 

 

To visualize these findings, Figure 1 presents a bar chart comparing the mean 
scores of the four variables. 
 

 

 

 

 

 

 

 

 

 

Figure 1. Average perceptions scores of students on ChatGPT use 

The bar chart confirms that Perceived Ease of Use (PEOU) is the highest-rated 
construct, followed closely by Perceived Usefulness (PU), reflecting students’ strong 
positive attitudes toward ChatGPT as a convenient academic tool. 

Reliability and Validity of Constructs 

Instrument reliability and validity were assessed using Cronbach’s Alpha, factor 
loadings, and Average Variance Extracted (AVE). All four constructs exceeded the 
minimum threshold for internal consistency (α ≥ 0.70) and convergent validity (AVE ≥ 
0.50). 

Table 3. Reliability and validity summary 

Construct 
No. of 
Items 

Cronbach’s Alpha 
(α) 

Factor Loading 
(Range) 

AVE Interpretation 

Perceived Usefulness 
(PU) 

5 0.87 0.70 – 0.85 0.63 
Reliable & 
valid 

Perceived Ease of Use 
(PEOU) 

4 0.85 0.68 – 0.84 0.61 
Reliable & 
valid 

Cognitive Offloading 
(COG) 

5 0.82 0.66 – 0.81 0.58 
Reliable & 
valid 

Reading Habits 
Impact (RHI) 

5 0.79 0.63 – 0.79 0.55 
Reliable & 
valid 

These findings confirm that the questionnaire is statistically sound and can 
reliably capture postgraduate students’ perceptions across all measured constructs. The 
internal consistency coefficients are high, particularly for PU and PEOU, showing that 
students’ responses were consistent regarding their evaluation of ChatGPT’s 
effectiveness and ease of use. 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

430 

 

Correlation Analysis 

To explore relationships among the four constructs, Pearson correlation 
coefficients were computed. Table 7 displays the correlation matrix, and Figure 2 
provides a heatmap visualization of these associations. 

Table 4. Correlation matrix among variables 

Variable PU PEOU COG RHI 

PU 1.000 0.67** 0.58** –0.42** 
PEOU 0.67** 1.000 0.61** –0.39** 
COG 0.58** 0.61** 1.000 –0.54** 
RHI –0.42** –0.39** –0.54** 1.000 

*Note: *p < 0.01 

The results indicate several significant relationships among the four main 
constructs measured in this study. A strong positive correlation was identified between 
Perceived Usefulness (PU) and Cognitive Offloading (COG) (r = 0.58), as well as between 
Perceived Ease of Use (PEOU) and Cognitive Offloading (r = 0.61). These findings 
suggest that postgraduate students who perceive ChatGPT as more useful and easier to 
operate tend to depend on it more frequently for academic tasks, thereby reducing the 
amount of mental effort required to comprehend or process information. In other 
words, the more students appreciate ChatGPT’s convenience and perceived benefits, the 
more they engage in cognitive delegation  allowing the tool to handle complex thinking 
or summarization processes on their behalf. 

Conversely, negative relationships emerged between Cognitive Offloading (COG) 
and Reading Habits Impact (RHI) (r = –0.54), showing that increased reliance on 
ChatGPT is associated with a decline in traditional reading engagement. This indicates 
that students who often use ChatGPT to aid their studies are less likely to sustain 
regular reading habits, such as reading academic books or scholarly articles in depth. 
Similarly, both Perceived Usefulness (r = –0.42) and Perceived Ease of Use (r = –0.39) 
were negatively correlated with Reading Habits Impact (RHI). This pattern implies that 
higher satisfaction with ChatGPT’s ease of use and functionality coincides with reduced 
motivation to engage with printed or original materials. Overall, these correlations 
illustrate a consistent trend: as ChatGPT becomes more embedded in students’ 
academic routines, it tends to enhance efficiency but simultaneously diminish their 
commitment to deep, reflective reading. 
  

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

431 

 

 

 

 

 

 

 
 
 
 
 
 
 
 

Figure 2. Correlation matrix among main variables (n = 108) 

The heatmap clearly highlights the direction and magnitude of the correlations: darker 

blue tones represent strong positive associations (PU–PEOU–COG), while deeper red tones 

show strong negative links (COG–RHI). This pattern visually demonstrates the trade-off 

between technological reliance and academic reading engagement. 

DISCUSSION 

The findings of this study provide meaningful insights into postgraduate students’ 
perceptions of using ChatGPT as a learning aid and a substitute for traditional book-
based reading in the English Education programme at Universitas Negeri Makassar, 
Indonesia. By integrating the Technology Acceptance Model (TAM), Cognitive Offloading 
theory, and Reading Literacy perspectives, this discussion seeks to interpret the results, 
connect them to prior research, and elaborate implications for academic practice and 
future research. 

Interpreting Results through the TAM Lens 

According to the Technology Acceptance Model, users’ behavioural intentions and 
actual usage of a technology are largely influenced by Perceived Usefulness (PU) and 
Perceived Ease of Use (PEOU) (Davis, 1989). In this study, the high mean values for both 
PU (M = 4.23) and PEOU (M = 4.35) demonstrate that the postgraduate students 
regarded ChatGPT as both highly useful and easy to use within their academic activities. 
These perceptions align with the significant positive correlations between PU and 
Cognitive Offloading (r = 0.58) and PEOU and Cognitive Offloading (r = 0.61). Thus, in a 
TAM framework, students who believe ChatGPT supports their learning effectively and 
with minimal effort are more inclined to use it frequently and rely on it cognitively. 

This result resonates with meta-analyses and reviews of TAM in educational 
technology contexts, which confirm that PU and PEOU remain major predictors of 
technology adoption (Schorr, 2023; Abbas & Al-Lawati, 2025). For instance, educational 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

432 

 

studies have found that ease of use can moderate the effect of usefulness on behavioural 
intention, and in our context, PEOU may have enhanced students’ readiness to use 
ChatGPT routinely. It is noteworthy that PEOU out-scored PU slightly in our research; 
this suggests that convenience and reduced effort may be more salient drivers for this 
cohort than outright performance enhancement. 

However, TAM alone does not fully account for the downstream effects observed, 
namely the negative correlation between Cognitive Offloading and Reading Habits 
Impact (r = –0.54). While TAM explains adoption, it does not explain how technology 
usage might alter cognitive or literate behaviours — hence the need to integrate 
additional theoretical lenses. 

Cognitive Offloading, ChatGPT, and Academic Reading 

Cognitive Offloading theory posits that individuals may shift part of their cognitive 
workload onto external tools or systems in order to reduce effort and increase 
efficiency (Risko & Gilbert, 2016). In the present study, the high mean for Cognitive 
Offloading (M = 4.08) and the strong negative relationship with Reading Habits Impact 
illustrate that students who offload cognitive effort to ChatGPT tend to engage less with 
traditional reading activities. 

This pattern echoes recent scholarship on the AI efficiency–literacy paradox — 
while artificial intelligence enhances efficiency and access, it may reduce deep cognitive 
engagement. For example, Hossain (2025) found in a study published in Frontiers in 
Psychology that increased reliance on AI tools correlates with decreased processing 
depth and fewer opportunities for active recall and problem-solving. Similarly, Gerlich 
(2025) observed that structured prompting and reflective engagement may mitigate 
offloading, but unstructured AI use tends to promote superficial processing. 

In the context of postgraduate English Education students, this effect carries 
particular weight. Reading academic texts — textbooks, journal articles, research 
reports — is a core component of disciplinary literacies in English education. When 
students substitute reading with ChatGPT prompts, they may bypass the labour-
intensive but cognitively rich process of deep reading, analysis, and synthesis. This 
result is confirmed by the moderate negative correlations between PU (r = –0.42) and 
PEOU (r = –0.39) with Reading Habits Impact (RHI). Thus, even as students perceive 
ChatGPT as effective, their traditional reading practices decline. 

The theoretical implication here is that cognitive offloading acts as a mediator 
between technology acceptance (TAM) and literacy behaviour (reading habits). In other 
words, PU and PEOU lead to higher offloading, which then leads to lower engagement 
with reading. This mediated pathway aligns with research showing that AI-usage can 
increase offloading, which in turn negatively affects critical thinking or reading depth 
(Kulal, 2025). The practical implication is that institutions should not only promote 
technology adoption but also monitor and guide how students engage cognitively with 
those tools. 

Implications of Reduced Book Reading 

Reading literacy, especially in higher education, is not merely an act of decoding 
text; it involves sustained concentration, reflective thinking, critical synthesis, and the 
internalisation of complex ideas (Guthrie & Wigfield, 2020). While we did not measure 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

433 

 

facets such as reading comprehension depth or metacognitive reading strategies, our 
RHI measure gives an indicator of reduced reading output and motivation. The 
correlation findings suggest that students who offload more tend to read less. This 
raises important concerns about the long-term development of disciplinary literacies. 

In English Education, students must engage with a wide range of texts: linguistic 
theories, applied pedagogy, curriculum standards, research methodology, professional 
literature, and more. The act of reading these texts builds background knowledge, 
disciplinary vocabulary, and the capacity for critical reflection — all of which underpin 
teaching competence and research skills. If ChatGPT substitutes reading, then there is a 
risk of a shallow reading culture emerging, one in which learners rely on summaries, 
prompts, and AI assistance rather than engaging firsthand with the nuances of the 
original texts. 

Moreover, the shift away from reading may also impair the development of what 
some researchers call deep reading routines — the sustained, uninterrupted 
engagement with complex texts that fosters insight and creativity. Sohn et al. (2022) 
found that digital reading initiatives often fail to replicate the comprehension benefits of 
printed texts unless paired with reflective practices. Although their study focused on 
leisure reading, the logic applies to academic reading: if students are skipping original 
texts, then the richness of disciplinary knowledge may suffer. 

Pedagogical and Institutional Implications 

Given these findings, several practical considerations arise for pedagogy, graduate 
programmes, and institutional policy. Firstly, educators should design tasks that require 
engagement with primary texts (books, articles) rather than only AI-generated 
summaries. For example, ChatGPT can be used as a supplementary tool, to clarify 
concepts, generate questions, or propose alternative examples, but not as a replacement 
for reading full texts. This aligns with Gerlich’s (2025) emphasis on structured 
prompting and reflective follow-up as key to mitigating cognitive offloading. Secondly, 
graduate programmes in English Education should incorporate AI-literacy training, 
emphasising not just tool use but critical engagement with AI-outputs. Students should 
be aware of the trade-offs: convenience may come at the cost of reading discipline, 
critical thinking, and knowledge depth. Thirdly, institutional policies should encourage 
reading-rich cultures. This might include dedicated reading times, reading groups, peer-
discussion sessions about texts, and assignments requiring the critique of AI-
summarised content. By weaving reading back into the curriculum, institutions can 
balance the technological efficiency with academic depth. Finally, a blended approach is 
advisable — integrating ChatGPT and other AI tools into learning design while 
preserving time, space, and incentives for traditional reading. The goal is not to resist AI 
but to harness it in ways that augment rather than replace human cognitive effort. 

Limitations and Future Research Directions 

While this study provides important insights, it has several limitations. The 
sample is limited to postgraduate English Education students at a single institution 
(UNM), which may restrict generalisability. Self-reported measures may be subject to 
social desirability bias. Also, the study did not measure actual reading comprehension 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

434 

 

outcomes, long-term retention, or the quality of reading engagement — only self-
perceived impacts. 

Future research should explore longitudinal designs to track how ChatGPT usage 
affects reading habits and comprehension over time. Experimental studies might 
compare groups using ChatGPT versus traditional reading and assess differences in 
reflection, synthesis, and knowledge retention. It would also be valuable to extend the 
model to other disciplines, institutions, and cultural contexts, and to incorporate more 
nuanced reading literacy measures (e.g., metacognitive reading strategies, deep reading 
routines). 

CONCLUSION 

This study explored postgraduate English Education students’ perceptions of 
ChatGPT as a learning aid and as a substitute for traditional reading at Universitas 
Negeri Makassar. Drawing upon the theoretical foundations of the Technology 
Acceptance Model, Cognitive Offloading, and Reading Literacy perspectives, the findings 
reveal a dynamic shift in how students engage with knowledge, technology, and literacy 
in the context of higher education. The results show that students perceive ChatGPT as 
both highly useful and easy to use, indicating a strong acceptance of AI as a learning 
tool. These perceptions are consistent with a growing tendency among postgraduate 
learners to integrate technology into their academic routines. However, the data also 
reveal a significant trade-off: as reliance on ChatGPT increases, the frequency and depth 
of students’ traditional reading practices decline. In other words, efficiency and 
convenience appear to come at the cost of sustained engagement with academic texts. 

This duality reflects the new learning reality faced by postgraduate students. 
ChatGPT has become a convenient partner in processing complex materials, providing 
instant explanations and summaries that accelerate understanding. Yet, this 
convenience risks diminishing the habits of deep reading and reflection that are 
essential for advanced scholarship. The more students depend on AI to interpret or 
summarise knowledge, the less they engage in the critical and interpretive thinking that 
reading independently demands. From a pedagogical standpoint, these findings call for 
a balanced approach to technology integration in postgraduate education. ChatGPT 
should be positioned not as a replacement for reading but as a complementary tool that 
supports comprehension and critical engagement. Educators can design learning 
activities that require students to interact with both AI-generated responses and 
original texts, encouraging them to question, verify, and reflect on the information they 
receive. Such practices will help maintain a healthy balance between efficiency and 
intellectual depth. 

For institutions, the challenge lies in fostering digital literacy that includes ethical 
and reflective AI use. Students need to be trained not only to use ChatGPT effectively but 
also to remain aware of its limitations. Encouraging structured reading programs, 
critical discussions, and reflective journals can help sustain reading motivation and 
ensure that AI tools enhance rather than diminish literacy. In essence, this study 
illustrates that the integration of ChatGPT into postgraduate learning represents both 
progress and caution. It enhances accessibility and learning speed but also risks 
narrowing the scope of intellectual engagement if used uncritically. The findings 
highlight the importance of guiding students toward mindful and responsible use of AI 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

435 

 

— one that promotes inquiry, understanding, and reflection rather than passive 
consumption of information. 

Ultimately, the success of AI integration in postgraduate English Education 
depends on balance. Technology should serve as an extension of students’ cognitive 
capacity, not a substitute for it. As universities move further into the era of AI-driven 
learning, the goal should be clear: to develop learners who are technologically 
competent, intellectually independent, and capable of combining the precision of 
machines with the depth of human thought. 

REFERENCES 

Abbas, A. F., & Al-Lawati, E. H. (2025). Revisiting artificial intelligence in start-ups: A 

theoretical perspective on integration, opportunities, challenges, and strategic 

advancement. Journal of the International Council for Small Business, 1–44. 

https://doi.org/10.1080/26437015.2025.2549059 

Ajayi, A., & Ogunleye, T. (2023). Digital resource usage and reading habits among 

university students in Nigeria. International Journal of Educational Technology in 

Higher Education, 20(4), 1–15. https://doi.org/10.5281/zenodo.10028965 

Albayati, H. (2024). Investigating undergraduate students' perceptions and awareness of 

using ChatGPT as a regular assistance tool: A user acceptance perspective study. 

Computers and Education: Artificial Intelligence, 6, 100203. 

https://doi.org/10.1016/j.caeai.2024.100203 

Aljanabi, A. R. A., Al-Maroof, R. A., & Salloum, S. A. (2024). Students’ acceptance of 

ChatGPT in higher education: Extending the technology acceptance model. Education 

and Information Technologies, 29(5), 5331–5348. 

Al-Seghayer, K. (2024). Comparative efficacy of digital and nondigital texts on reading 

comprehension and EFL learners’ perceptions of their merits. 

https://doi.org/10.64152/10125/73589 

Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of 

information technology. MIS Quarterly, 13(3), 319–340. 

https://doi.org/10.2307/249008 

Gerlich, R. (2025). Structured prompting and reflective engagement in AI-assisted learning: 

Reducing cognitive offloading among postgraduate students. Frontiers in Education, 

10, 1398721. 

Guthrie, J. T., & Wigfield, A. (2000). Engagement and Motivation in Reading. In M. L. 

Kamil, P. B. Mosenthal, P. D. Pearson, & R. Barr (Eds.), Handbook of Reading 

Research (3rd Ed.). New York, NY: Longman. 

Hossain, K. A. (2025). Emotional Intelligence and Artificial Intelligence: Relation and 

Exploring the World of Human Touch and Machine. 

Khan, A., Alam, S., & Chen, C. (2024). Exploring learners’ experiences and perceptions of 

ChatGPT as a learning tool in higher education. Computers and Education: Artificial 

Intelligence, 6, 100219. https://doi.org/10.1016/j.caeai.2024.100219 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689
https://doi.org/10.1080/26437015.2025.2549059
https://doi.org/10.5281/zenodo.10028965
https://doi.org/10.1016/j.caeai.2024.100203
https://doi.org/10.64152/10125/73589
https://doi.org/10.2307/249008
https://doi.org/10.1016/j.caeai.2024.100219


  
 

Copyright © The Author(s) 
Vol.6, No. 4, October 2025 

e-ISSN: 2723-4126 
p-ISSN: 2776-8880 

 

436 

 

Kulal, A. (2025). Cognitive Risks of AI: Literacy, Trust, and Critical Thinking. Journal of 

Computer Information Systems, 1–13. 

https://doi.org/10.1080/08874417.2025.2582050 

Mirza, K. B., & Jabeen, M. (2025). University library leaders' technological initiatives for 

reshaping reading habits in the digital era: A twin cities case study. The Journal of 

Academic Librarianship, 51(3), 103038. https://doi.org/10.1016/j.acalib.2025.103038 

Moradi, H. (2025). Integrating AI in higher education: factors influencing ChatGPT 

acceptance among Chinese university EFL students. International Journal of 

Educational Technology in Higher Education, 22(1), 30. 

Papadopoulos, P. M., Demetriadis, S. N., Stamelos, I. G., & Tsoukalas, I. A. (2009). 

Prompting students’ context-generating cognitive activity in ill-structured domains: 

Does the prompting mode affect learning? Educational Technology Research and 

Development, 57(2), 193–210. https://doi.org/10.1007/s11423-008-9105-6 

Pérez-Mira, B., López, D., & García, J. (2025). Study on students’ use of ChatGPT in higher 

education: Benefits, costs, and attitudes. Education Sciences, 15(2), 217. 

https://doi.org/10.3390/educsci15020217 

Ravšelj, D., Keržič, D., Tomaževič, N., Umek, L., Brezovar, N., Iahad, N. A., ... & 

Aristovnik, A. (2025). Higher education students’ perceptions of ChatGPT: A global 

study of early reactions. PLoS One, 20(2), e0315011. 

https://doi.org/10.1371/journal.pone.0315011 

Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 

20(9), 676–688. https://doi.org/10.1016/j.tics.2016.07.002 

Schorr, E. (2023). A meta-analysis of TAM studies in educational technology adoption 

(2010–2022). British Journal of Educational Technology, 54(5), 1452–1471. 

https://doi.org/10.1111/bjet.13227 

She, M., Xu, Y., Luo, Y., & Hu, P. (2025). Understanding university students’ acceptance 

and use of fragmented academic reading: An extension of the UTAUT2 model. SAGE 

Open, 15(3), 21582440251377881. https://doi.org/10.1177/21582440251377881 

Sohn, H., Acosta, K., Brownell, M. T., Gage, N. A., Tompson, E., & Pudvah, C. (2023). A 

meta–analysis of interventions to improve reading comprehension outcomes for 

adolescents with reading difficulties. Learning Disabilities Research & Practice, 

38(2), 85–103. https://doi.org/10.1111/ldrp.1230 

Zhou, T. (2012). Examining location-based services usage from the perspectives of unified 

theory of acceptance and use of technology and privacy risk. Journal of Electronic 

Commerce Research, 13(2), 135. 

 

http://u.lipi.go.id/1593190689
http://u.lipi.go.id/1593190689
https://doi.org/10.1080/08874417.2025.2582050
https://doi.org/10.1016/j.acalib.2025.103038
https://doi.org/10.1007/s11423-008-9105-6
https://doi.org/10.3390/educsci15020217
https://doi.org/10.1371/journal.pone.0315011
https://doi.org/10.1016/j.tics.2016.07.002
https://doi.org/10.1111/bjet.13227
https://doi.org/10.1177/21582440251377881
https://doi.org/10.1111/ldrp.1230

