I. INTRODUCTION
The advent of artificial intelligence (AI) has transformed English as a Foreign or Second Language (EFL/ESL) education, providing new pathways for both teaching and learning. Among these advancements, tools like ChatGPT have shown considerable impact. Despite its recent introduction, ChatGPT is now widely accessible and continues to illustrate AI’s expanding role in education (Koraishi, 2023).
Previous studies have demonstrated ChatGPT’s positive impact in diverse educational domains. For instance, it was found to generate text aligned with Common European Framework of Reference for Languages (CEFR) proficiency levels and analyze grammar in student compositions, often surpassing traditional grammar-checking tools like Grammarly or ProWritingAid (Schmidt-Fajlik, 2023). In language learning, particularly in writing, ChatGPT provided insights, suggestions, and organizational support, which enhanced students’ writing skills (Su et al., 2023). Moreover, its integration into interactive platforms, such as RECIPE (Revising an Essay with ChatGPT on an Interactive Platform for EFL learners), offered students continual feedback, fostering measurable progress across an academic semester (Han et al., 2023). In Han et al.’s (2023) study, two pre-designed prompts (i.e., a hidden prompt for ChatGPT to act as an English writing teacher and an open prompt for students to work on their essays) were used to help students stay focused, although some required additional clarifications from ChatGPT. Additionally, ChatGPT addressed the challenge of generating compelling essay topics by simplifying the writing process (Hwang et al., 2023).
While research on ChatGPT’s application in various educational contexts is growing, studies specifically examining its use in high-stakes exams like IELTS remain limited. Recent investigations have demonstrated ChatGPT’s potential in academic writing, with studies showing its ability to provide constructive feedback on essay structure, coherence, and language use (e.g., Han et al., 2023; Su et al., 2023). However, these studies have primarily focused on general academic writing rather than the specific demands of standardized tests like IELTS. Given that Task 2 in the IELTS writing section requires test-takers to construct a well-developed opinion essay within a limited timeframe, it remains a significant challenge for many candidates (Arefsadr & Babaii, 2023; IELTS, n.d.). Thus, there is a pressing need to investigate how ChatGPT feedback may influence students’ performance in this context.
The International English Language Testing System (IELTS) is a globally recognized English proficiency test designed to assess non-native speakers’ abilities in Listening, Reading, Writing, and Speaking. The Writing section consists of Task 1, which requires test-takers to describe visual data, and Task 2, which involves writing an opinion-based essay in response to a given prompt. Task 2 carries more weight in the overall writing band score and assesses candidates’ ability to present arguments coherently, support ideas with reasoning, and demonstrate grammatical and lexical proficiency (IELTS, n.d.). Many EFL learners find the IELTS Opinion Essay particularly challenging, as it demands both linguistic precision and effective argumentative structuring (Arefsadr & Babaii, 2023). Research suggests that feedback plays a critical role in improving IELTS writing performance (Algburi & Razali, 2022), yet access to individualized feedback is often limited in classroom settings. Henceforth, Task 2 will be referred to as the IELTS Opinion Essay for clarity.
This study aims to bridge the research gap by exploring Uzbek EFL students’ perceptions of ChatGPT feedback for the IELTS Opinion Essay. Understanding these perceptions can inform teaching practices and enhance learning outcomes by identifying specific ways ChatGPT supports students in high-stakes writing contexts. As AI-driven feedback becomes increasingly integrated into language learning, these insights may help educators enhance IELTS writing instruction, particularly in settings where access to individualized teacher feedback is limited. Furthermore, by examining how students interact with ChatGPT’s feedback, this study could lay the groundwork for future research on optimizing AI-assisted writing support in standardized test preparation.
The study addresses the following research questions:
1. What are EFL students’ perceptions of ChatGPT as a tool for IELTS essay writing support?
2. What aspects of IELTS essay writing do EFL learners believe ChatGPT could influence or assist with?
II. LITERATURE REVIEW
1. The Role of Feedback in EFL Writing
Feedback plays a critical role in language learning by supporting learners to refine their skills and address errors effectively (Hyland, 2019). In high-stakes exams like IELTS, timely, regular, and detailed feedback significantly improves EFL learners’ writing proficiency (Algburi & Razali, 2022).
Different types of feedback—direct, indirect, written, and oral serve various roles depending on learners’ proficiency levels, nature of errors, and pedagogical objectives (Benali, 2021). Research suggests that direct feedback with explicit corrections benefits less proficient learners, while indirect feedback encourages advanced learners to engage more deeply with content (Bitchener & Ferris, 2012; Ferris & Roberts, 2001). ChatGPT, which primarily offers direct feedback by identifying and correcting specific errors, is particularly suited for IELTS writing preparation, as it addresses key aspects like grammatical accuracy and task fulfillment while providing immediate support to learners.
Many studies have investigated various feedback mechanisms and their impacts on EFL learners. For instance, Zhang and Zhang (2022) found that teacher and peer feedback positively influence the use of metacognitive strategies in EFL writing, whereas automated feedback may hinder monitoring strategies. Gonzalez-Torres and Sarango (2023) reported that both teacher and peer feedback effectively enhance high school students’ writing skills, with students expressing a preference for a combination of both. Wanchid (2020) compared self-correction, paper-pencil peer feedback, and online peer feedback, concluding that different feedback types and learners’ English proficiency levels significantly affect writing achievement. Küçükali (2017) observed that oral feedback, through meaningful interaction and negotiation between students and teachers, significantly improves EFL learners’ writing performance compared to traditional written feedback.
These studies underscore the multifaceted role of feedback in EFL writing instruction. While automated feedback offers immediacy and consistency, teacher and peer feedback provide nuanced insights that can foster metacognitive strategy use and self-regulation among learners. Therefore, a balanced integration of various feedback types may be most effective in supporting EFL learners’ writing development.
2. From Automated Feedback Systems to AI-Driven Tools
Automated Writing Evaluation (AWE) systems, like Criterion and My Access, were created to provide consistent, instant feedback on grammar, syntax, and content by reducing the feedback burden on teachers (Woodworth & Barkaoui, 2020). However, traditional AWE systems lacked the capacity for nuanced interpretation, sometimes overwhelming students with overly mechanical feedback which did not address individual learner needs (Stevenson & Phakiti, 2014). These limitations have sparked interest in AI-driven tools like ChatGPT which is designed using deep learning techniques to offer personalized and context-sensitive feedback (Hwang, 2023).
Research supports ChatGPT’s potential to enhance language learning. In Ali et al.’s (2023) study, data from teachers and students using ChatGPT revealed a positive impact on reading and writing skills. Li et al. (2023) also found that ChatGPT effectively explains vocabulary and phrase usage, enabling learners to make better word choice and express ideas more naturally. These features are especially valuable for high-stakes exams like IELTS, where clear, precise language and strong task fulfillment are critical. Furthermore, ChatGPT provides detailed grammatical and structural suggestions by making it a particularly effective tool for enhancing writing skills in exam contexts. While recent studies (Li et al., 2023) have examined ChatGPT’s benefits for academic writing, they primarily focus on general writing proficiency and language enhancement rather than standardized test-specific writing requirements. This study, however, specifically investigates how ChatGPT feedback aligns with IELTS Opinion Essay scoring criteria and whether it effectively addresses the unique challenges EFL learners face in this high-stakes test context. By shifting the focus from general writing support to exam-oriented strategies, this research examines how AI-generated feedback can enhance test performance and inform more targeted preparation methodologies.
Given these potential benefits for IELTS preparation, it is also important to consider how ChatGPT compares to earlier AWE systems. By addressing the shortcomings of AWE systems, which often deliver standardized and rigid feedback, ChatGPT demonstrates greater adaptability by simulating human interaction. This flexibility allows it to provide responses specific to individual learner needs, a crucial factor for effective language learning (Kim et al., 2023). However, it is important to acknowledge that while AI-generated feedback offers higher levels of personalization, it still falls short of replicating the depth and nuance of feedback from human instructors, particularly when addressing complex grammatical or stylistic issues (Marzuki et al., 2023; Wang et al., 2024).
3. Advantages and Limitations of AI Feedback in EFL Writing
While AI tools like ChatGPT offer many benefits, including instant feedback and the ability to handle large volumes of student writing, concerns about students’ over-reliance on these tools have been raised. Marzuki et al. (2023) caution that dependence on AI tools may hinder the development of critical thinking skills, as students may rely too heavily on automated corrections without deeply engaging with their writing. Additionally, Wang et al. (2024) observe that students using ChatGPT sometimes become distracted by social media platforms (e.g., Instagram, WhatsApp, Facebook), further diminishing their focus and productivity during writing tasks.
Despite these concerns, ChatGPT’s structured feedback and adaptability demonstrate considerable potential in supporting students’ development of key writing skills. Its effectiveness can be amplified through the use of carefully designed prompts that provide clarity and focus for writing tasks. Koraishi (2023) highlights that well-constructed prompts align AI feedback with students’ needs and expectations. Such prompts can also reduce cognitive load for lower-proficiency learners by making it easier to elicit targeted responses. Similarly, Hwang et al. (2023) emphasize that effective prompts establish clear context and direction, ensuring that ChatGPT’s feedback aligns closely with the user’s goals and task requirements. These findings suggest that prompts designed specifically for academic writing, particularly in high-stakes contexts like IELTS, can positively impact the quality and relevance of AI-generated feedback.
While this study does not aim to evaluate the specific effectiveness of the pre-designed prompt itself, its inclusion in the methodology aligns with existing research emphasizing the role of prompts in improving AI-assisted learning outcomes. By providing structure and guidance, the prompt helps ensure that ChatGPT’s feedback addresses critical components such as argument structure, coherence, and linguistic accuracy in the IELTS Opinion Essay.
Research on the application of AI tools like ChatGPT in high-stakes assessments such as IELTS remains limited. This study seeks to bridge this gap by exploring how Uzbek EFL students perceive ChatGPT’s feedback for IELTS Opinion Essay writing. By incorporating a pre-designed prompt as part of the methodology, this research provides valuable insights into the potential of structured guidance to enhance ChatGPT’s feedback and support academic writing tasks in high-stakes contexts.
III. METHOD
1. Participants
The study took place in December 2023 and was conducted online using voluntary response sampling. The participants of this study were 26 EFL students preparing for the IELTS exam at a university language center in Uzbekistan. These students, aged 17 to 29, were provided with a detailed explanation of the study’s purpose, procedures, and confidentiality measures before voluntarily agreeing to participate. Considering the inclusion of minors, parental consent was also secured through coordination with the language center. Their motivations for studying IELTS included goals such as pursuing higher education abroad and improving employment prospects.
The sample represented a diverse academic background: 15.4% (n = 4) were secondary school students, 34.6% (n = 9) were undergraduates, 42.3% (n = 11) were graduate students, and 7.7% (n = 2) were postgraduates. Gender distribution was relatively balanced, with 53.8% female (n = 14) and 46.2% male (n = 12).
In terms of self-perceived English writing proficiency, 15.4% (n = 4) rated themselves as intermediate, 57.7% (n = 15) as upper-intermediate, 26.9% (n = 7) as advanced. To provide consistency in interpretation, participants were advised to consider the CEFR framework (e.g., B1 for intermediate, B2 for upper-intermediate, C1 for advanced) or their prior IELTS scores, if available. Most participants (73%, n = 19) had prior experience with high-stakes language exams, and their self-reported IELTS writing scores ranged from 5.5 to 7.0 (out of a maximum of 9.0). This range generally corresponds to upper-intermediate to advanced levels in the IELTS scoring system. Thus, while self-reported, the proficiency levels appear to align reasonably well with the participants’ IELTS writing scores.
2. Data Collection Procedures
1) The Prompt Development Process
At the outset of the experiment, a pre-designed prompt was developed by the researchers to ensure the students could obtain accurate and detailed feedback from ChatGPT on their IELTS essays. While many of them had prior experience using ChatGPT for general academic purposes, their familiarity with crafting effective prompts specifically for IELTS writing was limited. The students had primarily relied on instructor-provided feedback in their learning experiences and were unlikely to have developed the metacognitive skills necessary to maximize ChatGPT’s feedback through self-directed queries.
Without clear guidance, the students might face challenges in eliciting the specific and actionable feedback required for essay improvement. This issue is particularly critical for high-stakes writing tasks like the IELTS Opinion Essay, where structured feedback on argumentation, cohesion, and linguistic precision is essential. Recognizing these challenges, the researchers sought to develop a well-structured prompt that would enable the students to obtain more precise and useful feedback from ChatGPT. To ensure that the prompt was aligned with IELTS evaluation standards, the researchers first examined existing IELTS writing resources before constructing the final version. As part of this process, they conducted an initial test using an IELTS Opinion Essay available on the website of IELTS educator Liz (n.d.). While Liz is not officially affiliated with the IELTS testing organization, her instructional materials are widely used by IELTS candidates and structured in alignment with IELTS writing assessment criteria.
To evaluate the effectiveness of ChatGPT’s feedback, the researchers compared its assessment of an essay from Liz’s (n.d.) website under two conditions: first, using only a general instruction to evaluate the essay without specific guidelines, and second, applying the detailed, pre-designed prompt. In the first scenario, ChatGPT identified areas for improvement across most parts of the essay and assigned it an approximate band score of 7.0. In contrast, when the researchers used the pre-designed prompt, ChatGPT provided more targeted feedback that was better aligned with IELTS scoring criteria, assigning a band score of 8.0. This was compared to the 9.0 rating given by an IELTS examiner for the same essay.
To further assess the reliability of Liz’s (n.d.) materials, this comparison was repeated multiple times using different essays from her website. The consistency of results across these trials demonstrated that her essays align closely with IELTS scoring standards. These findings validated the use of her materials as a benchmark for developing the pre-designed prompt. As a result, a structured and well-crafted prompt enabled ChatGPT to offer evaluations and feedback that were more consistent with expert assessments, thereby improving its reliability in this context. Consequently, the pre-designed prompt (see Appendix 11) was provided to the students to ensure they could effectively use ChatGPT to evaluate their essays and receive targeted feedback, even without extensive training or experience in creating prompts.
2) Introductory Zoom Session and Prompt Modeling
To prepare the students for using ChatGPT effectively in the context of IELTS Opinion Essays, a one-hour introductory Zoom session was conducted. The primary aim of this session was to familiarize the students with the study’s procedures and demonstrate how to use the pre-designed prompt to obtain actionable feedback from ChatGPT on their essays. The researchers facilitated the session, guiding the students through the process of using ChatGPT effectively and ensuring they understood how to interact with the tool for writing improvement. This included overseeing the session, modeling sample activities, and addressing any questions or difficulties the students encountered in following the directions. During the session, the researchers did the following: (1) provided an overview of how ChatGPT could support their IELTS writing preparation, (2) demonstrated the use of the pre-designed prompt by inputting a sample IELTS Opinion Essay into ChatGPT and presenting the kind of feedback the students could expect, and (3) emphasized the importance of crafting clear and specific queries to maximize the effectiveness of ChatGPT’s feedback.
The students were encouraged to ask questions throughout the session to clarify any uncertainties. To accommodate those who wished to revisit the session, a recording was made available for later viewing.
3) EFL Students’ Use of ChatGPT for IELTS Opinion Essay Writing Feedback
After the Zoom session, the students were tasked with independently applying the skills they had learned. They were encouraged to engage fully with ChatGPT during the feedback process, including asking follow-up questions to clarify or expand on the feedback received and exploring the tool’s capabilities for iterative improvements in their writing.
To complete this task, the students were provided with three IELTS Opinion Essay prompts designed to reflect the typical requirements of Task 2 writing. These prompts were selected to cover common response types, including: (1) identifying causes and proposing solutions, (2) evaluating contrasting views and expressing an opinion, and (3) taking a position and justifying it with arguments. The essay prompts were as follows:
① Discuss reasons for the decline in animal and plant populations in many countries and propose solutions to address the issue.
② Compare the effectiveness of punishment versus rewards in teaching children good behavior and provide your opinion.
③ Evaluate whether universities should emphasize practical training over theoretical knowledge and justify your stance.
To ensure clarity and consistency in completing the task, the students were instructed to: (1) write an essay based on one of the given topics, (2) use the pre-designed prompt to obtain feedback from ChatGPT on their essay, and (3) save their ChatGPT interaction history, including the feedback received and any additional questions or responses exchanged with ChatGPT, for submission at the end of the study. Both the essay and the complete interaction history were submitted for review.
4) A Survey Questionnaire
Following the completion of their essays and ChatGPT interactions, all study participants (n = 26) were invited to complete a survey questionnaire, which was shared via a link (see Appendix 22). The survey, created on Google Forms, consisted of 13 main questions with additional sub-questions and was designed to collect insights on the students’ experiences and perceptions. It was divided into five thematic sections, each targeting specific aspects of the study.
The first section, Demographic and Academic Background (Questions 1-7), gathered details about the students’ profiles, such as age, gender, education level, and self-assessed English proficiency. These questions aimed to capture the students’ individual characteristics and learning contexts.
The second section, Previous AI Experience and Usage (Questions 1.1-2), explored the students’ familiarity with AI tools, particularly ChatGPT. It included questions about their general usage patterns and any prior applications of ChatGPT for academic or language learning purposes.
The third section, Perceptions of ChatGPT’s Feedback on IELTS Opinion Essays (Questions 3-10), focused on the students’ evaluations of ChatGPT’s usefulness for their IELTS writing tasks. This section addressed areas such as task response, coherence and cohesion, lexical resource, and grammatical range and accuracy. The students were asked to reflect on the relevance and helpfulness of the feedback provided in each of these areas.
The fourth section, Perceived Effectiveness for IELTS Writing Skill Development (Questions 11.1-11.3), assessed the students’ perspectives on how ChatGPT’s feedback contributed to their essay-writing abilities. This included their satisfaction with the feedback and its perceived benefits in preparing for high-stakes exams like IELTS.
The final section, Future Usage Intentions and Additional Comments (Questions 12-13), explored the students’ intentions to continue using ChatGPT for IELTS preparation. It also invited them to provide additional insights, reflections, or suggestions regarding their experiences with the tool.
The survey provided a structured approach to capturing the students’ insights on ChatGPT’s utility in IELTS writing preparation and contributed to a comprehensive understanding of their interactions with the tool.
3. Data Analysis
This study utilized a mixed-methods approach, combining both quantitative and qualitative research methods to explore Uzbek EFL students’ perceptions of the effectiveness of ChatGPT as an assistant tool in IELTS writing essays. To ensure comprehensive data collection, participants were instructed to submit both their essays and the full interaction history with ChatGPT, including any follow-up questions or responses. Based on these collected responses, the following analysis was conducted to address the research questions.
To answer the first research question, which explored the students’ perceptions of ChatGPT as a tool for IELTS essay writing support, survey responses to Questions 1, 2 (prior experience with AI tools and satisfaction), 3 (helpfulness of ChatGPT feedback in IELTS writing), 9 (challenges faced while using ChatGPT), 11.1, 11.3 (perceived helpfulness of ChatGPT in improving IELTS writing skills and the reasons for their views), 12 (intention to continue using ChatGPT), and 13 (general comments about the experience of using ChatGPT for IELTS writing feedback) were analyzed. It is important to note that while Questions 3 and 11.1 may seem similar, they serve distinct purposes in this study. Question 3 captures the students’ immediate perceptions of ChatGPT’s feedback on a specific writing task, while Question 11.1 assesses broader beliefs about ChatGPT’s potential to improve IELTS writing skills over time. This distinction enables a comparative analysis of short-term perceived usefulness and long-term expectations, contributing to a more comprehensive understanding of AI-assisted feedback in IELTS preparation.
Close-ended questions were processed using SPSS (version 25) to calculate descriptive statistics, including frequencies, means, and standard deviations. Open-ended responses were reviewed, categorized, and thematically analyzed to identify common perceptions, with selected comments included to provide qualitative insights that contextualized the quantitative results.
To answer the second research question, which investigated the specific aspects of IELTS writing influenced by ChatGPT feedback, survey responses to Questions 4.1, 4.2 (whether follow-up questions were asked after receiving feedback and reasons behind them), 5 (most helpful follow-up questions they asked), 6 (reasons for not asking follow-up questions), 7 (beliefs about ChatGPT’s accuracy in assessing essays based on IELTS criteria), 8 (the specific IELTS writing criteria for which ChatGPT’s feedback was perceived as most accurate and useful), 10 (overall satisfaction with ChatGPT’s assessment based on IELTS criteria) and 11.2 (perceptions of IELTS writing skills which ChatGPT can help with) were analyzed. Descriptive statistics were calculated for close-ended questions, while open-ended responses were reviewed and categorized to identify themes related to ChatGPT’s role in enhancing various aspects of writing. Additionally, interaction logs between the students and ChatGPT were examined to identify patterns in follow-up questions and the feedback received. These logs supplemented the survey data, particularly for Questions 7, 8, and 10, which focused on feedback quality and its alignment with IELTS criteria.
By integrating quantitative and qualitative data, the researchers attempted to provide a comprehensive understanding of the students’ experiences. Qualitative insights complemented the statistical results by adding context and emphasizing how ChatGPT influenced the students’ perceptions and writing, particularly in relation to IELTS-specific criteria.
IV. RESULTS AND DISCUSSION
1. Perceptions of ChatGPT as a Tool for IELTS Essay Writing Support
The first research question sought to explore the students’ opinions and beliefs about ChatGPT as an assistant tool for IELTS writing essays. Specifically, the study examined their previous experiences with AI tools, their satisfaction with ChatGPT’s feedback, and their overall impressions of its usefulness and ease of use.
1) Students’ Prior Experience With AI Tools
An analysis of the students’ responses (Questions 1-2) revealed that the majority (69.2%, n = 18) had prior experience with AI tools, predominantly using ChatGPT, while 30.8% (n = 8) reported minimal or no exposure to such tools. The students who had used ChatGPT highlighted its value in academic contexts, reflecting their familiarity and confidence in harnessing AI for learning. Despite varying levels of experience, satisfaction with ChatGPT was relatively high, with a mean score of 3.92 (SD = 1.29). This suggests that even among those with limited exposure to AI tools, ChatGPT was regarded as a useful and effective resource.
2) Helpfulness of ChatGPT for IELTS Writing
The students’ perceptions of ChatGPT’s helpfulness for IELTS writing were measured through their responses to Questions 3 and 11.1. As summarized in Table 1, the students generally viewed ChatGPT favorably for its ability to provide clear, structured, and actionable feedback.
The descriptive statistics for Question 3 (M = 4.04, SD = 1.22) indicate that while the students mostly found ChatGPT’s feedback helpful, there was some variability in individual experiences. On the other hand, responses to Question 11.1 (M = 4.27, SD = 0.83) demonstrated stronger agreement regarding ChatGPT’s potential to improve their IELTS writing skills. The lower standard deviation in Question 11.1 reflects more consistent satisfaction with ChatGPT’s capacity to support writing improvement and highlights its perceived reliability for addressing their learning needs.
This high satisfaction of ChatGPT’s feedback appears to stem from several key characteristics. First, clarity was frequently highlighted by the students, as one noted, “All the feedback was clear-cut for me” (S12). Second, the structured nature of the feedback was reflected in responses that emphasized its logical organization, such as S17’s comment, “Because its answers and feedback were accurate enough,” and S19’s statement, “As my writing proficiency is not so high, the feedback given by ChatGPT was enough for me.” These responses suggest that the feedback followed a systematic evaluation process that aligned with the students’ comprehension levels. Third, multiple students noted that ChatGPT’s structured feedback allowed them to easily pinpoint areas for improvement. Finally, the actionable nature of the feedback was indicated by comments like “Because the feedback I got was enough to me to realize my weaknesses” (S4), emphasizing that the students were able to directly implement ChatGPT’s suggestions to refine their essays. This perception of structured and actionable feedback was further supported by the students who described how ChatGPT’s evaluation method helped them assess their work based on predefined criteria, making the revision process clearer and more effective.
Overall, most students found ChatGPT’s feedback beneficial for their writing development, citing its clarity, structured format, and practical suggestions. However, a few students expressed concerns about its limitations. Interestingly, almost no one believed ChatGPT would be of little help (Question 11.3), except for one student who commented, “As it is AI, I don’t think that it knows better than teachers” (S22). This may reflect unfamiliarity with the tool or a preference for traditional teacher feedback.
3) Challenges With ChatGPT Feedback
Most students reported a smooth experience when using ChatGPT for their IELTS writing preparation (Question 9), with 84.6% (n = 22) indicating that they encountered no difficulties. However, a small portion of students (15.4%, n = 4) faced challenges, primarily related to technical issues or unrelated problems. For instance, one student noted difficulty when ChatGPT was used outside the writing context, commenting, “When it was a speaking partner (not writing) about fictional book analysis, it could not manage to describe the process in the book” (S23). These findings suggest that while ChatGPT generally supports the students effectively, its limitations in non-writing scenarios may influence their perceptions.
4) Future Intentions to Use ChatGPT
The students overwhelmingly expressed a willingness to continue using ChatGPT for IELTS preparation, particularly for the writing section (Question 12). Only one student (S22) maintained reservations about AI’s ability to surpass human teachers (Question 13). Conversely, the majority of students’ responses (as evidenced in responses to Questions 7, 11.1, and 12) suggested ChatGPT’s reliability and usefulness. They highlighted its role in improving writing skills (“I think ChatGPT can greatly help me improve my writing skills, especially about IELTS writing Task 2,” S1, Q11.1), its accessibility (“I recommend using ChatGPT for checking one’s IELTS writing, because you can get feedback anytime easily.” S5, Q12), and the quality of its feedback (“It is real checker.” S3, Q11.1). These positive perceptions align with the students’ high satisfaction ratings for ChatGPT’s feedback (Question 3, M = 4.04, SD = 1.22; Q11.1, M = 4.27, SD = .83), which further demonstrates its potential as a reliable support tool for high-stakes exam preparation.
The findings align with prior research on AI-assisted writing tools, which suggests that automated feedback can support students in refining their writing rather than directly improving overall writing proficiency (Ali et al., 2023; Li et al., 2023). This study provides insights into how students perceive ChatGPT’s role as a supportive tool for IELTS writing preparation. Some students explicitly noted that ChatGPT’s feedback was clear and easy to follow, with one stating, “ChatGPT’s feedback was straightforward and made it easier to see where I needed to improve” (S1, Q11.1). These responses suggest that students valued the accessibility and structured nature of the feedback.
However, some students voiced concerns about ChatGPT’s limitations compared to human instructors, particularly in addressing more nuanced feedback, echoing challenges identified in previous studies (Kim et al., 2023). Despite these reservations, the strong intent to continue using ChatGPT suggests its perceived reliability and effectiveness as a resource for IELTS preparation. Its ability to provide immediate, structured feedback makes it a valuable supplementary tool, especially for students aiming to improve their writing skills independently.
2. ChatGPT’s Impact on Specific Aspects of IELTS Writing
The second research question aimed to identify the specific areas of IELTS writing—such as task response, coherence and cohesion, lexical resource, and grammatical range—where ChatGPT’s feedback was perceived to have the greatest impact.
1) ChatGPT’s Evaluation Accuracy
The analysis, which focused on how the students perceived ChatGPT’s ability to evaluate essays based on IELTS Writing assessment criteria (Question 7), revealed that a majority of students (73.1%, n = 19) believed ChatGPT provided accurate evaluations. Meanwhile, 11.6% (n = 3) expressed neutrality, and 15.4% (n = 4) voiced skepticism. This skepticism often stemmed from misconceptions about IELTS examiners being subjective or doubts about AI’s ability to replicate human judgment effectively (see Table 2).
Among the criteria, Task Response was the most frequently highlighted, with the students (S5, S11, S15, S19, S25, S26) appreciating ChatGPT’s ability to identify weaknesses and suggest improvements. These students specifically mentioned that ChatGPT’s feedback helped refine their arguments and better address the task prompt.
Meanwhile, responses from S15 and S19 were more general in nature and did not explicitly reference Task Response. Instead, their comments reflected broader perceptions of ChatGPT’s assessment accuracy across multiple criteria. As such, these responses have been reclassified accordingly to avoid misrepresentation.
However, other criteria, such as Coherence and Cohesion (S5, S6, S15, S19), Lexical Resource (S5, S15, S19), and Grammatical Accuracy (S5, S15, S19, S24), were also valued. For example, S5 commented, “ChatGPT can provide a full assessment along with the weaknesses in the essay if a proper prompt of assessment criteria is provided.” It is also essential to note that few comments (S5, S15, S19) were wide-ranging, covering multiple aspects of IELTS Opinion Essay writing rather than being limited to a single criterion.
To illustrate this point, Figure 1 shows an excerpt from Student 5’s essay along with ChatGPT’s feedback. While the feedback provides balanced insights across all four IELTS criteria, it particularly underscores areas for improvement under Task Response. Specifically, it suggests further development in proposed solutions and more work on the underlying reasons, noting that “the prompt is adequately addressed,” rather than perfectly. These are major issues that need to be mended whereby the Task Response can be elevated to a higher score. In contrast, Coherence and Cohesion and Lexical Resource were assessed more positively with minor mistakes pointed out, reflecting strengths in organization and vocabulary use.
The comparison demonstrates ChatGPT’s detailed alignment with IELTS assessment criteria, offering constructive feedback tailored to the essay’s strengths and weaknesses of the essay. This level of feedback not only supports the student’s reflections on Task Response but also reinforces the tool’s utility in providing actionable insights across other criteria like Coherence and Cohesion and Lexical Resource.
However, while Task Response was frequently cited as a strength, the students’ overall satisfaction with ChatGPT’s assessment varied. Responses to Question 10 (M = 2.77, SD = 1.18) indicate a low level of satisfaction with ChatGPT’s evaluation accuracy, though not overwhelmingly negative. Some students expressed dissatisfaction, while others provided neutral responses, reflecting mixed perceptions of ChatGPT’s effectiveness in assessment accuracy. This variability highlights the need to further explore factors influencing students’ views on AI-generated feedback, including expectations, prior experiences with human feedback, and individual preferences. When comparing ChatGPT’s overall assessment with the ratings of other satisfaction factors, its rating was significantly lower. This suggests that while the students found certain aspects of ChatGPT’s feedback useful (such as grammar correction or coherence-related suggestions), they were more critical of its overall assessment accuracy. Higher expectations for AI-driven evaluation, along with the complexity of IELTS writing tasks, may have contributed to this discrepancy.
This discrepancy can be examined in three main ways. Firstly, while the students found specific aspects of ChatGPT’s feedback highly satisfactory (such as grammar correction or vocabulary enhancement), they might have been more critical when considering its overall effectiveness. Secondly, the students may have had higher expectations for ChatGPT’s overall performance, leading to a more stringent evaluation. Finally, the complexity of IELTS writing tasks might have influenced the students’ perceptions of ChatGPT’s overall usefulness, despite finding it helpful in specific areas. Therefore, this finding highlights the importance of considering both specific and overall evaluations when assessing the effectiveness of AI tools in educational contexts.
2) Perceived Strengths in Feedback
The analysis, which focused on identifying the areas of IELTS writing the students believed ChatGPT was most effective (Question 8), revealed that Task Response was particularly emphasized. Table 3 highlights these perceptions, illustrating the aspects of IELTS essay evaluation where areas where ChatGPT’s feedback was considered most impactful.
The findings indicate that Task Response was perceived as ChatGPT’s strongest area, with 53.8% (n = 14) of the students acknowledging its effectiveness in addressing essay content and relevance to the prompt. This aligns with earlier findings from the students’ qualitative responses (see Table 2), where many valued ChatGPT’s ability to highlight weaknesses and suggest improvements in this criterion.
Additionally, Coherence and Cohesion was identified by 26.9% (n = 7) of the students, indicating the tool’s utility in guiding logical structuring and the organization of ideas. Although Grammatical Range (11.5%, n = 3) and Lexical Resource (7.7%, n = 2) were mentioned less frequently, the results highlight ChatGPT’s broad applicability in providing feedback across diverse writing elements.
3) Follow-Up Questions and Their Themes
The analysis of Question 4.1 and 6 indicates that while most students did not ask follow-up questions after receiving ChatGPT’s feedback, this does not necessarily imply that they found the feedback clear or comprehensive. Specifically, 38.4% (n = 10) of students sought additional clarification or guidance, primarily for grammar-related or task-related issues (see Table 4). This suggests that while the pre-designed prompt effectively guided ChatGPT to deliver structured feedback, some students still desired more specific explanations in certain areas.
The follow-up question (Question 4.2) was related to grammar and task clarity. For example, one student (S14) asked ChatGPT to explain a specific grammar rule relevant to their essay, while another (S11) sought clarification on how to better meet the task requirements of the IELTS Opinion Essay.
In contrast, 61.5% (n = 16) of students did not ask follow-up questions (Question 6). The majority of these students explicitly stated ChatGPT’s feedback met their needs, as evidenced by responses such as: “It pointed out my exact weaknesses and offered clear fixes, so I didn’t have any more questions” (S2), “I am satisfied with ChatGPT response” (S7), “Everything is quite clear and there is no need to ask extra questions” (S14), and “Explanation was clear and fair” (S16). Additionally, two students (S19, S21) indicated that their limited English proficiency prevented them from asking further queries.
Among the 10 students who posed follow-up questions, most (n = 4) found inquiries related to essay structure and writing skill improvement to be the most helpful. Others highlighted the usefulness of ChatGPT’s feedback on grammatical accuracy (n = 3) or valued its general learning recommendations (n = 1). Additionally, two students (n = 2) appreciated all types of feedback equally, reflecting ChatGPT’s ability to cater to diverse learning needs (see Table 5).
These observations align with Koraishi’s (2023) findings, which emphasize the significance of structured prompts in optimizing AI-generated feedback. These findings do not conclusively establish that ChatGPT’s feedback clarity alone minimized follow-up questions. While some students explicitly stated that the feedback was sufficient, others may have had different reasons for not engaging further, such as a preference for independent learning or difficulty in formulating follow-up questions.
4) Perceptions of Writing Improvement
When the students were asked if they believed ChatGPT could improve their IELTS writing skills (Question 11.1), responses were overwhelmingly positive (see Table 1). To further explore this perception, the students were invited to elaborate on how ChatGPT could support their writing development (Question 11.2). Table 6 summarizes their comments, with 19 out of 26 students (73.1%) providing detailed responses. The most frequently cited benefit was Feedback and Error Identification (n = 7), followed by Vocabulary and Grammar Improvement (n = 5) and Idea Generation and Development (n = 4). Fewer students highlighted Writing Structure and Style Guidance (n = 2) and General Practice and Learning (n = 1), reflecting the tool’s versatility in addressing diverse aspects of IELTS essay writing. These findings indicate that EFL learners perceive ChatGPT as a multifaceted resource capable of enhancing both the linguistic and conceptual dimensions of their writing. This diversity of responses illustrates the broad applicability of ChatGPT as a resource for both linguistic and conceptual enhancement.
The comments gathered reflect how Uzbek EFL students view ChatGPT as a reliable tool for addressing their writing challenges. Its strengths in providing clear feedback on task response and coherence were particularly valued, alongside its role in supporting grammar and vocabulary improvements. These perceptions align with Shin et al. (2023), who highlight the importance of targeted feedback in fostering writing proficiency.
However, some students noted ChatGPT’s limitations in identifying advanced grammatical errors. For instance, Student 6 commented, “ChatGPT feedback is sometimes not very accurate. While it helps with basic grammar mistakes, it does not always recognize more complex sentence structures or subtle grammar mistakes that a human examiner might notice.” These concerns align with prior research (Kim et al., 2023) that highlights the limitations of AI in detecting complex linguistic nuances. Despite these concerns, the overwhelmingly positive responses underscore ChatGPT’s potential as a valuable resource for foundational writing skill development.
Additionally, the ease and clarity of ChatGPT’s feedback suggests that structured prompts may enhance the feedback process, particularly for EFL learners preparing for high-stakes assessments. However, further research is needed to directly examine the role of structured prompts in optimizing AI-generated feedback.
V. CONCLUSION
This study examined Uzbek EFL students’ perceptions of ChatGPT as a support tool for IELTS Opinion Essay writing and explored the aspects of writing that ChatGPT feedback most effectively influenced. The findings revealed that most students, including those with intermediate to advanced proficiency and varying levels of prior AI experience, held positive perceptions of ChatGPT. They particularly valued its ability to deliver clear, actionable, and structured feedback across key areas such as Task Response, Coherence and Cohesion, Grammar, and Vocabulary. However, a few students expressed concerns about ChatGPT’s limitations in replicating the detailed feedback typically provided by human instructors, especially for complex grammatical and stylistic issues.
1. Educational and Practical Implications
The findings of this study highlight several educational implications for integrating ChatGPT into EFL instruction. ChatGPT can function as a supplementary feedback tool, supporting students’ writing development by providing immediate, structured, accessible, and personalized guidance, particularly in contexts where teacher feedback is limited (Han et al., 2023; Hwang et al., 2023). This is especially valuable in large classrooms where instructors may struggle to provide individualized feedback to each student, as AI-generated responses can help bridge the feedback gap (Ali et al., 2023). Additionally, ChatGPT can foster learner autonomy by encouraging students to take an active role in revising and improving their writing based on AI-generated suggestions (Wang et al., 2024). By engaging in iterative feedback processes, students may develop stronger self-editing skills, promoting independent and reflective writing practices (Kim, et al., 2023).
Beyond its general application in EFL instruction, ChatGPT’s structured feedback may align with standardized writing assessment criteria, potentially making it a useful resource for test preparation courses such as those focusing on IELTS (Schmidt-Fajlik, 2023). However, to maximize its benefits, instructors should guide students in critically evaluating AI-generated feedback, ensuring they do not become overly reliant on the tool but rather use it as a complement to human instruction (Marzuki et al., 2023). Future pedagogical frameworks could explore blended approaches that integrate ChatGPT feedback with teacher intervention, striking a balance between AI efficiency and human expertise in language education (Koraishi, 2023).
Additionally, this study provides several practical implications for IELTS writing instruction. Educators can enhance ChatGPT’s role in IELTS preparation by incorporating AI-assisted review sessions that focus on the areas students find most beneficial. This could include developing targeted prompts that help students receive specific, constructive feedback on essay structure, argumentation, and language use. Training students to formulate effective prompts for AI interaction can improve the quality of feedback they receive, fostering greater autonomy in their writing development.
Furthermore, instructors could design scaffolded learning experiences where students first receive ChatGPT-generated feedback and then participate in reflective discussions about the suggestions. Another approach could be the integration of ChatGPT feedback into blended learning models, where students first analyze AI-generated feedback before refining their essays through peer review and instructor guidance. These strategies may lead to more personalized and efficient IELTS writing instruction, helping students develop metacognitive skills and improve their performance in high-stakes exam contexts.
2. Limitations and Future Directions
Several limitations were identified in this study. First, the relatively small sample size of 26 participants limits the generalizability of the findings. Additionally, all participants were Uzbek EFL learners from a single language center in Uzbekistan, which may not fully represent the diverse experiences of EFL learners from different regions or educational settings. Future studies could include a larger and more diverse participant pool to better capture variations in experiences and perceptions, particularly across different EFL contexts.
Second, the study focused primarily on the students’ immediate perceptions of ChatGPT feedback rather than its long-term impact on measurable outcomes, such as measurable improvements in IELTS writing scores. Future research could examine the sustained effects of ChatGPT use on writing proficiency and performance in high-stakes assessments like the IELTS.
Third, while this study utilized a pre-designed prompt to ensure consistent feedback, future research could investigate how pre-designed versus self-generated prompts influence the nature of feedback and student engagement. Such research could provide insights into optimizing prompting strategies for writing development. Furthermore, exploring the integration of ChatGPT feedback with human input could address advanced writing components, such as tone, style, and creativity, which remain challenging for AI tools to handle independently.
Finally, this study did not conduct a reliability test (e.g., Cronbach’s alpha) for the questionnaire, as its primary aim was to explore students’ perceptions rather than measure a specific skill or trait that requires statistical validation. Given the exploratory nature of this research, the survey was designed to gather qualitative insights rather than produce standardized statistical measurements. Future studies incorporating larger and more diverse participant samples may consider reliability testing to ensure consistency in findings.
Despite its limitations, this study offers valuable insights into the pedagogical potential of ChatGPT as a supplementary tool for EFL learners preparing for high-stakes exams like the IELTS. Its ability to provide immediate, structured feedback makes it particularly useful for students with limited access to human instructors or those requiring additional practice outside the classroom. Moreover, the findings emphasize the importance of well-designed prompts in optimizing AI feedback for learners, simplifying interactions, and supporting more effective learning experiences (Hwang et al., 2023; Koraishi, 2023).
In conclusion, this study highlights ChatGPT’s promise as a practical and accessible tool for EFL learners aiming to enhance their writing skills for high-stakes exams. While it addresses foundational aspects of writing effectively, further advancements in prompt design and AI-human feedback integration could maximize its potential to support advanced writing skills in language education.
PDF Links
PubReader
ePub Link
Full text via DOI
Download Citation
Print
