ChatGPT Integration in Higher Education for Personalized Learning, Academic Writing, and Coding Tasks: A Systematic Review

Abstract

The emergence of ChatGPT in higher education has raised immense discussion due to its versatility in performing tasks, including coding, personalized learning, human-like conversations, and information retrieval. Despite the rapidly growing use of ChatGPT, a dire need still exists for an overarching view regarding its role and implications in educational settings. Following the PRISMA guidelines, this study represents a systematic review of 26 articles exploring the use of ChatGPT in academic writing, personalized learning, and code generation. The relevant literature was identified through electronic databases, including Scopus, ACM Digital Library, Education Research Complete, Computers & Applied Sciences, Web of Science, and IEEE Xplore. Key details from each article were extracted and synthesized narratively to provide insights into ChatGPT’s efficacy in academic writing, personalized learning, and coding. The findings indicate that ChatGPT enhances tailored learning by adapting delivery methods to individual needs, supports academic writing through error detection and content refinement, and assists in coding by offering clarifications and reusable code snippets. However, there are concerns over its ethical implications, including the impact on academic integrity, overreliance by students on AI, and privacy concerns about data use. Based on these insights, this study proposes recommendations for the ethical and responsible integration of ChatGPT into higher education, ensuring its utility while maintaining academic integrity. In addition, the results are discussed based on the relevant learning theories to understand how students engage with, learn through, and adapt to AI technologies such as ChatGPT in educational contexts.

1. Introduction

ChatGPT has become a focal point of discussion in higher education in recent years. Developed by OpenAI, this advanced language model has attracted attention because of its diverse capabilities, including proficiency in essay writing, assistance with coding tasks, human-like conversation engagement, and seamless information retrieval [1]. These capabilities raise a number of questions concerning the effectiveness of ChatGPT in an academic environment and the ethical consequences of its use. A student’s learning experience quality impacts their academic performance level and satisfaction with their educational journey. This will further enhance learning experiences for students by offering them more personalized opportunities to learn through ChatGPT [2] as personalizing education can improve learning outcomes [3]. This personalized approach may enrich academic writing by offering tailored feedback and further support coding tasks with customized help in code generation and solving programming challenges, improving student learning outcomes in both areas.

Effective writing support is crucial for students to excel both academically and professionally. Imran and Almusharraf [4] showed that ChatGPT provides immediate and personalized assistance in writing tasks and delivers instant feedback to students [4]. With ChatGPT, personalization allows feedback on everything from individual writing styles to personal preferences. This may create a deficiency in contextual understanding and raise questions about the accuracy of the responses this tool provides. This has been highlighted in studies by Roumeliotis and Tselikas [5]. Moreover, it needs to be added that accidental plagiarism continues to occur due to the nature of the training on the different datasets. There can always be more over-reliability, which is the factor inhibiting one’s independent writing improvement. According to Ray [6], this tool cannot generate original ideas as it is supposed to do; it can also produce results that may sometimes be inconsistent and may have trouble with complicated commands.

According to Dwivedi [7], ChatGPT can be used as a coding buddy for students in generating code snippets. It also provides an avenue for students to get coding assistance at their own pace, hence enhancing independent learning [8]. Tian’s [9] study raises concerns regarding the limitations of AI in addressing complex programming challenges. Ethical considerations include the potential for academic misconduct through the use of machine-generated code. Attention must be paid to maintaining academic integrity, as indicated by Azeem and Khan [10].

The reason why academic writing and coding were chosen as distinct types of learning experiences in this paper is that they relate to different types of cognitive and practical skills. First, academic writing is essentially a verbal and conceptual operation based on substantial knowledge of language, structure, and reasoning [11]. Writing professionally in an academic context has been supported to require a student to apply such cognitive and metacognitive processes as critical thinking, organization, and reflection. Writing, in turn, implies linguistic precision and the ability to express complex ideas [12]. On the other hand, coding is a logical problem-solving ability to change abstract ideas into functional solutions using programming languages. Learning to code has enhanced problem-solving skills, whereby the learner breaks down tasks into smaller, manageable pieces and applies algorithmic thinking [13]. Coding also involves much technical understanding and is, therefore, iterative in nature, where errors and debugging form a part of learning to do it correctly [14]. While both are quite different in their approach to learning, academic writing and coding can shed much light on how these two modes of learning (i.e., one more language-driven and the other more technical) shape our cognitive development and our approach to problem-solving.

1.1. Theoretical Foundation: ChatGPT Within the Broader Context of Educational Technology

Personalized learning is an educational approach that tailors learning experiences to individual students’ needs, preferences, and interests. Instead of a one-size-fits-all curriculum, personalized learning looks to adaptive instruction, flexible pacing, and customized content to optimize each learner’s path to mastery. According to the theory, learning centered on the learner’s prior knowledge, goals, and learning styles is the most effective [15]. Key components of personalized learning include using technology to collect student learning data to provide real-time adjustment, multiple pathways to the learning content, and cultivating an autonomous and self-directed learner. Research supports that personalized learning can increase student engagement, motivation, and achievement by addressing individual learning gaps and enabling learners to take ownership of their educational journey. The approach often links back to adaptive learning technologies, which employ data-driven algorithms that deliver a tailored learning experience in real-time [16].

However, personalized learning is more contentious than this suggests because the educational community has debated its efficacy and practicability. For example, Pashler et al. [17] felt that there was limited empirical evidence on the effectiveness of personalized learning approaches, particularly related to the differential tailoring of instruction on individual learning styles. Moreover, Coffield et al. [18] analyzed several learning style models and found that many of these approaches lack a scientific basis, which raises concerns regarding their widespread usage in educational institutions.

The learning style theory postulates that individuals learn best when a material is presented in their preferred mode, such as visual, auditory, or kinesthetic. While the theory is well-practiced in education, it has come under considerable attack due to a lack of empirical support. For example, Pashler et al. [17] and Coffield et al. [18] did not find any substantial evidence linking learning styles to improved academic performance. Critics suggest that learning styles grossly simplify cognitive diversity, and that someone only learns through one modality that is not taken into context or is situational [19].

Despite these criticisms, the notion of personalized learning has value. Yet, it suggests that, rather than learning styles being so rigid, a more fluid approach would be far more advantageous. Education should draw upon an evidence-based approach: Hattie’s [20] emphasis on visible learning, where feedback and adaptability lie at the core of success, is a far more valued means forward. This nuanced perspective acknowledges that personalized learning is useful yet should not be bound to the model of learning styles. Ultimately, a more evidence-based and adaptive approach to personalization would allow diverse instructional methods tailored to diverse learner needs without using unsupported theories of fixed learning styles. ChatGPT may support such an adaptive approach to provide personalized learning, as identified in this paper.

ChatGPT relates closely to the principles of personalized learning and is, therefore, an effective tool to improve individualized education. Since it is an AI language model, ChatGPT can provide personalized interactions with a learner based on his input, knowledge level, and preference. The literature supports that AI technologies in education have been shown to support personalized learning through real-time, adaptive feedback that meets the individual needs of learners [21]. Immediate feedback, deeper explanations, and additional resources help ChatGPT support learners in progressing at their own speed, giving personalized guidance tailored to their distinctive learning needs. ChatGPT can also evaluate student responses and guide them through the learning process by asking probing questions, helping learners develop critical thinking and problem-solving skills—skills central to personalized learning environments [22]. For example, if a learner struggles with a particular concept in any subject, ChatGPT can paraphrase the explanations or offer simpler examples and adjust responses according to the learner’s understanding. It also aligns with research illustrating how AI dynamically adapts instructional content to meet the learner’s needs while increasing engagement and deepening comprehension [23]. With such a broad-ranging remit over subjects and topics, ChatGPT’s support could be flexible to meet nearly any learning needs in many ways through visual, auditory, or text-based means. Research has identified that adaptive technologies utilizing artificial intelligence provide tailored support across various areas of content, thus ensuring that learners gain swift access to fit-for-purpose education [24].

According to the socio-constructivist theory [25], knowledge is constructed through interaction with others, and tools like ChatGPT could play a central role in fostering dialogue, peer collaboration, and scaffolding learning experiences. Research shows that ChatGPT conversation engages students in language production and problem-solving, especially in teaching and learning English [26]. While using ChatGPT, students can actively negotiate meaning and co-construct knowledge, demonstrating active engagement [27] and learner autonomy [28] that supports constructivist learning theory.

Similarly, human–technology interaction theory is based on the interaction between users and technology [29]. It investigates how humans interact with technological tools and how such interactions affect user experiences and learning outcomes [30,31]. This theory also considers issues related to user agency, trust, and the potential of AI tools to empower or limit students in their learning journey. Research shows that ChatGPT can mimic human empathy and enhance user experience by providing sophisticated conversational capabilities [32].

1.2. Existing Reviews and Their Limitations

Scholars have conducted several reviews on the use of ChatGPT in education. For example, Lo [33] researched approximately 50 articles published from December 2022 to February 2023 and summarized how students and teachers benefit from ChatGPT. Mai et al. [34] conducted a strengths, weaknesses, opportunities, and threats (SWOT) analysis of 51 articles to understand ChatGPT’s contribution to teaching and learning. Their review suggests that it supports learning effectiveness despite the inaccuracy of information and the generation of limited answers. In a SWOT analysis of 12 studies conducted by Tur [35], the authors determined the utilization of ChatGPT in K-12 education. Tutors can use the ChatGPT to help students learn better, but there are concerns about the reliability of such sources [36]. Twelve articles were examined by Montenegro-Rueda et al. [37] for using ChatGPT in education. Ultimately, it was found that while ChatGPT can improve learning in the classroom, teachers must first undergo related training before they can assist learners.

Nevertheless, the reviews provided show how using ChatGPT may have advantages and disadvantages despite the limitations of these studies. For example, the study of ChatGPT is rapidly evolving, and most existing reviews have focused on a small number of studies published between 2022 and 2023 [34,38]. Some of the included studies were not peer-reviewed; for instance, Lo [33]. Montenegro-Rueda et al. included only 12 studies between May and June 2023. To the best of our knowledge, no published review has investigated the use of ChatGPT for personalized learning, academic writing, and coding tasks. Although Zhang and Tur [35] reported in their review that ChatGPT could be utilized for personalized learning, their review did not elaborate on how it has been utilized. Another study by Baig and Yadegaridehkordi [39] emphasized that ChatGPT offers a cost-effective alternative to traditional resources. However, it is still in the early stages and requires further investigation to understand this tool fully in the context of higher education. The ChatGPT’s support for writing is especially critical for coding tasks in higher education. It is now advisable to create a comprehensive summary of the available studies regarding ChatGPT’s capacity to improve individual learning, scholarly writing, and programming in higher education institutions. As ChatGPT came into being in November 2022, no review has directly targeted the above-mentioned areas; hence, this study aims to address four main research questions.

RQ1: How can ChatGPT be used for personalized learning?

RQ2: How does ChatGPT effectively enhance academic writing skills?

RQ3: How can ChatGPT be used for improved coding tasks?

RQ4: What are the challenges in the integration of ChatGPT in higher education?

2. Methods

The literature review was conducted considering the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [40].

2.1. Search Strategy and Inclusion Criteria

We searched six electronic databases to ensure that sources were comprehensive and diverse for our study. These were selected based on their relevance to the research topic, academic credibility, and coverage of the specific fields of interest. These criteria thus led us to choose databases that, in their coverage, include a wide variety of peer-reviewed journals, conference papers, and academic articles. The targeted databases were Scopus, for extensive multidisciplinary coverage and strong indexing of peer-reviewed literature; ACM Digital Library, a key resource for research related to computer science and technology; Education Research Complete, covering a wide range of educational research articles; Computers and Applied Sciences, applied computer science and technology; Web of Science, multidisciplinary in scope with high-impact journals; and IEEE Xplore, specialized in engineering, computer science, and technology research. These databases were selected to ensure both depth and breadth in the literature search, aligning with the objectives of this review.

Published articles on ChatGPT integration in higher education, academic writing, personalized learning, programming, and coding were included. The publication period was from January 2023 to February 2024. The exclusion criteria included non-English articles and articles from non-academic sources.

Table 1. Inclusion and exclusion criteria for article selection.

Criterion	Inclusion	Exclusion
Article topic	Addresses the integration of ChatGPT in higher education contexts, focusing on its role as an educational resource/tool	Does not address the integration of ChatGPT in higher education contexts, focusing on its role as an educational resource/tool
Article type	Academic articles	Non-academic articles, such as articles from mass and social media
Time	1 January 2023 to 1 February 2024	Articles outside the specified time
Language	English	Non-English

summarizes the inclusion and exclusion criteria in selecting the articles for review. In the final search, 551 papers were identified to enable a comprehensive literature review on ChatGPT integration in higher education.

2.2. Screening

After the collection of relevant papers, Zotero 6.0.30 was utilized for effective management. The first step involved eliminating duplicate records, totaling 148 in number. Next, papers published in non-English languages, those not aligned with the scope of the review, and those published in non-academic sources or conference papers were excluded, totaling 120 papers. Only the journal articles were retained during this process. The title and abstract screening process was subsequently performed on the remaining records (n = 283). In this phase, the title and abstract of each paper were reviewed independently by two authors to establish their relevance to the scope of the current review. Exclusion criteria involved papers that did not meet the objectives of this review or failed to fulfill set eligibility criteria. This intensive process selected only those studies directly relevant to the research question and excluded records that focused on technical aspects rather than educational implications, explored broader AI ethics without specific educational applications, or had limited relevance to personalized learning approaches (n = 113). The next step was a full-text eligibility screening. During this phase, non-full-text articles (n = 27) were excluded. A full-text reading of the remaining papers led to the exclusion of literature reviews and additional records outside the scope of the current review (n = 117). This further refinement resulted in a final set of papers (n = 26) (See Figure 1).

2.3. Data Extraction and Analysis

An extraction sheet was developed to include the key components of each study. See Supplementary Materials for the extracted data. These components included the study’s objective, concept of ChatGPT, discipline, context, methods, findings, conclusions, limitations, and recommendations. Each record was scrutinized for its journal source, research aims, and support for personalized learning, academic writing, coding, and ethical considerations. Each record was then reviewed by two independent authors on journal quality and source relevance. The reputation of the journal, based on its past reputation and linkage to known academic or professional bodies, was considered. For this reason, only credible sources directly relevant to the research found their way into the review. Thematic analysis, as outlined by Braun and Clarke [41], was considered to interpret the gathered data. This six-step process involves familiarization, coding, generating themes, reviewing themes, defining and naming themes, and documentation. Initially, we examined the extracted data to become familiar with its content before assigning codes for categorization. From each study, we extracted information (e.g., personalized learning, academic writing, coding, ethical issues, and others) based on our domain of interest, according to

Table 2. Our domains of interest with their categories.

Domain	Categories/Themes
Personalized Learning	Real-time feedback, adaptive learning, collaborative learning, student engagement, and self-regulatory learning.
Academic Writing	Academic writing process, content generation, writing tools, over-reliance, and critical thinking.
Coding	Code generation, code debugging, and code explanation.
Ethical issues	Plagiarism, biases, and academic misconduct.
Others	Privacy, accuracy, and recommendations

. These categories were developed based on our aim and research questions of the study. We then analyzed studies reporting similar issues and grouped them into a cohesive theme. We re-read the themes to be certain they reflected an accurate representation of the data. With the final list, we delineated each theme and gave it a name that was related to our research questions. See Appendix A (

Table A1. Summary of the included studies.

Author & Year	Aim	Country	Samples	Methods	Data Collection	Findings	Limitations
Abbas et al., 2023 [53]	To discuss the opportunities and challenges of using ChatGPT for personalized learning in higher education.	Malaysia	Not reported	A comprehensive literature review	Not reported	Educators, administrators, and policymakers should know the possible benefits and risks of using ChatGPT in personalized learning. An ethical framework and guidelines regarding its implementation and use in higher education should be developed.	This paper lacks clarity regarding the comprehensive literature review methodologies employed, as no explicit details are provided. While it recommends the development of ethical guidelines for integrating ChatGPT in higher education, it falls short of offering a thorough discussion to support these recommendations.
Albdrani and AI-Shargabi, 2023 [54]	To examine ChatGPT’s potential as a research tool for data science educators to investigate the effectiveness of AI in personalized learning experiences.	Saudi Arabia	20 students	A case study design with a mixed methods approach: quantitativedata from quiz results and qualitative observations	A control group and an experimental group	Its integration into the instruction of the experimental group, embedded in a 5E instructional model, proved reasonably successful concerning the students’ level of involvement with it and with well-ordered learning. Moving through its successive phases clearly promoted comprehensive learning while ChatGPT played the important role of the virtual co-teacher in encouraging agency and greater student engagement.	The sample size for the study was relatively small, comprising only 10 students in the control group and 10 students in the experimental group. Additionally, the study primarily focused on the short-term effects of using ChatGPT on student learning outcomes.
Basic et al., 2023 [63]	To examine students’ essay-writing performances with or without ChatGPT as an essay-writing assistance tool.	Croatia	18 s-year master’s students	Quantitative data using Excel and R studio	A control group and an experimental group	Results showed that on none of the two indicators did the ChatGPT group do any better; the students produced content of no higher quality, did not write faster, and did not have a higher degree of authentic text.	The primary limitation of this study was its small sample size, with only nine students per group. Moreover, the analysis was restricted to quantitative data, focusing solely on final essay scores evaluated based on mechanics, style, content, and format. Incorporating qualitative data, such as student feedback and responses to targeted questions, could provide a more comprehensive understanding of the outcomes.
Biswas, 2023 [65]	To outline the role and capabilities of ChatGPT, a language model developed by OpenAI for computer programming	USA	Not reported	Not reported	Not reported	ChatGPT could explain complex notions and technologies, provide examples, guide resources, and identify and solve different technical issues. Its use can further enhance overall satisfaction with the support services of an organization so that it’s considered a reputable and dependable company.	This paper lacks a transparent methodology for examining the role and capabilities of ChatGPT in computer programming, leaving significant gaps in the understanding of its application and effectiveness.
Buscemi, 2023 [67]	To investigate the coding proficiency of ChatGPT 3.5 to identify potential areas for development and examine the ramifications of automated code generation on the evolution of programming languages and the tech industry.	Luxembourg	Not reported.	Testing, 10 programming languages were chosen, and 40 tasks were employed to test.	A total 4 of 000 tests	ChatGPT 3.5 already demonstrates the ability to generate code that solves many tasks. The model is non-deterministic, meaning it may develop different code solutions to the same problem. This usually results in inconsistent performance: given a specific task, the model generates syntactically correct code in some instances, but in other cases, it produces either bugged code or no code at all. Also, the programming language choice influences comprehension of the task’s requirements.	This paper assessed various factors, including time performance, code length, ethical considerations, and others. However, it overlooked a critical aspect: evaluating the quality of code produced by ChatGPT.
Coello et al., 2024 [66]	To explore the effectiveness and efficiency of the popular OpenAI model ChatGPT, powered by GPT-3.5 and GPT-4, in programming tasks to understand its impact on programming and, potentially, software development.	Germany	Not reported	Testing, A quantitative approach	Four hundred sixty programming prompts were used from the Basic Python Programming dataset (1000 crowd-sourced programming problems.	It was observed that LLMs developed some difficulties in code generation that led to bugs and errors, and multiple solutions when a slight change or wrong query was provided. This may lead to inefficient time use by programmers. While ChatGPT and other LLMs can generate code effectively and thus can be used as programming assistant tools, they are not meant to replace human software developers, as they always need human feedback and monitoring.	This study did not discuss ethical issues related to using ChatGPT in coding. Also, they did not test the prompts with students and did not focus on evaluating the generating codes.
Ellis and Slade, 2023 [49]	To examine the potential of ChatGPT as an educational tool for statistics and data science.	USA	Not recorded	A literature review	Three prompts, statistical data (p values)	Educators can guide the use of generative AI tools in statistics and data science classrooms so that students and educators can leverage the benefits of this technology.	They briefly discussed the challenges of using AI tools in the classroom. More discussion would be useful in terms of limitations and legal and ethical concerns of using AI.
Esmaeil et al., 2023 [46]	To understand perception regarding the use of ChatGPT in their argumentative writing.	Malaysia	17 students	A qualitative research approach	Document analysis is used to collect data and analyze insights into complex topics related to student writing.	Although the students acknowledge the extensive capabilities of ChatGPT, including its ability to provide information and guidance and decrease both research expenses and time consumption, they also voice apprehensions.	The discussion section of this paper is brief, presenting the findings with some support from existing literature. Expanding this section to provide a more in-depth analysis and including practical recommendations based on their findings would significantly enhance its value.
Fernando et al., 2023 [58]	To ascertain what factors influence the length of time undergraduates receive individual tutoring	Poland	287 students	Quantitative questionnaires	The questionnaires were used to collect data.	ChatGPT can increase student engagement by utilizing multimedia and interactive teaching aids.	The participants were not selected randomly for conducting this study; hence, the results may be biased.
Fuchs, 2023 [45]	To discuss a range of challenges and opportunities for higher education, as well as conclude with implications that (hopefully) expose gaps in the literature, stimulate research ideas, and, finally, advance the discussion about NLP in higher education.	Thailand	Not reported	A comprehensive review	Not reported	Some of the possible benefits of using NLP models for personalized learning and on-demand support include the creation of customized learning plans, generating feedback and support, and providing resources to students at any time and from any location. However, there are also some challenges that NLP models may bring, including the loss of human interaction, bias, and ethical implications.	This study could benefit from a clearer explanation of the implications of its findings. Additionally, it should directly and precisely address the question posed in its title: “Is ChatGPT a blessing or a curse?”
Guleria et al., 2023 [71]	To highlight the use of AI and AI-assisted technologies such as the ChatGPT and other chatbots in scientific writing and research, which results in bias, the spread of inaccurate information, and plagiarism.	India	Not reported	Experiments were conducted to test the authenticity and accuracy of ChatGPT.	Not reported	The information provided by ChatGPT was not appropriate; it can also bring about implications for medical science and engineering. Here, critical thinking should be encouraged to show awareness of related privacy and ethical risks.	The research conducted experiments and shared insights on the use of AI; however, it lacks clarity regarding the students’ perspectives.
Halaweh, 2023 [64]	To present an argument in favor of incorporating ChatGPT into education, educators should be provided with a set of strategies and techniques to ensure the responsible and successful implementation of ChatGPT in teaching or research.	United Arab Emirates	Not reported	Literature review	Not reported	Teachers should permit the use of ChatGPT and even be the ones to start using it since students will use it anyway. Permitting them to use the tool puts them on an equal footing in developing ideas and improving their writing, as the faculty encourages.	This paper did not consider many peer-reviewed research papers on the use of ChatGPT in education. The only consideration is Google Scholar as their database for collecting papers.
Kiryakova and Angelova, 2023 [55]	To explore the opinion of university professors at a Bulgarian university regarding the possibilities and challenges of ChatGPT in carrying out teaching activities.	Bulgaria	87 University professors	Survey	A questionnaire was distributed by email to the participants.	ChatGPT is a means to support time-consuming teaching activities, provoke interest, activate and engage students, and stimulate their critical thinking and creativity.	A study focusing on learners’ attitudes toward the use of ChatGPT in education would provide valuable insights into the benefits and challenges of integrating ChatGPT into classrooms.
Mehmet Firat, 2023 [59]	To identify the implications of ChatGPT, an AI-powered language model, for students and universities by examining the perceptions of scholars and students.	Turkey	14 PhD students, and seven scholars	A thematic content analysis approach	Open question	Integrating AI in education offers many opportunities to enhance learning experiences, personalize instruction, and transform educators. However, this shift challenges assessment, digital literacy, and ethical considerations.	The study employed an open-ended question to identify the main themes and their frequencies. It would be beneficial to incorporate additional open-ended questions to gather more comprehensive data.
Qureshi, 2023 [56]	To explore the prospects and obstacles associated with using ChatGPT as a tool for learning and assessment in undergraduate Computer Science curriculum, particularly in teaching and learning fundamental programming courses.	Saudi Arabia	24 students	Quasi-experimental research	The control and experimental groups	Students using ChatGPT had an advantage in earned scores, but inconsistencies in the submitted code affected the overall performance.	Providing detailed recommendations for both educators and students regarding the use of ChatGPT would be highly valuable.
Rasul et al., 2023 [47]	To examine the potential benefits and challenges of using ChatGPT in higher education, in the backdrop of the constructivist theory of learning	Australia	Not reported	Perspective type study	Not reported	Tertiary educators and students must exercise caution when using ChatGPT for academic purposes to ensure its ethical, reliable, and effective use.	The discussion could be more thorough by exploring the benefits of incorporating constructivist learning principles alongside the use of ChatGPT in the classroom.
Richards et al., 2024 [61]	To provide a baseline understanding of how the public release of generative AI is likely to impact quality assurance processes significantly.	UK	Not reported	A dual-anonymous study protocol	A mix of descriptive statistics and graphing of the data	In most cases, across a range of question formats, topics, and study levels, ChatGPT is at least capable of producing adequate answers for undergraduate assessment.	The study design lacks sufficient detail, as it does not clearly outline the sample size or the procedure followed in conducting the study.
Rudolph et al., 2023 [81]	To present the technology’s implications for higher education and discuss the future of learning, teaching, and assessment in higher education in the context of AI chatbots such as ChatGPT.	Singapore	Not reported	A desktop analysis approach	Not reported	ChatGPT can be used to check sentences for plagiarism input by the user and then modify them so that anti-plagiarism software reports a low originality index score.	This paper only considered two peer-reviewed and eight academic papers on ChatGPT and higher education.
Rudolph et al., 2023 [52]	To systematically compare selected chatbots across a multi-disciplinary test relevant to higher education.	Singapore	Not reported	A Systematic comparison within the Chatbots.	15 Test questions	There are currently no A-students and no B-students in this bot cohort, despite all publicized and sensationalist claims to the contrary.	The study could discuss the implications of their research.
Silva et al., 2024 [70]	To gauge the viability of ChatGPT in programming education and sustainability.	Brazil	40 students	Three distinct stages: Initial tests on ChatGPT on chatbot, learning and teaching code on ChatGPT, and student experience using ChatGPT in coding	A Questionnaire	Of particular note, the majority of students involved in this research showed interest in the usage of the tool as a supportive tool for teaching in the classroom to develop sustainable and improved learning. The integration of ChatGPT into coding and programming courses changes students’ perceptions regarding educational support, sustainability, and individual learning experiences.	Half of the students struggled to effectively utilize ChatGPT’s resources, highlighting the need for a deeper understanding of these challenges in order to enhance the AI tool’s support for their learning.
Singh et al., 2023 [50]	To compare selected chatbots across a multi-disciplinary test relevant to higher education.	UK	430 students	A Survey	A Questionnaire	ChatGPT can be helpful in learning/teaching activities, but better guidelines should be provided for the students in using the tool.	This study did not compare student performance based on their level of proficiency in using ChatGPT.
Sridhara et al, 2023 [68]	To explore how ChatGPT can be used to help with common software engineering tasks.	India	Not reported	Testing and observation	15 common software engineering tasks	While ChatGPT indeed does a credible job for many tasks, its response is detailed, often beyond what even a human expert would have come up with or even state of the art. However, it also turns out that ChatGPT provides incorrect answers for a few other tasks and, hence, is not suited for such tasks.	This study could offer more insight into the academic integrity concerns and risks associated with using ChatGPT for coding tasks. Additionally, conducting a test with students to gather their perspectives on using this tool would provide valuable information.
Stojanow, 2023 [82]	To examine ChatGPT’s use as a tool aiding the learning process has not been examined.	New Zealand	Not reported	An autoethnographic study examining	Not reported	ChatGPT gave users sufficient content to form an overall impression of its technical features, and users felt the response it provided to be engaging and relevant. Answers were, however, somewhat superficial; generated text was not always logical, even contradictory at times.	Focusing on a single instance, particularly the feature of autoethnographies, is a limitation of this study. Learning approaches, experiences, and how students interact with technology vary among individuals, which means the insights gained by the author may not be applicable to others.
Sullivan et al., 2023 [48]	To examine news articles (N = 100) about how ChatGPT is disrupting higher education, concentrating specifically on Australia, New Zealand, the United States, and the United Kingdom.	Australia	Not reported	Content analysis/a systematic search	Not reported	There was mixed public discussion and university responses, with a focus mainly on academic integrity concerns and opportunities for innovative assessment design.There has also been a lack of public discussion about the potentialfor ChatGPT to enhance participation and success for students from disadvantaged backgrounds.	They analyzed coverage in mainstream news databases but did not explore alternative news sources.
Yilmaz and Karaoglan Yilmaz, 2023 [43]	To analyze the students’ perspectives on using ChatGPT in the field of programming and programming learning	Turkey	41 students	The case study method	A questionnaire and a form consisting of open-ended questions	It would be useful to integrate generative AI tools into programming courses, considering the advantages they provide in programming teaching.	This research was limited to an 8-week period. Future studies could benefit from a longitudinal approach to examine the perspectives of students who use ChatGPT over a longer duration.
Zou and Huang, 2023 [44]	doctoral students’ acceptance toward ChatGPT in writing and the factors that influence it.	China	242 doctoral students	Online survey	descriptive analysis and correlation analysis	There was powerful evidence for the applicability of the Technology Acceptance Model in the acceptance of ChatGPT in writing	This study was exploratory and relied solely on survey questions to gauge students’ acceptance of ChatGPT. It did not employ a case study or mixed-methods research design, nor did it collect multiple sources of data to gain a more nuanced and in-depth understanding of students’ actual processes and outcomes when using ChatGPT for writing.

) for the summary of included studies.

3. Results

3.1. General Description of the Included Studies

The bar chart (Figure 2) illustrates the included studies’ coverage of different higher education disciplines in which the ChatGPT was used. Among these disciplines, general higher education is the most prominent, and Computer Science and Education are the most frequent, considering a specific study area. Six papers closely trail the domains of the Humanities and Social Sciences, and the fields of business and social sciences, medicine, and interdisciplinary science are represented by one paper each.

The pie chart (Figure 3) illustrates the distribution of the studied papers across various countries and continents. In terms of continents, Asia appears to have the highest frequency of papers (45.2%), followed by Europe (31.0%) and Oceania (11.9%).

Research methods such as quantitative (e.g., surveys) (n = 11), qualitative (e.g., observations and interviews) (n = 12) studies, and mixed methods (n = 3) were used in the included studies. The key aim of most of the studies (n = 11) was to understand ChatGPT’s applications across various fields. On the other hand, a significant chunk of studies (n = 5) also focused on the effectiveness of ChatGPT, majorly in essay writing and programming. The interest in researching the issues and limitations in using ChatGPT, above all in view of academic integrity and biases, was high (n = 4). Comparison with other tools or methods, the effect of ChatGPT on teaching and learning and student perception was discussed, but lower in frequency (n = 3).

3.2. RQ1: How Can ChatGPT Be Utilized for Personalized Learning?

The included studies pointed out that ChatGPT can provide feedback at times of need, facilitate learning changes over the course of education, and create conditions for students to work together. In addition, ChatGPT has been shown to increase learner engagement through interactive platforms, which also supports self-regulation. The tool encourages students to monitor their progress and make adjustments.

Real-time feedback refers to responses, either for action, question, or performance, provided for immediate and continued action. That would imply time for adjustment, improvement, constant learning, and development of skills [42]. Of the 26 studies analyzed, ten focused on exploring ChatGPT as a personal tutor and, more specifically, on the function of its real-time feedback. Several studies (n = 7) have emphasized ChatGPT’s role in enhancing personalized learning. Wang et al. (2021) noted not only real-time feedback but also goal setting and progress monitoring. Several other researchers have noted similar benefits, such as quick and effective responses [43], effective personalized support across learning tasks [44], on-demand support [45], and real-time writing guidance [46]. Rasul et al. [47] state that ChatGPT’s just-in-time feedback provision fosters understanding and access to resources. Sullivan et al. [48] highlighted its value in providing plain language explanations, organizational suggestions, grammatical feedback, and quiz-question development.

However, the limitations of real-time feedback systems should be understood and acknowledged. Seven of the ten articles studying real-time learning acknowledged that ChatGPT sometimes gave incorrect answers. Ellis and Slade [49] in their study observed its inability to adjust instructions that were not in text format, and occasional delays in its responses. Furthermore, the efficacy of ChatGPT is contingent on the complexity of the problem at hand [49,50]. It points to a more profound problem in educational technology: how ChatGPT responds poorly to complexity, especially within diverse and changing learning contexts. Such an educational tool, like ChatGPT, would fail to respond appropriately to the complexities of the needs of learners, particularly those requiring more nuanced approaches. While AI tools provide useful features, they also have limitations that may affect learning outcomes. Pedagogical seriousness designates the importance of accurately understanding and conveying knowledge in an educational setting. The mistakes of AI or the inability to handle complex or nuanced information undermine such learning, mainly when students rely on AI’s output without fully mastering concepts. In other words, overlooking these weaknesses in AI applications risks compromising the pedagogical value of these tools, which may go so far as to accidentally mislead students or fail to engage them in critical thinking about the content. Therefore, although GPT tools assist greatly, limitations in their functionalities should not be avoided, particularly in learning situations that are quite complex or subtle.

Adaptive learning is an educational approach in which instructional materials, resources, and activities are tailored to the needs of each learner [51]. However, as seen with ChatGPT, the inability to adjust to complex learning needs—such as non-text formats or varied cognitive styles—can present significant limitations. Out of the 26 reviewed papers, 15 emphasized adaptive learning. However, again, in that context, integration with ChatGPT depends on either the complexity of the task or the learner’s familiarity with that tool. In various educational systems, such biases and limitations may be further exacerbated, for example, by difficulties faced by students who are hands-on or experiential learners, leading to inequalities whereby some learners may not be able to engage effectively with ChatGPT [52]. Therefore, in an educational context where complexity or learner diversity is not addressed, AI tools like ChatGPT may inadvertently reinforce existing biases, making it harder for all students to benefit equally.

The NLP in ChatGPT encourages student independence and makes their learning experience active and engaging [53]. When used as a co-teacher virtual assistant, ChatGPT provides individualized support to make learning more engaging [54]. Kiryakova and Angelova [55] and Qureshi [56] found it to be an intelligent tutoring system with potential benefits for creating customized instruction. Natural language processing models, including ChatGPT, can be used to construct highly individualized learning plans based on performance and feedback [45]. The integration of ChatGPT appears to enhance educational support and sustainably create positive learning experiences [50]. ChatGPT can be beneficial to students in such a way that proficient users become better at using this tool effectively through critical engagement with materials [57]. In addition, once learners grasp how to properly use ChatGPT, it becomes more useful and effective in augmenting their learning experiences [57].

Applying the learning phases based on the 5Es model (Engagement, Exploration, Explanation, Elaboration, and Evaluation), researchers showed that ChatGPT provides personalized assistance in each phase, which increases learner commitment. Through instant access to data, scholars’ enthusiasm and self-belief are boosted in ways that make them more committed to their work [46]. Moreover, the use of the tool is believed to create more enjoyment and enthusiasm, as shown by the positive perceptions of ChatGPT and personalized coaching [58]. According to Fuchs [45], the support provided by ChatGPT upon request contributes a great deal to the learning among students, especially those who study online. This enables teachers to concentrate on upper-level abilities as well as mentorship aided by AI systems, such as ChatGPT. The shift from educators’ roles can result in increased satisfaction among learners [59]. In a way that enhances learning experiences for students who study together in interactive learning environments, ChatGPT fosters teamwork and the verification of solutions during problem-solving processes [50]. However, there are concerns that learners may rely solely on interactions with the ChatGPT. In this scenario, collaborative learning and critical thinking can be hindered [47]. Therefore, ChatGPT’s promotion of robust student engagement may vary depending on the instructional context and implementation strategies [43].

Self-regulatory learning is a process whereby students set goals, monitor progress, and adjust strategies to manage their learning [60]. Of the 26 papers analyzed, 11 discussed self-regulatory learning. ChatGPT plays an important role in self-regulated learning by responding to different learning styles and preferences [53]. Personalized learning materials enhance ownership and agency for learners in the learning process [54]. ChatGPT’s judgment-free environment serves as a private space for learners to seek clarification, which alleviates classroom anxiety [46]. Moreover, personalized tutoring by ChatGPT brought more satisfaction, engagement, and self-confidence. This underlines the importance of subjective factors for effective learning interventions [58]. ChatGPT provides respectful and motivating comments and supports engaging with the content better [57]. This type of interaction is particularly beneficial for students with prior knowledge, who are developing critical thinking skills and can enhance their learning efficiency [57].

Despite Esmaeil et al.’s [46] finding that ChatGPT can reduce classroom anxiety by providing a judgment-free zone to clarify doubts, Stojanow [57] found that ChatGPT may not fully support students in achieving their learning goals independently. Furthermore, although Yilmaz and Yilmaz [43] indicated that incorporating ChatGPT in programming tasks may enhance self-confidence and motivation, Zou and Huang [44] found that it may not adequately promote self-regulated learning strategies. See

Table 3. Summary of ChatGPT’s support of real-time feedback, adaptive learning, and self-regulatory learning based on 15 studies.

Aspect of Learning	Summary of Support from Literature	Limitations and Considerations
Real-time Feedback	Provides immediate guidance, corrections, and explanations, enhancing continuous learning [42].	Occasionally generates misinformation and incorrect answers (Various sources).
	Offers personalized feedback, goal-setting assistance, and progress monitoring [2].	Unable to accommodate non-text prompts; occasionally has slow response times [49].
	Facilitates on-demand support and guidance in various tasks [45,46].	Efficacy is dependent on task complexity [50].
Adaptive Learning	Tailors learning paths, providing personalized guidance and support (Various sources).	Efficacy in offering customized instructions is uncertain [56].
	Acts as a virtual co-teacher, offering customized instructions and enhancing interactivity [55].	It is not suitable for students who prefer hands-on, experiential learning [52].
	Constructs individualized curricula and generate customized learning plans [58].
Student Engagement	Empowers learners, fosters ownership, and increases motivation through personalized learning [53].	Potential over-reliance on ChatGPT could hinder collaborative learning and critical thinking [47].
	Sequential progression guided by the 5Es model enhances engagement [54].	Varies in effectiveness depending on instructional context and implementation strategies [43].
	Immediate access to information boosts motivation and confidence [46].
Self-regulatory Learning	Facilitates diverse learning styles and preferences [53].	May not fully support students in independently setting and achieving learning goals [57].
	Provides a judgment-free zone for clarifying doubts and reducing anxieties [46].	Incorporation in programming tasks may not adequately promote self-regulated learning strategies [44].
	Promotes active learning and critical engagement through respectful and encouraging feedback [57].

for the summary of ChatGPT’s support of real-time feedback, adaptive learning, and self-regulatory learning.

These studies have a number of recommendations that can ensure the effective usage of ChatGPT for personalized learning: training educators and providing them with clear guidelines on how to integrate ChatGPT into their practice is necessary [45,53]. Such training may encourage responsible and transparent use of the tool. Besides, digital competencies at both the student and faculty levels need to be developed for the full exploitation of AI tools in educational settings [55]. Knowledge empowerment of stakeholders in the use of such a tool as ChatGPT is the key to building an enabling environment for personalized learning. Future research should explore collaborative learning environments that leverage ChatGPT’s capabilities to collectively solve complex problems collectively [53]. Another important avenue of research is the investigation of the long-term impact of ChatGPT on learning outcomes beyond the classroom [53].

3.3. RQ2: How Can ChatGPT Effectively Enhance Academic Writing Skills?

Nine studies discussed ChatGPT’s usage across a full academic writing process and described effective content produced within instances of this AI performing a standard tool’s function. In contrast, however, they express several concerns since overuse will decrease motivation toward developing skills on their own.

All nine studies that explored the role of ChatGPT in writing addressed its ability to generate content. ChatGPT provides invaluable assistance in idea generation and summarization. It offers students a platform for brainstorming and quickly generating texts that would otherwise require substantial time and effort [46]. Zou and Huang [44] further suggested that, in addition to brainstorming, ChatGPT enhances content creation through the provision of individual feedback to students in developing their written drafts. The effectiveness of ChatGPT in content generation also varies across different contexts. For instance, it has been observed that it generates adequate scripts for modules at the undergraduate level, while differences in quality vary with formats and topics of questions [61]. Additionally, as McMurtrie [62] suggests, ChatGPT’s integration into everyday writing seems inevitable, akin to incorporating calculators and computers into mathematics and science. While this use of ChatGPT for writing is undoubtedly an increasingly common and continuing phenomenon, it must be placed within the perspective that ChatGPT is not the only AI model available. Various companies are integrating other GPT models and many other language models into their products. For instance, Microsoft has already integrated GPT models into its products, such as Microsoft Word and Excel, to show that this broader trend of integrating AI into everyday tasks goes beyond just ChatGPT. ChatGPT is only a small part of a bigger trend in which many AI technologies, including scores of variants of GPT, are finding their way into all sorts of products.

Contrary to expectations, there is no evidence that this would speed up writing. The essays written with ChatGPT are not faster than conventionally written essays [63]. Apart from text stimuli, responses by ChatGPT cannot be initiated with pictures, for instance [49]. It has still been an open question whether such generated text material is credible at all. This is a point of great importance that the student will learn to check and give sources in general so that the content is valid [55].

Of the 26 papers analyzed, 5 discussed the capability of ChatGPT as a writing tool. The advantages pointed out include how ChatGPT saves time in aggregating and summarizing information and in paraphrasing [64]. Furthermore, ChatGPT is a good reference for grammatical feedback; it can serve as a quasi-translator for non-native English-speaking people, especially in very complex terms that may present problems [48]. Meanwhile, several aspects of ChatGPT were found useful across the writing process: brainstorming, personalized feedback, translation, and draft enhancements [44]. However, ChatGPT struggles to consistently generate high-quality content, especially for non-native English speakers [48].

Based on the recommendations, this study suggests various ways in which ChatGPT can be used to enhance writing skills. First, users need to understand how their prompts can affect the quality of the responses they get from ChatGPT [63]. The nature and depth of students’ prompts present a good starting point for understanding how prompts can be optimized [63]. For example, instructors can look at students’ prompts to get an idea of the type of questions they are asking and at what level of specificity. Instructors can then adjust their teaching to better position students to avoid the knowledge gaps and get higher value responses with ChatGPT. ChatGPT is recommended for collaborative learning because it provides a context for developing writing skills in teamwork and problem-solving work [53]. For example, students can work together to generate and refine written content while leveraging the capabilities of ChatGPT, thus enhancing their collaboration skills. For the proper integration of this tool into academic writing, instructors also need training in how to do so effectively with ChatGPT, and educators should receive training to integrate this tool into their teaching [45,53]. Addressing the ethical considerations associated with ChatGPT use is also essential [45,53]. This involves educating students about the limitations and potential biases [49] and emphasizing critical thinking alongside ChatGPT usage [50]. See

Table 4. A summary of ChatGPT’s support of content generation process, grammar and clarity, idea generation, and writing tool based on the literature.

	Academic Writing Skills
	Summary of Support from Literature	Limitations and Considerations
Content generation process	ChatGPT supports students in their writing process by providing personalized feedback and improving drafts [44]	Although ChatGPT proves useful in content generation, it does not necessarily enhance or shorten writing time. Its impact on writing efficiency varies depending on individual writing styles and preferences [63].
	ChatGPT provides students with a platform for brainstorming and quickly generating written content [46].	The credibility of ChatGPT-generated text remains an open question [55].
Grammar and Clarity	ChatGPT effectively identifies and rectifies grammatical errors, improving clarity, coherence, and overall quality of written content [48].	ChatGPT lacks the expertise to provide detailed guidance on subject-specific terminology, non-text prompts, non-English content, citation styles, or content relevance, particularly in specialized fields requiring precise terminology and logical reasoning [48,55].
Idea Generation	ChatGPT proves an invaluable platform for exploring different perspectives and enriching their writing [46,64].	ChatGPT may propagate biases present in its training data and provide not up to date information due to its reliance on pre-existing text data [53].
Writing tool	ChatGPT has proven to be useful in assisting various aspects of the writing process, from brainstorming ideas, and personalized feedback to translating language items, paraphrasing and language improvement [48].	There is a risk of students becoming overly dependent on ChatGPT, potentially hindering the development of critical thinking, creativity, and independent writing skills [63].Additionally, the risk of plagiarism and academic dishonesty is increased if students use ChatGPT-generated content without proper citation [53].
	ChatGPT streamlines the process of aggregating, summarizing, and paraphrasing information, saving valuable time [64].	ChatGPT struggles to consistently generate high-quality content, especially for non-native English speakers [64].

for the summary of ChatGPT’s support for writing-related activities.

3.4. RQ3: How Can ChatGPT Be Used for Coding Tasks?

Seven of the included studies highlighted ChatGPT’s ability to aid students with code generation and explanation, particularly at the introductory level. However, the accuracy of ChatGPT in this regard remains contingent upon further advancements. Its efficacy for code debugging is limited. Seven of the analyzed papers discussed ChatGPT’s capabilities in generating code. The utility of ChatGPT extends across a spectrum of programming tasks, from code completion and correction to document generation and chatbot development [65]. ChatGPT shows potential as a comprehensive tool for assisting programmers in software development. Coello et al. (2024) delved deeper into the dynamics of code generation by comparing GPT-based models with non-GPT-based approaches [66]. Their findings highlight the tendency of GPT-based models to produce shorter, more concise codes [66]. Despite this advantage, challenges such as compatibility issues and the need for human oversight persist. AI models should play a complementary role along with human developers [66].

The effectiveness of ChatGPT in generating executable code varies among programming languages. While it excels in languages like Julia, in other languages such as C++, it performs considerably worse [67]. This leads to inconsistent results and raises ethical issues based on the choice of language [67]. Moreover, if students abuse the tool and use it without acquiring the concepts themselves, academic integrity becomes a problem [56,59]. Sridhara et al. [68] highlighted ChatGPT’s proficiency in natural language processing tasks related to data analysis but noted limitations in feature engineering and accuracy.

Of the 26 papers analyzed, 5 discussed the explaining code capability of ChatGPT. These studies showed the adaptability of ChatGPT in generating documentation for programming tasks [65]. It can automatically generate explanations available for functions, classes, variables, and API references, thus helping in their comprehension [65]. A clear explanation of code and programming concepts with the help of ChatGPT reinforces its role as a learning aid for novice students [43]. ChatGPT works well for method name suggestion, log summarization, and commit message generation [68]. It can generate code snippets adeptly and provide clear explanations, which will be particularly helpful for novice learners working to grasp the basics of programming [66,67].

However, there were instances where the model did not follow some instructions. For example, it would give comments in the code generated even when instructed not to comment [67]. Other problems arise when trying to convert the model-generated explanations into codes that can be executed [56]. Coello et al. [66] stressed that human feedback is necessary for fine-tuning the performance of the model. While ChatGPT can indeed make a much-needed boost for better coding practices, it works best with, not instead, human developers themselves [66].

Out of the 26 papers, 3 were on the capability of ChatGPT to debug code. The ChatGPT chatbot indeed helps in the real-time detection of syntax errors and provides ways of correcting some common programming errors. Thus, it can improve the accuracy and quality of code [65]. However, the suggested solutions sometimes do not compile successfully in integrated development environments. This frustrates users [56]. Moreover, when the capabilities of ChatGPT were tested in complex debugging tasks, its effectiveness was limited. ChatGPT may provide generic debugging suggestions for complex issues. Additional difficulties arose in tasks such as duplicate bug report detection due to issues such as normalized identifier names [68]. See

Table 5. Summary of ChatGPT’s capabilities in code generation, explanation, and debugging based on the literature.

Aspect of Code Support	Summary of Support from Literature	Limitations and Considerations
Code Generation	Aids in code completion, correction, documentation, and chatbot development [65].	Effectiveness varies across programming languages, with inconsistent outcomes and ethical considerations [67].
	It demonstrates proficiency in generating code solutions for common coding problems, providing valuable examples for learners, and assisting in code completion, correction, documentation, and chatbot development [65,69].	Risk of academic dishonesty if used without understanding underlying concepts [56,59].
	Tends to produce shorter, more concise code than non-GPT-based approaches [66].	Limited in feature engineering and accuracy [68].
Code Explanation	Generates comprehensive documentation for programming tasks, aids understanding of code elements, and simplifies complex codebases [65].	Sometimes, it deviates from provided instructions and has difficulty translating generated explanations into executable code [56,67].
	Proficient in specific tasks such as method name suggestion and log summarization [68].	Human feedback and collaboration enhance effectiveness [66].
Code Debugging	Provides real-time detection of syntax errors and suggests solutions for common mistakes [65].	Suggestions may fail to compile in integrated development environments [56].
	It may offer generic debugging suggestions for complex issues, with effective constraints for more intricate tasks [70].	Faces challenges in tasks like duplicate bug report detection due to issues such as normalized identifier names [68].

for a summary of ChatGPT’s capabilities in code generation, explanation, and debugging.

3.5. RQ4: Challenges in the Integration of ChatGPT in Higher Education

We identified several challenges and barriers to the integration of the ChatGPT in education. These include concerns regarding accuracy, privacy, and plagiarism. There are concerns about the tool’s impact on critical thinking skills and that its responses can be influenced by biases or by the quality of training data. There is also the potential for privacy risks associated with misuse of personal data and the potential to facilitate academic misconduct, such as plagiarism.

Of the 26 papers analyzed, 15 discussed the impact of the ChatGPT on students’ critical thinking. This topic is of significant concern in educational research [46]. Some perspectives suggest that ChatGPT can stimulate critical thinking and creativity [47] because crafting effective prompts for ChatGPT requires a deep understanding of the tool and underlying content. Therefore, Crafting prompts demand students’ critical thinking skills [56]. Rasul et al. [47] suggested that educators can promote discovery learning (i.e., through play and investigation, students come up with new concepts) to enhance problem-solving skills so that students’ use of ChatGPT involves deeper reflection. While ChatGPT offers unparalleled ease and convenience in generating responses, there is growing apprehension that its use may promote intellectual laziness among students [45]. Becoming reliant on suggestions could hinder students’ development of critical thinking and problem-solving abilities [55]. Students could fall into passiveness, whereby they might receive whatever ChatGPT provides without ever assessing either the veracity or relevance of such responses [45]. However, its impact on critical thinking still requires investigation [47].

Of the 26 papers analyzed, 20 discussed the accuracy of the ChatGPT responses. Several studies (n = 9) have pointed out instances where ChatGPT responses are inaccurate or inconsistent [49,53,54,57,63,67,71]. Many inaccuracies seem to arise from model biases or variations in the quality of the training data [45,49,53]. Privacy concerns and copyright issues also affect the accuracy of ChatGPT responses [47,53]. Accuracy may also vary depending on the complexity of the task and the programming language involved [66,67]. Although ChatGPT can generate code for a broad range of tasks, its performance is inconsistent, resulting in syntactically correct code in some instances and bugged or no code in others [67].

Of the 26 analyzed papers, 4 discussed privacy issues related to the ChatGPT. The use of ChatGPT in educational settings has raised significant privacy concerns [53]. A primary concern is that personal information can be used for purposes beyond personalized learning, such as targeted advertising or other commercial activities [53]. Moreover, data breaches or disclosure of sensitive information cannot be completely eliminated [53]. Recent incidents have increased the AI technology’s privacy vulnerabilities, including in ChatGPT, where leaks took place to raise the alarm about the protection of user information [71]. In this regard, ChatGPT often operates without explicit user consent [71]. ChatGPT may also raise copyright concerns, as the model may have been trained on content protected by copyright and may have provided responses similar to that content [47].

A total of 5 of the 26 analyzed papers discussed bias in the ChatGPT responses. Biased training data may lead to ChatGPT perpetuating inaccuracies, thus hindering genuine learning [45,47]. Without proper verification, students may internalize misleading content [55]. There is much concern that biased information could reinforce negative stereotypes and harm learners [57]. Efforts have been made to address biases, such as implementing plugins for updated data, but this issue currently persists [47].

Educators are at the forefront of developing habits of using ChatGPT among students. Various studies recommend the design of training programs that could assist instructors in effectively integrating ChatGPT into teaching. Such training should focus on the supplementary nature of ChatGPT and the need to encourage critical thinking [45,50]. To avoid overdependence on ChatGPT, educational institutions should focus on active learning experiences that foster critical thinking, problem-solving, and independent inquiry [49,59]. Institutions should establish clear policies and ethical guidelines for using AI models like ChatGPT to protect student privacy and academic integrity [45,71]. To incentivize student engagement and discourage passive reliance on AI-generated content, assessments should measure the application of concepts and not rote memorization [56]. Assessment strategies must evaluate learning processes and foster collaborative learning while safeguarding against plagiarism [47]. Educational initiatives should prioritize students’ development of AI literacy, enabling them to navigate ethical considerations and potential biases in AI-generated outputs [43].

4. Discussion

This study explored the potential of integrating ChatGPT into education to enhance learning experiences, academic writing, and coding tasks. The key findings are discussed in the following sections, and recommendations are provided for the successful, ethical, and responsible integration of ChatGPT in higher education.

Educators increasingly use ChatGPT to personalize learning by tailoring content to individual needs and fostering engagement and self-regulation of learning processes [72]. As such, ChatGPT provides targeted explanations and activities according to students’ proficiency and learning style. On the other hand, challenges such as response accuracy and reliability need consideration; uncritical acceptance of AI suggestions can harm cognitive development in learners [49,53]. Ethical use guidelines and accuracy measures should be integrated into the curriculum to ensure responsible use of AI while enhancing personalized learning and supporting student success [73].

ChatGPT has the potential to help improve academic writing by providing brainstorming and suggesting alternatives; however, it also brings a number of challenges, such as over-dependence on the tool, which reduces critical thinking and independent writing [36]. This should be used not as a solution but as a supporting tool for writing improvement. Educators should guide students in responsibly using ChatGPT, emphasizing information verification, critical thinking, and ethics in AI use. This includes integrating AI content with the students’ styles and voices in their writing, ensuring a culture of integrity to avoid overreliance, and enhancing the educational value of the tool [63].

The integration of ChatGPT into coding tasks is limited; it may help explain code and solve frequent problems mainly, as stated in [43,69]. While ChatGPT can develop code snippets and explain concepts in programming, critical thinking, and problem-solving need to remain central. This encourages students to review and verify the code generated by ChatGPT for better understanding and peer learning. The educator can guide students in workshops or labs on using AI technology responsibly and interpreting the responses correctly.

4.1. Recommendations for the Integration of ChatGPT in Higher Education

Our findings on the integration of ChatGPT into higher education are replete with opportunities and challenges. First, there is the implication of how the instructors contemplate the continually developing role that AI technologies, such as ChatGPT, take within a classroom environment. While ChatGPT is able to support teaching with its ability to provide fast feedback and individual attention, limitations related to its level of accuracy and depth create great pedagogical concerns. Educators have to find a delicate balance between using the capabilities of ChatGPT and making students develop critical thinking skills and domain-specific knowledge. Another teaching implication in higher education could be that traditional methods of instruction will need to be revisited due to the growing demand to integrate with AI. As AI technologies in education continue to increase, so too must strategies to have students interact with and critically evaluate content generated by AI. This includes the integration of overt instruction in information literacy, digital ethics, and critical thinking within the curriculum. Moreover, the results highlight the need for a collaborative culture and an interdisciplinary dialogue between educators, AI developers, and domain experts. Working together allows stakeholders to co-create innovative approaches that harness the power of AI technology while guarding against bias, misinformation, and over-reliance.

While ChatGPT holds great potential for personalized learning, educators must ensure that such a utility does not undermine student autonomy. Students need to be encouraged to critically use AI-generated content. Teachers should encourage active learning techniques for self-regulation, critical analysis, and problem-solving rather than passively relying on ChatGPT for an answer. According to Hattie [20], students have to develop higher-order thinking skills, which can only be achieved if students are engaged in active learning. Therefore, educators should integrate ChatGPT in a manner that would spur curiosity and independent thinking, yet also give guidance on how to verify information, question responses, and integrate AI-generated content with their ideas.

The integration of ChatGPT into the classroom requires teachers to shift from traditional providers of knowledge to facilitators of learning. As AI increasingly supports the process of information retrieval and content creation, educators’ roles will shift toward fostering critical thinking, ethical decision-making, and collaborative learning. Teachers must guide students in evaluating and interpreting the vast amounts of information provided by AI [74]. This shift requires a professional development process for teachers to equip them with the skillsets needed to responsibly integrate AI into teaching practices. Teachers should, in turn, work on training them to use AI to support learning processes while not forging academic creations. We summarized recommendations from the studies that are listed in

Table 6. A summary of the recommendations for institutions, educators, and students.

For Institutions	For Educators	For Students
Formulate wide-ranging policies and directives on ethical utilization of ChatGPT [45,48,52,53,59].	Establish clear guidelines on when and how ChatGPT should be used [45,52].	Recognize the limitations of ChatGPT and not rely solely on its responses [43,44,49,52].
Offer faculty training programs on ChatGPT’s capabilities and limitations [50,55].	Design assessments that focus on critical thinking skills [52,59].	Cross-check information obtained from ChatGPT with reliable sources [43,44,52].
Foster transparency in ChatGPT usage and encourage reporting of concerns [52].	Educate students on ChatGPT’s capabilities and limitations [44,50,52].	Evaluate ChatGPT suggestions critically, considering context and relevance [43,44].
Conduct regular evaluations of ChatGPT’s impact on teaching and learning [50,59].	Prompt students to reflect on ChatGPT interactions [52].	Incorporate ChatGPT insights into the learning process alongside other resources [52].
Address ethical and legal considerations related to ChatGPT usage [45,52].	Leverage ChatGPT’s adaptive learning features for personalized feedback [48,59].	Seek clarification from instructors or consult additional sources if unsure about ChatGPT responses [50].
Encourage collaboration among educators and departments for effective ChatGPT integration [52,59].	Educate students on privacy and data security when interacting with ChatGPT [45,50,52,63].	Avoid plagiarism by adequately citing ChatGPT-generated content in academic writing [71].
Stay informed about advancements in AI technology and adapt institutional policies accordingly [52].	Continuously review and update guidelines based on feedback and advancements [45,63].	Seek guidance from instructors or peers when uncertain about ChatGPT usage [50].

below, along with the implications of our findings.

4.2. Implications

The findings of this study form a basis for integrating ChatGPT into educational practice, as it may enhance educational outcomes under certain conditions. Personalization in learning and immediate feedback raise the level of participation of students for better learning outcomes [62]. Therefore, this section presents recommendations for institutions, educators, and students in integrating ChatGPT into higher education and discusses ethics and future research directions.

4.2.1. Implications for Institutions

To this end, an institution is expected to provide an enabling environment through which AI tools, including ChatGPT, can be utilized. Therefore, they need to make broad policies and guidelines on effective ways of using AI technologies for teaching and learning while ensuring institutional ethical practices and fostering responsible AI practices [45,48,52,53,59]. Institutions should also focus on infrastructure development and resource allocation to train educators regarding these technologies. In this regard, they would be able to create synergies between academia and the technology industry to support the integration of AI tools, like ChatGPT, into curricula in a way that is both effective and responsible [50,55]. Moreover, institutions must keep themselves updated on the developments of AI technologies and adjust their policies accordingly. As ChatGPT and other AI tools continue to evolve, institutions must also be nimble and revise guidelines in light of emerging best practices to maintain an educational environment conducive to innovation [50,59].

As universities increasingly use AI technologies like ChatGPT, it becomes a balancing act between embracing innovation and ensuring such uses of AI tools do not undermine academic integrity or traditional teaching methods. The challenge lies in regulating the use of ChatGPT in ways that foster educational enhancement while maintaining standards of quality, fairness, and ethical conduct. With clear policies on academic integrity, personalized learning, and ethical use, universities would, in effect, be able to control ChatGPT’s use and plant the seeds for an innovative educational environment. The secret to success in higher education will be finding a balance between integrating AI and traditional teaching practices so that AI enhances learning experiences without undermining critical thinking and independent learning. Following best practices, universities can ensure that ChatGPT and similar AI tools become valuable, ethical partners in the classroom. For example, a university can state that students may use ChatGPT in draft form or to see how to say something in the text. Students always need to locate where the AI-generated content is used and edit and refine such output themselves, demonstrating critical thinking in idea development. This ensures that AI is a tool to enhance, not replace, students’ creative and academic abilities. Universities can use ChatGPT as AI tools that grade or provide comments on assignments but ensure human checks to eliminate objectivity and equity. A particular institutional policy may define how the faculty members will review assessments or feedback from AIs so that the educational objectives would align and have academic rigor intact.

4.2.2. Implications for Educators

For educators, this provides another avenue in which ChatGPT can aid in student-centered learning. The teacher can utilize ChatGPT to provide feedback and encourage students toward active learning. It is also critical, however, that educators be aware of the technology’s limitations: it cannot recognize context fully, nor does it allow for many prompts in critical thinking ability [44,50,52]. Thus, educators should design assessments emphasizing critical thinking skills and encourage students to cross-check information obtained from ChatGPT with other reliable sources [52,59]. To fully exploit the potency of ChatGPT in teaching, there is a need to provide training for educators on both its capabilities and limitations [45,52], which would enable it to be better integrated into an educator’s practice in such a way as to promote learners’ autonomy rather than dependency on the AI tool itself.

4.2.3. Implications for Students

The role of ChatGPT is to inform students about their strengths and weaknesses. Although ChatGPT can support students richly, a student has to realize that it cannot replace critical thinking and personal effort, as stated by [36,37,47]. Students should be encouraged to evaluate ChatGPT’s suggestions critically about the context and relevance of the information [43,44]. In addition, ChatGPT should be used in combination with other resources, not as a primary source of information. Furthermore, students should be taught about privacy and data security when using ChatGPT or other AI platforms [43,44,49,52]. They should also be warned against plagiarism by correctly referencing any AI output in academic work [71]. As a last resort, they should also check with instructors for anything they are unsure of by using ChatGPT or seek additional sources for further clarification to fully understand what is being presented to them [50].

4.2.4. Ethical Considerations

The ethical implications of using ChatGPT in education are significant and need to be considered by educators and institutions alike. Overcoming biases and privacy concerns is crucial in gaining trust in AI-powered educational tools. Ethical considerations should be embedded in policies and training programs so that AI will be used responsibly in educational settings [75]. It is essential to conduct more longitudinal research showing how ChatGPT influences student learning over a longer period and leads to the workforce. Research can also be used to investigate how ChatGPT can be further customized for different kinds of learners, providing personalized support for a wider population of students. By focusing on this, further research can lead the way to the appropriate development of AI tools to achieve students’ learning outcomes while infusing academic integrity.

4.2.5. Extending ChatGPT’s Tutor or Assistant Role and Future Research Agenda

We provide a matrix (See Figure 4) that gives a theoretical framework for positioning AI (ChatGPT) beyond the typical tutor role. Incorporating key critical AI perspectives, like Human-in-the-Loop learning models and AI as an Epistemic Agent, ChatGPT, and other tools, can be further developed toward collaborative and reflective partners in educational processes. This not only extends the role of AI but also ensures students develop critical thinking, autonomy, and deeper engagement with content, with the teacher guiding the ethical and reflective aspects of the learning process.

Passive Engagement: Information Retrieval and Contextualization

At the level of passive engagement, ChatGPT acts classically as an assistant. ChatGPT engages the student upon demand by retrieval or explanations; it is about retrieving information with AI while the teacher controls against compliance and norms for consistency on AI use or purposefulness in pursuing learning objectives. This supports the findings by Luckin et al. [76], who stated that the earliest phases of adopting AI in education often position the role of AI as supportive, providing information to students or advice. Here, the teacher is a facilitator who ensures that AI is appropriate and benefits the student’s learning process. According to Baker et al. [77], the teacher’s role in contextualizing AI for the student is important in making the interaction between the student and the AI tool responsible and productive. The teacher positions ChatGPT as a passive engagement partner to ensure its capabilities align with pedagogical objectives.

Active Collaboration: Guidance and Epistemic Agent

As the student becomes more familiar with AI, the model shifts into Active Collaboration, where ChatGPT takes on a more involved role. Beyond information retrieval, AI facilitates Collaborative Learning: it may suggest novel ideas, stimulate questions, and assist the learner in exploring several perspectives on one issue. Based on the replies, the learner actively interacts with the AI as he refines his understanding of the subject of interest. The role of the teacher in this phase is to guide and monitor the students to ensure they are not passively following but rather thinking deeply about what AI suggests. The Human-in-the-Loop (HITL) model comes into play during this phase, where the AI and student work in tandem, each influencing the other’s input. Holstein et al. [78] emphasize how such HITL models encourage more active and adaptive learning processes where AI does not just present the student with inert content but modify that content to the developing needs of the student. The collaborative interaction underlines that AI can do much more than answer questions; it can create an environment where students engage with AI as collaborators in dynamic inquiry, not just as tutors.

Critical Thinking—Reflection and Ethical Oversight

In the Critical Thinking stage, students interact with the AI tool to analyze and reflect on the information provided. In this respect, ChatGPT acts as an Epistemic Agent at this stage, leading the student through processes of inquiry, critical evaluation, and synthesis. ChatGPT does not deliver answers but rather creates a dialogical space where the student can explore ideas, question assumptions, and construct deeper understanding. In this model, the teacher’s role is further emphasized. Teachers are responsible for facilitating critical thinking and ensuring that AI is used to support autonomous development. The teacher encourages students to question and assess AI responses, fostering a reflective process that helps the student construct knowledge in a deeper, more meaningful way. Popenici & Kerr [79] emphasize the teacher’s role in encouraging students to question the AI output and reflect on their own thinking skills, which are considered higher order in nature.

Knowledge Construction—Collaboration and Reflection

The last stage is Knowledge Construction, where co-construction involves the student, AI, and teacher. With ChatGPT now fully placed as an epistemic agent, it works with the student to guide new knowledge creation. Students synthesize information and develop insights into applying learning to real-life situations through active collaboration. At this stage, the teacher assumes the role of overseer and facilitator of reflection to keep the process in line with academic and ethical standards. The teacher gives students space to reflect on how they construct their knowledge, manages the application of critical thinking, and ensures that the learning process is rigorous and ethically sound. Stages also align with views [80] that AI most powerfully contributes to education in knowledge creation, where all participants—AI, students, and teachers—collaborate to shape new understandings.

Considering the above discussions, the following research questions need to be explored in future work.

(a) How does the use of ChatGPT in educational settings influence students’ academic performance compared to the traditional teaching methodology? (b) How does ChatGPT enhance the development of students’ critical thinking, problem-solving, and analytical skills? (c) How does ChatGPT influence student participation and interaction in individual and group learning environments? (d) How do educators balance AI-generated content with human feedback to promote meaningful student learning?

These research questions represent ways ChatGPT, and similar AI tools will continue shaping educational experiences, creating possibilities for improved learning outcomes, student engagement, and deeper development of students’ paths. In working out how to answer such areas, researchers will better understand how AI within the classroom should be leveraged to support students and educators alike.

4.3. Limitations

This review had several limitations. The data collection for this study was conducted over a period of only one year, and more papers could have been published at the time of writing. Additionally, the number of articles in the three categories was not evenly distributed: personalized learning, academic writing, and coding tasks. Similarly, the distribution of articles across categories-personalized learning, academic writing, and coding-is not uniform, which may affect the recommendations suggested in each domain. Given these limitations, further research should be done to extend the search period and include more studies to analyze the long-term effect of ChatGPT on student learning outcomes at different levels of academic disciplines. Additionally, assessing the attitudes and perceptions of students toward ChatGPT would shed light on the usage of this AI tool in educational contexts. There are some methodological limitations of the included studies. Limitations identified in the reviewed studies are small sample sizes and lack of representativeness (e.g., [54,63]). Many focus on narrow data types or short-term effects (e.g., [63,66], while methodological clarity (e.g., [53,65]) is often insufficient. Quite oftentimes, quality output generated by AI, particularly in coding, is not assessed properly, as noted in two studies (e.g., [66,67]). Also, some of these studies have sampling bias [58] and lack of diverse data sources [64], limiting the generalizability of the findings. These recurrent issues raise the need for more holistic, transparent, and ethical research in AI applications in education.

5. Conclusions

This review points out ChatGPT’s potential to enhance educational practices, especially writing, coding, and personalized learning. While it is a tool that provides valuable support, challenges include the risk of overreliance, the propagation of biases, and the inability to fully encourage critical thinking, self-directed learning, or the development of problem-solving skills necessary to pursue autonomous learning. Suitable integration into the curriculum with special emphasis on ethical guidelines by educators would contribute to mitigating these risks. This review highlights areas where ChatGPT might benefit education. However, most of the studies reviewed included small samples, and the studies lack longitudinal data, limiting the generalizability of findings. Although ChatGPT has shown promising results in some contexts, there is a need for more comparative studies regarding its effectiveness compared to other learning tools. Long-term research on learning outcomes of ChatGPT use, for example, is warranted, as is capturing student perceptions and experiences of this tool to define best practices in its integration into higher education. Future studies should involve more extensive and diverse samples and include longitudinal designs to assess the long-term impact of ChatGPT on learning outcomes.