Introduction

Time efficiency is an essential aspect of productivity in academic research. Recently, artificial intelligence (AI) technologies have been able to provide new mechanisms for optimizing scholarly activities. Although the development of artificial intelligence-based language models began some years ago, the awareness and engagement of the general public and institutions regarding these technologies have been very significant since the release of Chat Generative Pre-Trained Transformer (ChatGPT) by OpenAI in November 2022. The recently developed ChatGPT as example of AI, uses the Generative Pre-trained Transformer (GPT) model. This model uses natural language processing (NLP) a subdiscipline of computer science dedicated to the computational treatment of human language aimed at providing human-machine interaction [1].

Currently, AI systems can be used for literature searches, citation management, language improvement, as well as the creation of entire manuscript sections like abstracts and introductions [2,3]. For many authors, especially non-native English writers, AI systems can be helpful in improving clarity, coherence, and grammatical correctness in their writing. Such improvements can be highly advantageous in creating scientifically sound manuscripts [4]. At the same time, AI writing assistants can be highly efficient in performing tasks like scanning the entire body of literature to provide relevant references and creating structured introductions that place the research in the context of the broader body of research [5,2]. In this regard, AI systems have been identified as having the potential to be highly transformative in the process of writing. For instance, AI systems can be used to improve linguistic accuracy, structural organization, citation accuracy, as well as journal-specific guidelines. In doing so, they can free up the researcher to focus on higher-level tasks like critical thinking, interpretation, as well as concept creation. In this regard, the contribution of AI to research efficiency can be highly significant. In recent times, generative AI has been identified as a new paradigm that can create content independently. Its ability to assist in idea creation, research gaps identification, literature synthesis, research proposal creation, methodological structuring, as well as preliminary data interpretation suggests that there has been a major improvement in its functionality compared to its predecessors [6,7].

Despite these possible benefits, there are concerns regarding accuracy, integrity, and transparency when artificial intelligence is employed for writing academic work. For instance, there is evidence of 70% of false references and citations when artificial intelligence is solely employed for drafting academic work, and it is necessary for fact-checking before it meets scientific standards [8,9]. Similarly, there is evidence of poor performance of GPT-4 when employed for writing academic work, especially regarding American Psychological Association (APA) style academic writing, where there is a need for validation of artificial intelligence-generated work [10,11]. These findings suggest that artificial intelligence is not able to replace human expertise and judgment for writing academic work and is only able to work under human supervision [12,13]. Considering these possible benefits and risks of employing artificial intelligence for writing academic work, there is an urgent need for defining its use for research writing. For instance, there is evidence of possible risks and benefits of employing artificial intelligence for writing academic work, where there is a need for balancing its possible benefits and risks [12,14]. Based on these possible risks and benefits of employing artificial intelligence for writing academic work, this paper seeks to discuss the use of artificial intelligence in academic writing and research, including its advantages and disadvantages.

Methods

Study Design

This study was adopted through a structured narrative review, which aims at synthesizing and critically examining the existing literature on the application of AI, particularly generative AI systems, in the process of academic writing. This review seeks to synthesize the existing literature on the benefits, risks, and challenges of using AI in the process of academic writing.

Literature research strategy

A comprehensive literature search was carried out using the electronic database PubMed, Scopus, Web of Science, and Google Scholar. The time frame of the study was set from January 2019 to February 2026, a period that was carefully selected to capture the development and usage of generative artificial intelligence tools in the wake of the public release of large language models such as ChatGPT. The process of reviewing was applied using a combination of keywords connected by Boolean operators to retrieve pertinent information. The main keywords used in the study were ("artificial intelligence" OR "generative AI" OR "large language models") AND ("academic writing" OR "scientific research" OR "scholarly publishing") AND ("ethics" OR "benefits" OR "risks" OR "challenges"). To ensure a thorough and nuanced retrieval of pertinent studies, a variety of terms associated with particular benefits, risks, and challenges were used in the study. To ensure a comprehensive retrieval of pertinent studies, the reference lists of pertinent studies were used to retrieve studies that could not be obtained using the database search process.

Study Selection Approach

Studies were selected using a narrative design, where studies were selected on the basis of their thematic pertinence to the use of artificial intelligence in academic writing, including benefits, limitations, ethics, and policy implications. No particular study selection protocol was used, nor was a particular set of inclusion or exclusion criteria applied, nor was a risk of bias tool used to assess the eligibility of the studies. Both investigations, including quantitative, qualitative, and mixed-methods studies; and scholarly works, including narrative, systematic, editorial, legal, and ethics studies, were used to ensure a comprehensive representation of the pertinent studies. The studies obtained were categorized into thematic domains.

Results:

Benefits of Artificial Intelligence in Research Paper Writing

Regarding the benefits of using AI in research-paper writing, the reviewed studies indicate several benefits, as presented in Table 1. Some of the most reported benefits were generating ideas, helping with drafting, improving the quality of language used when writing and making writing more efficient in terms of time taken to write. Various research found that AI can improve the grammar, vocabulary, coherence and clarity of all written documents [4,15-18]. Also, it has been well documented by researchers that use of AI will lead to much less drafting and revising time, thus improving the overall efficiency of the writing process [8,19,8,3,20]. AI is used for generating ideas, outlining and organizing a research paper [21-24] as well as scanning literature, summarizing literature, and organizing information [14,4]. The use of AI has been empirically shown to improve writing performance, writing engagement, writing self-efficacy, decrease writing cognitive load, and decrease writer's block [15,25-27]. Thus, AI has the potential to improve how well and how much a writer writes, and how efficiently a writer conducts research, while acting as a writing assistant [28-32]

Table 1. Summary Table of Studies: Benefits of AI

Risks and Challenges of Artificial Intelligence in Research Paper Writing

Concerning risks and challenges of using AI in writing research papers, the present study reviewed A total of 21 studies published between 2019 and 2025 examined the risks of using AI in research paper writing, as shown in Table 2. The studies revealed numerous risks and challenges, including the following: Plagiarism and threats to academic integrity were the most frequently reported concerns [12,25, 13, 18,30, 33-39]. Closely related risks include uncertain authorship, transparency, and attribution, which challenge traditional academic standards [40-42]. Several studies also highlighted AI hallucinations, misinformation, and fabricated references, which may reduce the reliability of scientific manuscripts [30,33,34,43,44]. In addition, bias and inequitable access to AI technologies were reported as ethical concerns that could influence research outputs and widen disparities between researchers [13,35,37,34,40]. Another recurring theme was over-reliance on AI and the potential decline in critical-thinking and writing skills [12,33,44,38,45]. Furthermore, legal and regulatory uncertainties, including copyright ownership, privacy risks, and policy gaps, were emphasized in several analyses [46, 47].

Table 2. Summary Table of Studies: Risks of AI

Discussion

The integration of Artificial Intelligence into academic writing represents a significant paradigm shift, offering both transformative benefits and considerable challenges to scholarly practice. This discussion synthesizes contemporary findings to illuminate the multifaceted impact of AI on research paper writing, drawing upon recent literature that explores its contributions to linguistic refinement, workflow efficiency, and cognitive support, while simultaneously addressing critical concerns related to academic integrity, data reliability, and the potential degradation of critical thinking skills.

Benefits of AI Integration in Scholarly Writing

The writing of research papers has been enhanced by AI tools in many ways through three broad areas: progress made by using AI-based tools to augment language and productivity and the ability of such tools to improve cognitive supports for writers. Language enhancement and coherent structure were found to be the most significant areas where AI tools improved the overall quality of written material, such as using AI to improve the accuracy of grammar and improve how vocabulary is used and improve how sentences are constructed and how text as a whole is organized with respect to flow [15,22, 16]. While not exclusively found in the English language, these enhancements provided a reliable means of ensuring equal opportunity for all researchers to communicate with clarity and precision within the context of the global academic community [4,16]. Research has also demonstrated that writing at the sentence level and decreasing cognitive load while creating drafts enhances writing fluency and overall coherence, again, under continual human supervision [17,31,8].

In addition to helping improve language, AI also aids in making workflows run more efficiently and increasing productivity. The ability to automate repetitive and lengthy tasks, like scanning for relevant literature, summarizing articles, abstracting papers, and editing document structure, allows researchers to focus on applying cognitive resources towards two areas—conducting conceptual analyses and refining methodology [28,8,21,19,14]. In addition, studies report similar efficiency improvements due to the use of AI technology in writing research, particularly at the initial stages of writing a manuscript [12,28-30].

Moreover, AI provides significant ideational, conceptual, and cognitive assistance. Qualitative studies suggest that AI is useful for brainstorming, creating outlines, expanding themes and reframing arguments [21-23]. Also, there is an association between iterative interaction with generative AI and enhanced writing performance; thus, careful integration rather than passive use of AI yields the best results for users [26]. Also, AI technologies can function as cognitive tools that reduce writer's block and create intellectual momentum while developing a manuscript [27,19]. Thus, AI should not simply be viewed as a means of producing a document; instead, it is a collaborative partner for reasoning and conceptualizing.

Finally, AI provides valuable literature management and editorial support, assisting with identification, summarization, citation formatting, plagiarism detection, and structural editing [3,14,29]. These functionalities streamline the complex process of synthesizing extensive scholarly evidence into coherent narratives, improving manuscript organization and adherence to journal standards, provided that automatically generated references are rigorously verified [28,30]. The affective and pedagogical benefits are also noteworthy, with studies reporting enhanced writing self-efficacy, perceived competence, and academic confidence among students using AI tools, alongside reductions in anxiety and increased motivation [15,25]. These psychosocial outcomes are crucial for sustained scholarly productivity.

Risks and Challenges of AI Integration in Scholarly Writing

Despite the mentioned advantages, the uncritical application of AI in academic writing carries a complex range of risks and challenges, which require evaluation. The first risk is the challenge to academic integrity and originality. The use of AI-generated texts without adequate disclosure and significant modification by the human writer threatens the validity and originality of the content [12]. The core integrity concerns include plagiarism, misattribution, factual errors, and algorithmic biases, which highlight the need for disclosure of AI use in writing [13]. The questions surrounding authorship [36], detection of secret source generation [25], and the increased editor burden in quality control are significant considerations. The use of AI without disclosure may undermine trust in the peer review process and is incompatible with conventional notions of intellectual accountability and authorship accountability [41,42]. The use of AI in paraphrasing tools also poses significant risks in the detection of plagiarism, potentially evading conventional plagiarism detection tools and thus undermining academic integrity in true tests [39,38,45].

Another major risk is the accuracy, reliability, and existence of hallucinated material. It has been observed that generative AI models are prone to the occurrence of "hallucinations," which include the creation of false information, unsubstantiated facts, and non-existent sources that, though grammatically correct, may not be substantiated with empirical evidence. Such inaccuracies represent major risks, especially in healthcare simulation, which requires structured control [30]. The generation of false sources in scientific writing has major implications for the integrity of evidence-based practices [43]. Over-reliance on generative AI models may lead to a decrease in the alertness and diligence of researchers to verify sources [33]. The propagation of false information generated by AI models, without proper validation, may lead to the creation and dissemination of misinformation within the academic environment [35,13]. Systemic reviews have identified hallucinations, misinformation, and the ineffective nature of AI detection tools as major risks, which need to be addressed with effective human verification methods [34,44].

The increased accessibility of artificial intelligence tools also presents the risk of over-reliance and the erosion of critical thinking abilities. Over-reliance on generative AI can undermine critical thinking abilities and hinder the development of a distinctive writing style and voice in scholarship and research writing [12,33]. Over-reliance on AI has been linked to the erosion of critical thinking abilities and the lowering of intellectual standards in academic writing and research [44]. Over-reliance on generative AI can undermine critical thinking abilities and hinder the development of a distinctive writing style and voice in scholarship and research writing [38], and can undermine the development of authentic skill sets in students [45]. These findings point to the fact that while AI can increase efficiency in research writing, its uncontrolled use can undermine basic competencies that are essential in research writing and scholarship.

In addition, there are numerous ethical, legal, and data governance challenges that are presented by the integration of AI in research writing and scholarship. Key among the ethical challenges that are presented by the integration of AI in research writing and scholarship include issues of intellectual property, attribution, transparency, and equity concerns in access and use [40]. The necessity of transparent reporting standards in scientific research and publication is of prime importance in research writing and scholarship [41]. In terms of legal challenges, the use of generative AI presents risks of copyright infringement, data leakage, and breaches of privacy and the development of a credibility illusion, whereby the use of AI in research writing and scholarship can undermine the accuracy and truth of the research and scholarship produced by researchers and scholars due to its fluency and ease of use in research writing and scholarship [46,47]. In terms of data governance, the use of AI input systems presents risks of leaking confidential research data and undermining compliance with data protection laws and regulations. Equity concerns in access and use are also a major issue in the use of AI in research writing and scholarship, as unequal access to high-performance AI can undermine the development of research and scholarship in different parts of the world and exacerbate the global knowledge gap in research and scholarship in different parts of the world [37,34,35].

Lastly, methodological and epistemological implications in the integration of artificial intelligence. In fact, AI systems usually lack expertise in a particular field, and this could bring about methodological distortions in their use in processes such as peer review and research synthesis. This creates epistemological concerns with regard to the legitimacy and comprehension of machine-based analyses. In fact, the persuasive qualities of machine-written text can create a credibility illusion [46]. In this respect, there is a tendency to inadvertently influence peer reviewers and editors, leading to a compromise in the quality control processes that are in place in the publication of research articles.

Conclusion

The literature reviewed reveals that the main domains of benefits for the use of artificial intelligence in the process of research writing include: linguistic enhancement, which refers to the improvement of grammatical accuracy, vocabulary, coherence, and style; structural support, which refers to outlining, sectional construction, abstract creation, and other structural elements; ideation support, which refers to the creation of ideas, avoidance of writer’s block, and cognitive support for the writer’s ideas; productivity, which refers to the saving of time in the writing process; and auxiliary support, which refers to scanning, summarization, plagiarism detection, and other types of support. Therefore, the above domains of benefits illustrate the multifaceted nature of the value that is attached to the use of artificial intelligence in the enhancement and development of the process and the product of research writing. At the same time, the main domains of risks for the use of artificial intelligence in the process of research writing include: the risk of plagiarism, the risk of authorship confusion, the risk of hallucination and fabrication, the risk of misinformation, the risk of the erosion of critical thinking skills, the risk of the erosion of writing skills, the risk of bias, the risk of infringement of copyright, the risk of infringement of data privacy, the risk of authorship, the risk of undermining the editor, and the risk of undermining the writer. Therefore, the above domains of risks illustrate that, despite the obvious benefits that can occur from the use of generative artificial intelligence in the authoring of the research paper, the risks that can occur from the use of artificial intelligence in the process of research writing cannot be ignored.

Recommendations for the Responsible Use of Artificial Intelligence in Academic Writing and Research

1.Using AI as a support tool rather than the author: AI is best utilized for idea generation, outlining, wording, and formatting. However, the researcher should not rely solely on AI for information and ideas in the research paper. All the information obtained from AI should be critically evaluated and adapted according to the researcher’s needs and goals.

2.Human Originality and Interpretation: Conceptual frameworks, interpretation, methodology, and reasoning should be the responsibility of the researcher. However, the researcher should utilize the information obtained from the AI for drafting the paper and edit the information to adapt it according to the research paper needs.

3.Fact-Checking, Citations, and Methods: Researchers should be aware of the fact that the information obtained from the AI could be false and even include false citations (hallucinations). Researchers should fact-check all the information, methods, statistics, and interpretations obtained from the AI and included in the research paper.

4.Maintaining Academic Integrity and Avoiding Plagiarism: Researchers should avoid using the information obtained from the AI directly in the research paper and should edit the information heavily before using it in the research paper. Researchers should be careful about the citation of the information obtained from the AI and should be careful about the similarity of the paraphrased information and the similarity index.

5.Ensure Transparent Disclosure of AI Use:

The extent and functional utility of AI assistance, e.g., in language editing, summarization, or outline creation, need to be mentioned in the methods or acknowledgments sections. AI tools should not be included in the list of authors; only human authors should be held responsible.

6.Adhere to Journal and Institutional Policies:

Researchers need to adhere to AI policy guidelines set forth by various journals, research institutions, and organizations, with special emphasis on the policies governing authorship, figure creation, and AI content acceptance.

7.Develop AI Literacy and Iterative Evaluation Skills:

For the efficient and appropriate incorporation of AI tools, researchers need to be trained in the skills of prompt engineering, critical evaluation of AI output, and iterative refinement of AI output. Training programs need to focus on the recognition of AI output limitations, potential biases, and ethical implications of AI tools.

8.Safeguard Scholarly Competence, Privacy, and Equity:

Overdependence on AI tools results in the loss of scholarly competence; hence, the need for the appropriate and balanced use of AI tools is critical. Researchers need to ensure that they do not upload confidential information to unregulated AI platforms.

9.Develop and Implement Institutional Disclosure and Governance Frameworks:

Universities and research organizations should adopt robust institutional frameworks for the governance of AI, including the definition of appropriate usage, copyright issues, data protection, and ethical oversight mechanisms. The need for disclosure is essential to promote editorial trust.

10.Enhance Editorial and Peer-Review Safeguards:

Academic journals should ensure the declaration of AI usage, deploy AI-detection tools and human expertise, and provide appropriate guidelines to the peer-reviewers to identify inaccuracies resulting from AI usage. The editorial mechanism should promote intellectual accountability and methodological rigor.

11.Integrate AI Literacy into the Research Training:

The graduate and postgraduate research curricula should include AI literacy components, including the identification of fabricated references, the assessment of AI model bias, the evaluation of the epistemic reliability of AI models, and the appropriate usage of AI models in iterative research processes.

12.Promote Data Security and Regulatory Compliance:

Institutions must ensure that the AI systems they provide are secure, institutionally approved, and compliant with relevant data security legislation. Researchers must ensure that the use of AI is compliant with ethical norms and the applicable law on data privacy.

13.Advance Equitable Access to AI Technologies:

Policymakers and academic institutions must ensure that there is equitable access to AI tools. Proactive measures must be taken to ensure that existing inequalities in the productivity of researchers, the competitiveness of funding, and the global participation of scholars are not exacerbated.

Funding

The authors declare that this research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Conflicts of Interest

The authors have no conflicts of interest to report.