ChatGPT in academic writing: Maximizing its benefits and minimizing the risks

Abstract

This review article explores the use of ChatGPT in academic writing and provides insights on how to utilize it judiciously. With the increasing popularity of AI-powered language models, ChatGPT has emerged as a potential tool for assisting writers in the research and writing process. We have provided a list of potential uses of ChatGPT by a novice researcher for getting help during research proposal preparation and manuscript writing. However, there are concerns regarding its reliability and potential risks associated with its use. The review highlights the importance of maintaining human judgment in the writing process and using ChatGPT as a complementary tool rather than a replacement for human effort. The article concludes with recommendations for researchers and writers to ensure responsible and effective use of ChatGPT in academic writing.

Academic writing is an essential skill that plays a crucial role in academic and professional success. However, writing a research paper is often a daunting task, even for experienced writers. It requires a substantial amount of time, effort, and skill to conduct research, organize ideas, and present findings effectively.[1,2] As a result, many authors face difficulties during the research and writing process. Non-native speakers face more challenges in writing in English.[3]

With the recent advancements in artificial intelligence (AI), ChatGPT has emerged as a potential tool for assisting writers in the research and writing process.[4] ChatGPT is an AI-powered generative language model that can generate human-like text based on the input provided by the user. Its ability to understand natural language and generate coherent and relevant responses has made it an attractive option for writers looking to streamline their writing process.[5]

However, there are concerns regarding the reliability of ChatGPT and the potential risks associated with its use. This review paper aims to explore some features of ChatGPT that a novice author can use for academic writing. We also discuss the judicious use of the model to get maximum help for generating scientific content.

Research Map

The steps involved in research writing can vary depending on the nature of the research project, the discipline, and the requirements of the intended audience or publisher. However, here are some common steps involved in the research writing process. The steps are shown in a flowchart in Fig. 1.[6]

Figure 1.

A research map showing steps involved in the publication process where novice authors may need help, from identifying the research question to publication in a journal

The first step of a research project is to find a unique research question or problem that has not been explored before. This question should be clear, concise, and relevant to the field of study.[7] The thought process of novice authors may be so involved with minute details about the research that they may find difficulty in framing a clear research question.

The next step is to conduct a literature review. In this phase, researchers perform a comprehensive review of existing literature related to the research question or problem. This helps to identify if the question that came to mind is already answered or if there are gaps in the literature.[8] The research question can be modified accordingly. This step also helps in choosing the research methodology.

After finding the research gap, researchers find a suitable methodology for the study. This includes selecting a sample population, calculating the minimum requirement of sample size, data collection methods, methods to reduce bias, and data analysis techniques. This is one of the most important tasks as the whole study depends on choosing the appropriate methodology.[9]

In the next phase of research, the researchers collect data from the sample. Here, the researchers need to write an easy-to-understand informed consent form.[10] This may also involve conducting surveys, interviews, experiments, and collecting data from secondary sources. While conducting a survey, drafting the questions or making interview questions and probes also requires expertise.

The next phase is data analysis. This may involve using statistical software or other analytical tools. However, the statistical output should be clearly stated so that the actual meaning is understood by the audience.[11] Simplification of the complex statistical result is also important for the dissemination of the study results to common people.

After analysis and interpretation, researchers start writing a research report or manuscript for publication. This involves writing the research report, which typically includes an introduction, literature review, methodology, results, discussion, and conclusion. The report should be written in a clear, concise, and organized manner, and should adhere to the formatting and style requirements of the intended audience or publisher. A quick guideline about the sections of a research paper is found in the article by Mondal et al.[12] In addition to writing the manuscript, researchers need to write a cover letter to provide a glimpse of the research paper. This letter may be the first thing the journal editor would see. Hence, it should be written with utmost care to impress the editor with the novelty and scientific vigor of the study.[13]

When the manuscript is sent to a journal, it is rarely accepted as it is. One or more rounds of revisions are required in the publication process. This involves revising and editing the research report to ensure that it becomes more accurate, clear, and well-organized. Along with revising the manuscript, a letter is also written to reply to queries raised by reviewers. This is also an important task as reviewers are human and so are biased to some extent; a letter with humility always wins the hearts of the reviewers.[14]

Difficulty Faced

Authors often face several difficulties when preparing a research protocol or writing a research article.[15] In Indian medical undergraduate medical courses, research is rarely taught. Hence, after joining post-graduation, students are suddenly exposed to a research environment that may seem alien. Immediately after joining post-graduation, students are asked to submit a synopsis of the research project or thesis. Many students might have heard the term “synopsis” first time in their lives.

Novice authors may not have enough experience in writing a research proposal or synopsis, which can make the process seem overwhelming and confusing. They may find difficulty in searching the literature and may end up with unclear research questions. This is common for even experienced researchers. Hence, the struggle to define a clear research question or hypothesis is universal.[16] However, it is crucial for guiding the research and writing process.

Furthermore, after studying too many subjects in preparation for the entrance examination, a newly joined post-graduate student may have limited knowledge of the subject matter they are writing about. This can make it challenging to conduct research and synthesize information. The next problem is the lack of adequate time or poorly managed time for research work. The majority of Indian medical institutions have a high patient load; as such, clinicians may face time shortages for research work.[17]

When students start writing a synopsis or an article, it is a new endeavor. They may be unfamiliar with the formatting and style requirements for academic writing, which can lead to errors and inconsistencies in the paper. In addition, almost all medical journals are written in English. Hence, those who are not native English speakers may face additional challenges in expressing their ideas clearly and concisely in English.[18] Furthermore, authors may not have access to the necessary resources, such as research databases (e.g. Scopus and Embase) or specialized software (e.g. SPSS), which can limit their ability to conduct thorough research. Moreover, they may not get sufficient feedback from their peers or mentors, which can hinder their ability to improve their writing skills and produce high-quality research.

These difficulties can make the research and writing process challenging for novice authors. However, with guidance and support, they can overcome these challenges and produce a high-quality research article. In this context, AI may help a lot in gliding through the journey of a research cycle.[19]

Solution with CHATGPT

ChatGPT, as a large language model trained by OpenAI, has proven to be a powerful tool for research and academic writing. Its ability to generate human-like text, answer questions, and summarize information has made it a valuable resource for researchers and academics across a wide range of disciplines. ChatGPT can assist in tasks such as literature review, data analysis, and even writing entire sections of academic papers. Furthermore, it can help scholars save time and energy by generating ideas, providing summaries, and offering suggestions for further research. As ChatGPT continues to improve and evolve, it is likely to become an increasingly important tool for researchers and academics in the years to come.[20] In this section, we discuss its potential use by novice researchers and authors. Some of the capability and related examples of prompts we checked are enlisted in Table 1 and described briefly below.

Table 1.

Components of research map and related example prompts

Component of research		Prompt
Research question		Write the research question of the study titled “Comparative study of large language models in writing a case report in ophthalmology”
Hypothesis		Write the hypothesis of the research article titled “Knowledge, Attitude, and Practice of Artificial Intelligence in Writing Research Articles Between Residents and Faculties of Ophthalmology.”
Literature search		Write a literature review for the study highlighting the importance of research writing, the difficulty faced by Indian doctors in writing a research article, the application of AI in research writing, and why assessing knowledge, attitude, and practice between residents and faculties is important.
References		Write a paragraph of text about the “Influence of gut microbiota on eye diseases” with four references to journal articles.
Select methodology		I would like to conduct research to test the effect of drug X on treating diabetic retinopathy. What study method would be appropriate for the study?
Sample size		In a nationwide survey for assessing the perception of medical students about the use of AI in ophthalmology, how many research participants to recruit if the total number of students in India is 105000, the confidence level is 95%, and the margin of error is 5%?
Data collection		Write an informed consent form in English for recruiting participants for research on screening of refraction errors. The language should be understandable by 6th-grade students.
Data analysis		What statistical tests would be appropriate to compare a continuous variable between males and females when the data is not normally distributed?
Result interpretation		Interpret the result presented in the table and write a paragraph with major findings. [Table copied from MS Word file and pasted]
Writing drafting	Title	Write three titles of a research paper where researchers would like to assess the accuracy of three AI-based screening devices to detect diabetic retinopathy in comparison with the gold standard in a rural hospital situated in the eastern part of India.
	Abstract	Write an unstructured abstract of 250 words for the study titled “A narrative review on applicability of AI rural healthcare in India” with an emphasis on disease screening, early diagnosis, and referral.
	Introduction	Write an introduction of a research paper titled “Accuracy of a non-invasive home glucose monitor for measurement of blood glucose” describing the importance of home glucose monitoring, the necessity of a non-invasive glucose monitor, and why the accuracy of market device needs surveillance accuracy.
	Methodology	Write materials and methods of a research paper in the following subsections: Study type and setting, sample size calculation, participants and recruitment process, tools used in the study, data collection, statistical analysis, and ethics for the study where researchers aimed to assess knowledge, perception, attitude, and practice of self-directed learning by YouTube videos for cataract surgery among postgraduate students. The sample was 40 students from a single institute, a convenience sample. An online survey was conducted on Google Forms. A questionnaire was developed for this study with 10 questions each for knowledge, perception, attitude, and practice. It was pilot-tested.
	Result	Write in a paragraph the summary of the result in the table below: [Table data]
	Discussion	Despite having a positive attitude toward AI in ophthalmology, the practice is very low. Write potential reasons for the finding.
	References	Covert the following reference in APA style “[reference]”
	Grammar	Correct grammar and punctuation of the text: [Text]
	Paraphrase	Paraphrase the text: [Text]
	Simple language summary	Write a plain language summary so that a child in 6th standard can understand the text: [Text]
	Keywords	Find 10 keywords from the abstract below. [Abstract text]
	Cover letter	Write a cover letter to the editor for submission of a research paper titled “ABC” explaining the importance and implication of the study in improving healthcare delivery.
Revision	Text revision	Rewrite the text to make it easier to understand.
	Rebuttal letter	Write a rebuttal letter to the editor for submission of a revised manuscript to [Journal name] for the article titled “[article title]” where we have addressed all the comments by the reviewers except the one that asked for an example of a case report. Start with thanking the editor and reviewers for their effort in editing and reviewing.
	Copy editing	Correct only the grammar and punctuation of the provided text: [Text]
Research promotion		Write a catchy social media post for promotion of the article titled “ChatGPT in academic writing: Maximizing its benefits and minimizing the risks”

Research concept

The most difficult task perhaps in a research cycle is the selection of a research topic that is novel and feasible to conduct. ChatGPT can suggest some relevant topics based on your area of interest. It is important to remember that this language model may not always provide novel ideas but can help generate some for further consideration and modification. An example of such a question is shown in Fig. 2. As ChatGPT remembers the previous conversations, authors can continue chatting with threaded questions. For example, after generating some research titles as shown in Fig. 2, one can ask to “write in detail about the third proposal.” Researchers need not worry about asking grammatically correct questions. It can understand the questions like a human does.

Figure 2.

Response of ChatGPT to a request – “Provide some novel research topics on diabetic retinopathy”

Hypothesis generation

When the research title is provided to ChatGPT, it can synthesize the components in the title and provide the related hypothesis. However, the title should contain the most important tenets of the study; otherwise, it may provide the wrong hypothesis.

Literature review

Perhaps the domain where ChatGPT can help authors at a large scale is the literature review. With its vast knowledge base and ability to comprehend natural language, ChatGPT can quickly access numerous academic papers, books, and articles to identify relevant sources to write text for a topic. However, specific keywords or research topics should be provided for efficient output. In addition, ChatGPT can suggest new angles or lesser-known studies to enrich the literature review. For a better response, authors may divide the topic of the literature review into several segments and get the text generated for each segment at a time and then combine those. In many cases, ChatGPT starts providing text with bullets or numbering. In those cases, ask ChatGPT to “write the text in a paragraph.” It is important to remember that the literature may not be updated as the training data of the AI may be old.[21]

Generating references

The literature review is never complete without citing the source of the text. However, the generation of references from a language model is risky as it often provides fictitious references.[22] Furthermore, ChatGPT's responses are based on its training data, and it does not have real-time access to current databases or sources to verify references. Therefore, it is essential to manually verify and cross-check any references provided by ChatGPT before using them in academic or research work to ensure their accuracy. For the example in Table 1, two references were correct and the other two were incorrect.

Selecting methodology

Based on the research objectives, ChatGPT can suggest the most suitable study design that can best address the research questions and provide the desired outcomes. It can explain various types of research designs, such as experimental, correlational, cross-sectional, longitudinal, and qualitative approaches. It can provide insights into the strengths, weaknesses, and appropriate contexts for each design.

Sample size

One of the challenging tasks for the majority of the researchers is the calculation of the sample size. If the relevant information is provided, it can calculate the sample size along with the steps of calculation and details of the formula used. However, we observed that although the formulas were correct, some calculations were wrong. Hence, the calculations should be checked when finally using it in the synopsis, thesis, or research paper.

Data collection

Although language models may not help in data collection directly, they can help generate content used for data collection. They can help in writing an informed consent form for research. If the attributes are provided, they can draft a questionnaire for a survey. In addition, they can translate the questionnaire or consent form into various languages such as Hindi, Spanish, French, German, Japanese, and Chinese. It is also capable of translating regional languages such as Assamese, Bengali, Kannada, Odia, Tamil, and Telugu. However, those translations are not accurate and need further modification for use in a consent form or any other material used for research purposes.

Data analysis

After data collection, researchers generally use a spreadsheet to store the data and conduct descriptive statistical tests to find the central tendency of the data.[23] In this phase, ChatGPT can help generate Excel formulas that can be copied and used. Those who are not aware of the basic inferential statistics that can be computed in Excel can get the idea to run tests such as t-tests, and Pearson correlation. Hence, ChatGPT can help in data analysis in Excel. In addition, those who are using SPSS or other statistical software packages and want guidance for any statistical test method can ask ChatGPT. As a trained language model, ChatGPT is a good option for analyzing the text for sentiment or finding themes from the text in qualitative data analysis.

Interpretation

Result interpretation, although not a daunting task, many novice authors find difficulty in writing to convey the findings effectively. After researchers have conducted their data analysis, ChatGPT can assist in making sense of the findings by providing detailed explanations and insights. In addition, ChatGPT can answer specific questions about the research findings, assist in summarizing key points, and even suggest possible avenues for further investigation.

Manuscript drafting

Title: The title of a research paper should be catchy to attract readers by providing the whole research expressed in a line. ChatGPT has the capability of generating a title if the study details are provided. In addition, authors can get alternative title suggestions and choose the most appropriate one from the alternatives.

Abstract: The abstract is the next part of a published paper where the research is summarized in approximately 250–350 words. ChatGPT can help write the outline of the abstract, whether structured or unstructured. It can efficiently write the abstract if the exact prompt is provided, especially the sample and methodology details and key findings from the analyzed data.

Keywords: ChatGPT can generate the keywords from the abstract of a research paper. However, it may not provide keywords according to Medical Subject Headings, which is a criterion of the majority of the journal for better indexing of their articles.[24]

Introduction: The background of the study constitutes the base of the study. Here, the authors need to provide the answer to the question - “Why was the study conducted?” along with already available literature to provide a solid background. ChatGPT can write the introduction as a single prompt. However, we suggest dividing the prompt into several parts for better text output. Alternatively, clear instruction can be provided in a single prompt as shown in Table 1. However, this language model may not have updated literature as it is not yet trained using real-time data that comes daily from the internet, biomedical databases, and journal websites.

Methodology: The overall methodology or an outline is the next step. ChatGPT can assist in creating an outline for a research paper. Main points such as sample size, method of recruitment, and sample population can be provided to get the methodology written in detail. In the methodology of a research paper, many researchers face some challenges in calculating the sample size. ChatGPT can suggest the formula suitable for sample size calculation or can do the calculations and provide the calculated values. In this section, researchers need the highest input in their prompts to provide all essential components to make ChatGPT understand the study.

Result: It is a common practice to write the major findings of the study in text format and provide tables and figures presenting the major as well as minor findings. ChatGPT can write the textual interpretation of the result presented in a table and can summarize it.

Discussion: The discussion section of a study typically includes the main findings, meaning or reason of the finding, comparison with other studies, and novelty and limitation of the study along with future direction. If the discussion is not adequately written in one prompt, it can be divided into subsections, and texts can be generated for each subsection and compiled for the comprehensive discussion.

References: Although ChatGPT can generate references from journal articles or books, its accuracy is very low. It may generate fictitious references. If authors would like to find relevant literature in addition to the literature they have, they can take help from Journal/Author Name Estimator available at https://jane.biosemantics.org. ChatGPT can also assist in formatting citations and references according to the specific style guide required for your research paper. For example, if a reference needs to be changed from the NLM style to the APA style, it can be done within seconds by ChatGPT.

Grammar: This is another segment where non-native speakers can benefit a lot from the language models. ChatGPT can correct the grammatical errors of a given text and improve the presentation of the text to bring more clarity.

Paraphrasing: In many situations, authors need to paraphrase the sentences of a part of the text to present the message in a more coherent manner to avoid textual plagiarism. From our personal experience, we suggest that if authors need to present the concept in a better manner, they can use ChatGPT. However, if the objective is to avoid textual plagiarism, QuillBot (the free version allows checking 125 words at a time; users can check multiple times) available at https://quillbot.com is a better alternative.

Plain language summary: Many journals require a plain language summary of the study, which can be generated using ChatGPT. The abstract can be submitted to ChatGPT, and a prompt can be provided to generate a plain language summary from the abstract. In addition, further instructions can be provided to make it suitable for various levels of learners.

Cover letter: The cover letter appraises the editor about the research and gauges its suitability in the journal for publication. It should be short yet coherent to inform the importance of the study. ChatGPT can generate it for you. However, it may generate a long letter if not prompted properly. Those who are writing a cover letter themselves can use ChatGPT for correcting grammar and/or paraphrasing sentences for fluency and clarity.[25]

Manuscript revision

When a manuscript is rejected, it is revised either for submission in the same journal or to other journals. From our personal experience, we found that the majority of peer reviewers comment about grammar and punctuation errors. Hence, while revising, copy editing is a must, and ChatGPT can help in correcting grammar and language errors.

In addition, writing a rebuttal letter is a must in the resubmission process. It is written by an author to respond to the reviewers explaining how and where they amended the manuscript, clarify any misunderstandings, or justify why reviewers' suggestions were not addressed. Whatever may be the content, the letter should thank the editor and the reviewers for their effort in handling and reviewing the article. As this letter contains a detailed tabular reply to individual comments, a short description of points on what authors addressed (if not addressed all) and what did not address is sufficient. The letter can be generated by ChatGPT. Along with it, the textual reply to the comments of reviewers can be edited using ChatGPT for grammar and presentation.

Research promotion

After the publication of a research paper, authors can generate text for social media posts or generate highlights of the paper to share in social messenger with the link of the study. For example, it can write a catchy Twitter (now known as X) post with relevant emojis if the title of the paper is provided to it. One such example is shown in Fig. 3.[26]

Figure 3.

A social media post written by ChatGPT for the promotion of an article

Precautions to Take

Although ChatGPT can be a useful tool for academic writing, it is important to take precautions to ensure that it is used effectively and responsibly. ChatGPT should be a complementary tool, not a replacement for human effort. ChatGPT can assist in the writing process; however, it is important to maintain human judgment and critical thinking in the writing process. Authors should not rely solely on ChatGPT to write a paper. ChatGPT can sometimes perpetuate biases or inaccuracies, particularly if it has been trained on biased data.

The program does not take any responsibility for generated content. Furthermore, while ChatGPT can suggest phrasing and sentence structure, it is important to check the accuracy and intended message in the sentence as literary accuracy and scientific accuracy may differ.[27]

ChatGPT-generated text may also have plagiarized content with variations in the percentage, sometimes beyond the tolerance level as it generates text from its training data. If the text is of a common topic of interest and research in this field has been conducted since a long time, the chances of plagiarism may be high.[28,29] Hence, it is wise to check the plagiarism of the text or idea when it is being used for a research paper. We suggest the use of a model shown in Fig. 4 for judicial use of the ChatGPT-generated textual content. In addition, writers should be careful to properly cite all sources and avoid copying and pasting text directly from ChatGPT without proper attribution. Furthermore, writers should ensure that any sensitive data is kept secure and that they are aware of the legal implications of using ChatGPT for academic writing.

Figure 4.

A guideline for using ChatGPT-generated text on research articles

There is some risk if confidential data is shared during a conversation with ChatGPT or other AI-powered tools. The model is continuously being improved with user interactions and may compromise intellectual property rights.[30] Hence, data that need full security should not be shared with any AI-powered tool. In addition, any research-related work should not be done on public computers, which are at risk of being misused. This is true for ChatGPT too. It is recommended to use ChatGPT in a secure environment that is not vulnerable to hacking or data breaches.

Although machines are fascinating, they cannot replace human judgment in academic writing. ChatGPT can provide useful information and suggestions on various topics; however, it cannot think critically and creatively in the same way that humans can. Academic writing involves not only providing accurate information but also interpreting that information, evaluating it, and synthesizing it into a coherent argument or analysis. These complex tasks may require human expertise and judgment.[31,32]

Many authors may have a question about whether the use of AI in writing an article is allowed in academia.[33] Before the rise of LLMs, many authors used to take the help of medical writers or companies to write the draft on their behalf; this is well accepted in academia. However, authors need to clearly mention why and to what extent the help was taken from whom. Such service providers used to charge a hefty fee for the task. Now, LLMs can help authors who cannot afford third-party writing services. The only matter is that the authors need to take responsibility of the text.

Conclusion

ChatGPT has emerged as a potentially revolutionary tool for academic writing. It can increase efficiency and accuracy and can assist writers in producing high-quality research papers. However, it is important to recognize the potential risks associated with its use and take appropriate precautions to mitigate these risks. To ensure this, it is crucial to maintain human judgment and critical thinking in the writing process and to use ChatGPT as a complementary tool rather than a replacement for human effort. By using ChatGPT responsibly, researchers and writers can take advantage of its benefits while avoiding its potential pitfalls and contributing to the ongoing evolution of academic writing. However, excessive use of AI in writing would alter the cognitive abilities of humans. As an example, it was fascinating to write letters to family, friends, and even pen-friends to express love, care, and affection in the bygone era. How many young individuals of the present era can or have written a letter on paper to communicate with friends or families? The answer points toward the future of medical writing!

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.