The DOI matches its Crossref record; 34 of the 34 references it cites were found in a registry or an index; 7 of the 21 sources its claims cite were checked for support.
ChatGPT in academic writing: Maximizing its benefits and minimizing the risks
Open accessAbstract
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.
Overview
CoChat Summary
TL;DR
This narrative review describes how novice researchers can use ChatGPT across the academic writing cycle, from research questions and literature review to manuscript drafting, revision, and promotion. It argues ChatGPT should be a complementary tool, with human verification and judgment, because of risks such as fabricated references, inaccurate calculations, plagiarism, and confidentiality concerns.
Key findings
- ChatGPT can assist with many research-writing tasks, including generating research topics, hypotheses, literature review text, methodology suggestions, sample-size calculations, informed consent forms, data-analysis guidance, result interpretation, manuscript sections, cover letters, rebuttal letters, grammar correction, paraphrasing, plain-language summaries, and social media promotion.
- Reliability problems are reported: in one reference-generation example, “two references were correct and the other two were incorrect”; for sample-size calculation, “the formulas were correct, some calculations were wrong”; translations into several languages were “not accurate and need further modification.”
- The paper warns that ChatGPT may produce fictitious references, outdated literature, biased or inaccurate content, and plagiarized text, and that confidential data should not be shared with AI tools.
- The authors recommend using ChatGPT as a complementary tool rather than a replacement for human effort, maintaining human judgment and critical thinking, and taking author responsibility for the final text.
Methods
- Design: narrative review article exploring ChatGPT’s use in academic writing and offering recommendations for judicious use.
- Data/sample: no systematic sample or dataset; the paper draws on a research map, example prompts in Table 1, figures showing ChatGPT outputs, and the authors’ “personal experience.”
- Analysis: descriptive discussion of potential uses, precautions, and a proposed guideline for using ChatGPT-generated text in research articles.
Limitations
- The paper states that ChatGPT’s training data may be old, it lacks real-time access to current databases, it may generate fictitious references, calculations may be wrong, translations may be inaccurate, and it may perpetuate biases or inaccuracies.
- The paper states that ChatGPT cannot replace human judgment, critical thinking, or creativity in academic writing and that authors must verify accuracy, plagiarism, citations, and sensitive data security.
- Evident from the text: this is a narrative review without a systematic search strategy or empirical evaluation; many claims rely on examples, personal experience, and selected prompts rather than controlled testing.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures




Verification
Claims the paper makes
6 Valid · 27 Not valid · 14 Not checked · 11 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“It is important to remember that the literature may not be updated as the training data of the AI may be old.[21]”
§ ChatGPT in academic writing: Maximizing its benefits and minimizing the risks > Solution with CHATGPT > Literature review
What the sources say
- Supportedcites [21] Arif TB Munaf U Ul-Haque I The future of medical education and research: Is ChatGPT a blessing or blight in disguise? Med Educ Online 2023 28 2181052 36809073 PMC9946299 source
Passage [2] states that a significant barrier to using ChatGPT is its existing training data, 'which is updated till 2021,' compounded by restricted access to primary databases such as PubMed and Cochrane — i.e., its outputs cannot reflect current literature. The paper adopts this point in its own argument about the bot's limitations rather than merely crediting it, so the paraphrased claim is supported. The only nuance is that the source frames the training-data cut-off as being highlighted by Halamka and Cerrato (ref [8], Mayo Clinic Platform), which the citing sentence drops, but the substance attributed to this source is present. (LLM, conf=0.86)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the 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]”
§ ChatGPT in academic writing: Maximizing its benefits and minimizing the risks > Solution with CHATGPT > Manuscript drafting
What the sources say
- Partly supportedcites [24] Mao Y Lu Z MeSH Now: Automatic MeSH indexing at PubMed scale via learning to rank J Biomed Semantics 2017 8 15 28412964 PMC5392968 sourceExperimental — accuracy not yet measured
The Background section says MeSH terms are used to index PubMed articles and improve literature retrieval. It does not address whether ChatGPT provides MeSH keywords or whether most journals require them, so those parts of the citing sentence are unsupported. (LLM, conf=0.82)
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Authors often face several difficulties when preparing a research protocol or writing a research article.[15]”
§ ChatGPT in academic writing: Maximizing its benefits and minimizing the risks > Difficulty Faced
What the sources say
- Partly supportedcites [15] Shah J Shah A Pietrobon R Scientific writing of novice researchers: What difficulties and encouragements do they encounter? Acad Med 2009 84 511 6 19318791 PMC6035752 sourceExperimental — accuracy not yet measured
Our copy fully supports the 'writing a research article' half: the Abstract (Purpose/Conclusions) states 'Writing scientific articles is a daunting task for novice researchers' and the study set out 'to elucidate the main difficulties and sources of encouragement they encountered', with Results themes such as cognitive burden (Results, Cognitive burden) documenting time constraints, procrastination and anxiety, and the Introduction noting that these factors 'make scientific writing a daunting task' so that 'many researchers shy away from this critical element of research'. However, nothing in our copy addresses difficulties in preparing a research protocol — the only mention of a protocol is a caveat that a flawed backward-design protocol's limitations may carry into the manuscript (Discussion), and the study itself concerned sections of scientific manuscripts/term papers written by 16 novice-researcher students at Duke (Methods). The claim's protocol-preparation limb, and its unqualified extension to 'authors' generally beyond the novice/student population studied, are not supported by this source. (LLM, conf=0.72)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“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.”
§ ChatGPT in academic writing: Maximizing its benefits and minimizing the risks > Precautions to Take
What the sources say
- Supportedcites [28] Mondal H Mondal S Podder I Using ChatGPT for writing articles for patients' education for dermatological diseases: A pilot study Indian Dermatol Online J 2023 14 482 6 37521213 PMC10373821 source
The Discussion states that 'the text may have a high level of similarity (i.e., text plagiarism)' and explains the mechanism: ChatGPT is 'trained on large datasets of the text of existing written contents,' so 'there is a risk that the model may generate text that is similar to existing content.' The Results (Table 1, p. Results) give the variation in percentage (overall similarity 27.07 ± 11.46%, first quartile 19% to third quartile 35.5%) and the Abstract/Conclusion state the similarity index exceeded the expected/tolerable limit of 15% used by most scientific journals. All components of the citing sentence — potential plagiarism, variable percentage, sometimes beyond tolerance, and the training-data explanation — are in our copy. (LLM, conf=0.92)
- Partly supportedcites [29] Vanathi M Realms of publication ethics Indian J Ophthalmol 2023 71 2929 37530255 PMC10538838 sourceExperimental — accuracy not yet measured
The closing passage of 'Realms of publication ethics' (final paragraph) does support the core idea that AI-produced text can carry plagiarism and that human authors must confirm its absence: 'With the advent of artificial intelligence (AI)-assisted technologies (such as Large Language Models [LLMs], chatbots, and image creators) ... it is the responsibility of human authors to be able to confirm the absence of plagiarism in all parts of their submission, both in text and images produced by AI.' However, our copy nowhere mentions ChatGPT by name, nor any similarity-percentage variation, nor a 'tolerance level' threshold, nor the mechanism that the tool generates text from its training data; the rest of the editorial treats plagiarism only as a general category of misconduct (clear plagiarism, minor copying, self-plagiarism). Those added specifics are not attributable to this source. (LLM, conf=0.70)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“If the relevant information is provided, it can calculate the sample size along with the steps of calculation and details of the formula used.”
§ ChatGPT in academic writing: Maximizing its benefits and minimizing the risks > Solution with CHATGPT > Sample size
plainstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
2 more claims
Create a free account to see all 58 claims and to verify the 14 references still open.
- Probably not a substantial claim. 48
- It may state two separate claims as one. 15
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Among these technologies, large language models (LLMs) such as ChatGPT have demonstrated remarkable fluency in producing coherent and technically structured scientific texts across various disciplines, including health sciences and rehabilitation domains.”
From Fabricated or accurate? Ethical concerns and citation hallucination in aI-generated scientific writing on musculoskeletal topics (2025) · cites this paper as [1]
Experimental — accuracy not yet measured
Citations
Cites
What if the devil is my guardian angel: ChatGPT as a case study of using chatbots in education
Tlili, A., Shehata, B., Adarkwah, M. A., Bozkurt, A., Hickey, D. T., Huang, R., & Agyemang, B. (2023) · Smart Learning Environments · doi:10.1186/s40561-023-00237-x · Cited by 2,003 · cited 1×
“[31,32] Many authors may have a question about whether the use of AI in writing an article is allowed in academia.”
BackgroundArtificial hallucinations in ChatGPT: implications in scientific writing
Alkaissi, H., & McFarlane, S. I. (2023) · Cureus · doi:10.7759/cureus.35179 · Cited by 1,540 · cited 1×
“[31,32] Many authors may have a question about whether the use of AI in writing an article is allowed in academia.”
BackgroundChatGPT is fun, but not an author
Thorp, H. H. (2023) · Science · doi:10.1126/science.adg7879 · Cited by 1,437
Can artificial intelligence help for scientific writing?
Salvagno, M., Taccone, F. S., & Gerli, A. G. (2023) · Critical Care · doi:10.1186/s13054-023-04380-2 · Cited by 827
Haleem, A., Javaid, M., & Singh, R. P. (2022) · BenchCouncil Transactions on Benchmarks, Standards and Evaluations · doi:10.1016/j.tbench.2023.100089 · Cited by 694
ChatGPT and the future of medical writing
Biswas, S. (2023) · Radiology · doi:10.1148/radiol.223312 · Cited by 571
Practical tips for surgical research: Research questions, hypotheses and objectives.
Farrugia, P., Petrisor, B. A., Farrokhyar, F., & Bhandari, M. (2010) · PubMed · Cited by 347
The future of medical education and research: Is ChatGPT a blessing or blight in disguise?
Arif, T. B., Munaf, U., & Ul-Haque, I. (2023) · Medical Education Online · doi:10.1080/10872981.2023.2181052 · Cited by 317
High Rates of Fabricated and Inaccurate References in ChatGPT-Generated Medical Content
Bhattacharyya, M., Miller, V. M., Bhattacharyya, D., & Miller, L. E. (2023) · Cureus · doi:10.7759/cureus.39238 · Cited by 294
Informed consent process: A step further towards making it meaningful!
Kadam, R. A. (2017) · Perspectives in Clinical Research · doi:10.4103/picr.picr_147_16 · Cited by 268
Cited by
Kurtz, G., Amzalag, M., Shaked, N., Zaguri, Y., Kohen-Vacs, D., Gal, E., Zailer, G., & Barak-Medina, E. (2024) · Education Sciences · doi:10.3390/educsci14050503 · Cited by 181
“…knowledge and learning processes on diverse subjects [36,37], promote personalized learning experiences [38], and facilitate the writing process [39,40], thus reducing the pressure of writing papers [39] and cognitive load [36] while giving learners autonomy that will increase their sense of…”
BackgroundAlmarashdi, H. S., Jarrah, A. M., Khurma, O. A., & Gningue, S. M. (2024) · Eurasia Journal of Mathematics Science and Technology Education · doi:10.29333/ejmste/15739 · Cited by 79
Aydın, Ö., Karaarslan, E., Erenay, F. S., & Bačanin, N. (2026) · Turkish Journal of Engineering · doi:10.31127/tuje.1838765 · Cited by 74
Artificial intelligence in applied (linguistics): a content analysis and future prospects
Alaqlobi, O., Alduais, A. M. S., Qasem, F. A. A., & Alasmari, M. (2024) · Cogent Arts and Humanities · doi:10.1080/23311983.2024.2382422 · Cited by 64
A thorough examination of ChatGPT-3.5 potential applications in medical writing: A preliminary study
Dergaa, I., Saad, H. B., Glenn, J. M., Aissa, M. B., Taheri, M., Swed, S., Guelmami, N., & Chamari, K. (2024) · Medicine · doi:10.1097/md.0000000000039757 · Cited by 60
Metadata
- Authors
- Mondal, H., & Mondal, S.
- Affiliations
- All India Institute of Medical Sciences, Deoghar (IN)
- Department of Physiology, Raiganj Government Medical College and Hospital, West Bengal, India
- Corresponding author
- Mondal, Himel
- Mondal, Shaikat
- Venue
- Indian Journal of Ophthalmology
- Published
- 2023-11-20
- Volume
- 71
- Issue
- 12
- Pages
- 3600-3606
- ISSN
- 0301-4738 (ISSN-L)
- 1998-3689
- PMC
- PMC10788737
- OpenAlex
- W4388832642
- Cited by
- 80
- References
- 34
- Citations by year
- 2026: 13 · 2025: 34 · 2024: 32 · 2023: 1
- FWCI
- 13.66
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- cc-by-nc-sa
- Locations
- Indian Journal of Ophthalmology (publishedVersion, cc-by-nc-sa)
- PubMed Central (publishedVersion, cc-by-nc-sa)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Social Sciences › Library and Information Sciences › Academic Writing and Publishing
- Social Sciences › Education › Artificial Intelligence in Education
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Keywords
- ChatGPT, academic writing, large language models, AI-assisted writing, research proposal writing, manuscript preparation, AI reliability, responsible AI use
- MeSH terms
- Humans, Language, Reproducibility of Results
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: ChatGPT in academic writing: Maximizing its benefits and minimizing the risks. https://app.cochat.ai/papers/10.4103/ijo.ijo_718_23
- The paper
- Mondal, H., & Mondal, S. (2023). ChatGPT in academic writing: Maximizing its benefits and minimizing the risks. Indian Journal of Ophthalmology. https://doi.org/10.4103/ijo.ijo_718_23
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.