Is it acceptable to get ChatGPT’s help with a research paper? And what of a grammar checker, or an AI tool that locates papers? What does my journal permit — and what might get my paper rejected or retracted? Every researcher, PhD scholar and academic author is asking these questions in 2026.
AI tools now play a role throughout research, from searching the literature to drafting and answering reviewers. Most journals do not ban them. Even so, what is expected on disclosure, authorship and acceptable use has become clearer and stricter, and errors can bring rejection, retraction or lasting damage to your reputation. Below we explain what is generally permitted and what is not, how to disclose AI use, and the mistakes that cause the most trouble.
Why This Is More Important Than Ever
Rejections, corrections and retractions are increasingly traced to careless or undisclosed AI use. Fabricated references — citations with believable titles, authors and even DOIs for papers that do not exist — are among the simplest problems to spot, since reviewers and editors can just follow the links. Odd phrasing, made-up sources and text that fails to match the data all prompt questions during screening and review.
Meanwhile, appropriate use of AI is widely accepted. What separates acceptable from unacceptable is not the tool but whether AI supports your thinking and communication or replaces it — and whether the use is disclosed openly, specifically and in the proper place.
What Is Permitted and What Is Not
| AI use case | Generally allowed? | Why |
|---|---|---|
| Grammar checks and language polishing | Yes, widely accepted | Making text easier to read is not a claim of contribution; the content is still yours |
| Literature search assistance | Yes, with verification | Locating papers is a research aid; reading and judging them remains your job |
| Brainstorming research questions | Yes, with transparency | Ideas serve as a starting point; the design and execution have to be your own |
| Summarising a paper you have already read | Yes, as personal notes | Acceptable for your own understanding; passing off an AI summary as your analysis is not |
| Translating text you wrote | Yes, if declared and checked | The translation needs your verification, technical terms above all |
| Producing whole sections of your paper | No | Passing off AI-generated text as your own writing violates authorship standards |
| AI named as an author or co-author | No | Authorship demands human accountability, something AI cannot carry |
| Using AI in a manuscript without disclosure | No | Journals expect AI use to be declared, and concealing it amounts to a false declaration |
| Generating citations with AI | No, unless each one has been verified | Made-up references count as serious misconduct |
| Undisclosed AI in data analysis | No | Any role AI plays in analysis should be stated in the Methods section |
What Publishers and Ethics Bodies Expect
Wording differs from publisher to publisher, but their core stance is remarkably uniform. Both the Committee on Publication Ethics (COPE) and the International Committee of Medical Journal Editors (ICMJE) say that AI tools cannot be credited as authors, that using them should be disclosed, and that responsibility for the content stays entirely with the human authors. Leading international publishers have released policies of their own built on these same principles. Reading a few publisher policies side by side quickly shows how much they have in common.
| Principle | What this means for you |
|---|---|
| AI cannot hold authorship | Only individuals meeting the authorship criteria are named |
| Use has to be disclosed | Name the tool and its purpose in the place the journal specifies — often Methods, Acknowledgements or a dedicated statement |
| Humans are responsible | You are answerable for each sentence, number and citation |
| Simple proofreading is handled leniently | Numerous policies do not ask you to declare spelling and grammar checkers, though some do — check the journal’s rule |
| AI-generated images and data face restrictions | Most policies prohibit AI-created figures or results passed off as real |
| Reviewers are bound to keep confidentiality | Manuscripts should never be uploaded to public AI tools |
The ICMJE Recommendations are especially influential for medical and health papers, asking authors to mention writing assistance in the acknowledgements and to report AI used for data collection, analysis or figures in the methods. COPE supplies the wider ethical framework for many journals.
A Short Summary of JMVRA’s AI Policy
Generative AI is not prohibited by JMVRA. The requirement is that its use be declared and that a person stays answerable for each sentence, number and citation in the published paper. The complete text sits on the JMVRA AI policy page, and the key points follow.
| Topic | JMVRA position |
|---|---|
| Acceptable uses | Grammar, spelling, readability and flow in text you wrote yourself; translation of your own writing; reformatting of references; wording suggestions that you go on to verify; coding help for analysis you designed; summaries of papers you have already read |
| Not acceptable | Naming AI as an author; AI-generated data, images or figures passed off as real; a literature review built on sources you have not read; AI citations left unverified; an entire manuscript written by AI; rewording copied text to dodge similarity screening |
| Where to declare | A brief statement ahead of the reference list giving the tool, its version if known, and the purpose it served |
| If no AI was used | No statement is required; for that reason, undeclared use is regarded as a false declaration |
| Reviewers | Must not paste any portion of a manuscript into a public AI tool; the evaluation has to be their own work |
| Editorial decisions | Taken by people; an editor interprets similarity and AI-detection scores rather than applying them as automatic thresholds |
See also: Publication Ethics · Authorship Policy · Plagiarism Policy · Peer Review Policy.
AI Across Every Stage of Research
| Research stage | Acceptable AI use | What has to stay human |
|---|---|---|
| Choosing a topic and drafting a proposal | Probing for gaps, locating related work, brainstorming questions | The final question, the original contribution and critical judgement about the gap |
| Literature review | Locating papers, organising references, suggesting keywords | Reading and assessing each paper, synthesising arguments, pinpointing the gap |
| Research design | Surveying method options, getting to grips with statistical approaches | Selecting and defending the method, designing instruments, making ethical decisions |
| Data collection | Transcription, help with translation, initial coding suggestions | Verifying accuracy, ethical conduct, interaction with participants, interpretation |
| Data analysis | Software with built-in AI features, early pattern detection, coding assistance | Reading the results, reaching conclusions, explaining findings |
| Writing the manuscript | Grammar, sentence recasting, language polishing | Each substantive idea, argument, finding and conclusion |
| References and citations | Formatting citations, supplying missing metadata | Confirming that each reference exists and backs the claim |
| Responding to reviewers | Making sense of complex comments, polishing the response letter | Your own replies, methodological defence and revision decisions |
The principle holds throughout: AI is a tool that supports your work. What counts as intellectual substance — your thinking, judgement, conclusions and the writing that forms the contribution — has to come from you.
Everyday AI Tools and What They Can Acceptably Be Used For
Large language models (such as ChatGPT, Claude and Gemini)
Conversational language models can legitimately help you sharpen the clarity of text you wrote, brainstorm directions or counter-arguments, draft a plain-language summary for non-specialists, clarify a hard concept for your own understanding, and propose a better paragraph order. They ought not to write a paper’s substantive sections, generate your reference list, write your literature review or discussion, or take over your own critical analysis. Any use while preparing a manuscript should be declared in the way your journal specifies. Keeping a note of your prompts and the tool’s output lets you describe the use accurately later.
Grammar and language-editing software
Of all the kinds of AI help, grammar and language checkers enjoy the widest acceptance. A number of policies regard basic spelling and grammar correction as equivalent to a spell-checker and need no statement, whereas other journals want every AI tool declared. Paraphrasing tools call for extra caution: applied to text you wrote, they offer language support, yet applied to another person’s writing they become a means of hiding copying, which counts as misconduct.
Literature search tools powered by AI
Tools that surface related papers or chart citation networks are fine for discovering sources. What matters is that you yourself read, assess and synthesise each paper you cite. It is unacceptable to cite a paper you have not read, or to pass off an AI summary as your own analysis. Our Literature Reviews That Synthesise, Not List shows how to assemble a review from sources you have really read. Keeping a short note on each source as you read makes this much easier.
AI for data analysis and statistics
Statistical packages with AI features, and AI assistance with analysis code, are acceptable provided the Methods section describes them as it would any other software and you have validated what they produce. Deciding what the results mean remains your job. Whenever AI was involved in processing data, state this precisely in Methods rather than relying on a general thank-you note. Saving your scripts and settings also lets others reproduce the work.
Prompts That Are Responsible and Risky
How you phrase a request makes a difference. Asking for help with your own material is normally fine, whereas asking the tool to provide content, sources or results is where problems begin.
| Responsible prompt | Risky prompt |
|---|---|
| “Improve the grammar of this paragraph I wrote, without changing its meaning.” | “Write the discussion section for my paper on this topic.” |
| “Explain the difference between these two statistical tests.” | “Give me results that show a significant difference.” |
| “Suggest keywords I could use to search for papers on this topic.” | “Give me 30 references on this topic with DOIs.” |
| “Check whether my abstract matches my conclusions.” | “Rewrite this published paragraph so it passes plagiarism check.” |
| “Translate my methods section into English.” | “Create a figure showing these lab results.” (for data you do not have) |
Writing Your AI Disclosure Statement
Where a journal wants the statement varies: Methods, Acknowledgements, a separate declaration, the cover letter, or several of these. Treat the templates below as a starting point and tailor them to the journal you are targeting. At JMVRA, place the statement immediately before the reference list.
| Where to disclose | Template statement |
|---|---|
| In the manuscript (Methods, Acknowledgements or a declaration before the references) | “The authors used [tool name and version] to [specific use, e.g. improve the grammar and readability of the manuscript]. All content was written, reviewed and verified by the authors, who take full responsibility for it.” |
| For code assistance | “[Tool name] was used to assist with writing the analysis scripts. The study design, data and interpretation are entirely the authors’ own, and all output was independently validated.” |
| In the cover letter | “This manuscript was prepared with the assistance of [tool name] for [specific purpose only]. This use is disclosed in the [section]. All research design, data collection, analysis and conclusions are the work of the authors.” |
| In a data or ethics statement | “No AI tool was used to generate, fabricate or interpret the research data. All authors reviewed and are accountable for the final content.” |
Principles to follow when disclosing:
- Be specific — give the tool’s name and spell out precisely what you used it for.
- Be honest — never play down how much you used AI; a use that goes undeclared and is found later does much greater harm.
- Confirm responsibility — say that the authors checked all of it and accept full responsibility.
- Avoid vague wording — a line such as “AI was used to assist in writing this paper” fails to say at which stage or to what extent.
Disclosures: Strong Versus Weak
| Weak disclosure | Strong disclosure |
|---|---|
| “AI tools were used.” | “[Tool, version] was used to correct grammar in the Introduction and Discussion.” |
| “ChatGPT helped with writing.” | “[Tool] suggested alternative phrasings for two paragraphs, which the authors rewrote and verified.” |
| Nothing, although a translation tool was used | “The Methods section was translated from Hindi using [tool] and checked line by line by the authors.” |
| “AI was used for analysis.” | “[Tool] assisted in writing R scripts for the regression analysis; outputs were checked against manual calculations.” |
The 5 AI Errors That Lead to Retractions
| Misuse of AI | Reason it results in rejection or retraction | How to avoid it |
|---|---|---|
| Fake references generated by AI | Reviewers and editors can easily spot citations of papers that do not exist | Do not accept any reference from AI output until you have opened and checked the source |
| AI-written sections that go undeclared | Odd phrasing, made-up facts and inconsistencies prompt questions, and concealing the use amounts to a false declaration | Declare every use of AI, and rework assisted text so it carries your own ideas and voice |
| AI named as an author | Authorship carries accountability, and AI cannot be held accountable | Name only people who satisfy the authorship criteria |
| Images or data made by AI and passed off as genuine | That is fabrication, and it is uncovered by image screening and scrutiny after publication | Do not use AI to produce experimental data, figures or results |
| AI-assisted submissions that are recycled | Near-identical AI-assisted papers sent to multiple journals may be flagged as duplicate publication | Every paper has to be original; send it to only one journal at once |
Why AI Makes Up References and How to Check Them
A general-purpose language model produces text that is probable given your prompt; unless it is linked to a search tool and you check its sources, it does not look anything up. When you ask for references, it may output citations that seem perfectly correct, with plausible authors, a genuine journal, a convincing title and page range, for papers nobody ever wrote. Even where a tool names a real paper, that paper may not contain what the tool says it does. Treat every AI-supplied citation as unconfirmed until you have seen the source.
- Compile the bibliography only from works you have personally read, importing each one into a reference manager such as Zotero or Mendeley using the record held by the publisher or database.
- Click each DOI. A citation is unusable if the DOI does not lead to the paper you cite.
- Confirm that the source backs up your sentence. Citing a genuine paper for a finding it does not report is still a false citation.
- Make the text and the list agree — each in-text citation should appear in the list, and each listed item should be cited in the text.
Our DOI guide describes how DOIs resolve, while the Scopus search guide explains how to verify that a paper exists using its title or DOI.
Paraphrasing Tools and Tortured Phrases
A “tortured phrase” is an odd stand-in for a standard technical term, such as “counterfeit consciousness” in place of “artificial intelligence”, or “bosom peril” in place of “breast cancer”. Integrity researchers discovered that these phrases usually show up when copied text has been run through automatic paraphrasing (“spinning”) tools to conceal where it came from. Many examples have been catalogued, and screening tools now use them to mark suspicious papers. Because they are so noticeable, such phrases undermine an editor’s trust in the whole manuscript.
Whether they arrive by accident or on purpose, these phrases will attract an editor’s notice. Go through the entire manuscript before you submit, and check that each technical term is the standard one in your discipline. A paraphrasing tool should never be applied to text you did not write.
AI Detectors: What They Can and Cannot Show
Certain journals and institutions rely on AI-writing detectors. Such tools estimate how likely it is that text was produced by a machine, and that is not proof. They may flag writing done by people, particularly authors whose first language is not English, and they can overlook AI-assisted text that was edited heavily.
| What a detector can do | What it cannot do |
|---|---|
| Highlight passages worth a closer look | Prove that AI wrote a passage |
| Prompt an editor to ask questions | Replace an editor’s reading of the manuscript |
| Encourage authors to declare their AI use | Distinguish declared, acceptable assistance from misuse |
At JMVRA, an editor reads similarity and AI-detection scores together with the manuscript, and they are never used as an automatic cut-off. Should you wish to check your own manuscript before submitting, JMVRA’s similarity check service provides plagiarism reports, and AI writing detection can be requested. Even so, the strongest safeguard is a simple one: say what you used.
Is It Possible to List AI as an Author?
No. Journals, publishers and ethics bodies, among them COPE and the ICMJE, agree on this more clearly than on almost anything else, and there are three reasons:
- Accountability: authors have to answer for the work, whereas an AI tool carries no legal or professional responsibility.
- Consent and communication: authors must sign off on the final version and respond to questions about it, which an AI tool is unable to do.
- Defending the work: authors need to be able to answer reviewers, readers and criticism made after publication.
When AI assisted with your paper, you, the human authors, are still answerable for every word. Using AI does not lessen that responsibility; it adds the obligation to disclose and to verify. Refer to JMVRA’s authorship policy. Keep a record of how each tool was used so that you can answer any later question.
AI and Images, Figures and Data
Images and data demand the greatest care. Using AI to create or modify experimental images, microscopy, gels, scans or photographs and then presenting them as genuine is fabrication. Synthetic data passed off as collected data falls under the same rule.
Fabricated results
AI-generated data, images or figures presented as real observations.
Illustrations
Conceptual diagrams made with AI help, if the journal allows them, clearly labelled and declared.
Data that is synthetic by design
If a study knowingly relies on simulated or synthetic data, the methods section has to state that fully.
AI and Peer Review: Guidance for Reviewers
Reviewers have duties as well. A manuscript under review is confidential and unpublished. Pasting all or part of it into a public AI tool sends it to a third party and breaks that confidentiality. A JMVRA reviewer must not do so; AI may be used to tidy the wording of a report, but the evaluation itself has to come from the reviewer’s own judgement. People always make editorial decisions, whether to accept, revise or reject. See also our guide to open access peer review. Reviewers who are unsure about a tool should ask the editor before using it.
Indian Researchers and AI Use
AI policies at journals bind every author regardless of country: an Indian author sending work to an international journal is held to the same rules as everyone else. Indian journals differ, since some have published AI policies and others have not, so consult each journal’s author guidelines.
- Theses and dissertations: similarity in theses is governed by the UGC’s 2018 regulations on academic integrity and plagiarism, and numerous universities have added their own advice on AI tools. Look up your university’s present rules and consult your supervisor before using AI when writing a thesis.
- Appraisal and accreditation: a paper is valued according to the journal and the integrity of the work itself. Being retracted for undisclosed AI use or invented references can damage an appraisal file much more than publishing in a modest venue ever would.
- Language support: plenty of Indian researchers write in a second language. Turning to AI to edit or translate your own text is generally acceptable, as long as you declare it and check the outcome.
For similarity levels and thesis rules, see our plagiarism guidelines and plagiarism in research guides.
Safeguarding Your Paper Before Submission
- Go through every AI-assisted section yourself and make sure each claim, argument and conclusion expresses your own research.
- Verify each reference by hand — open every DOI and check that the title, authors and content match.
- Do a similarity check ahead of the journal, so you can resolve any issues first.
- Compose a disclosure that is specific and honest — naming the tool, its purpose and the humans’ full responsibility.
- Consult the target journal’s up-to-date AI policy on its author guidelines page.
- Make sure the journal suits your paper — a mismatch of scope is still a frequent cause of desk rejection. The JMVRA Journal Finder and our journal selection guide can help.
Submitting to JMVRA?
JMVRA is open to papers written with responsible AI help that has been declared. Put your AI statement ahead of the reference list, check every citation, and comply with the author guidelines.
- Review: transparent, double-blind peer review. Specialist reviewers assess your paper, and you receive the acceptance or rejection decision by email and WhatsApp.
- Cost: ₹971 INR (Indian authors) or $41 USD (foreign authors) is due after acceptance only; nothing is paid at submission.
- Access: licensed CC BY 4.0 for open reading, ISSN 2984-8911, and a Crossref DOI is available if asked for.
- Indexing, stated openly: JMVRA appears in Google Scholar and is registered with Crossref and the ISSN Portal / ROAD; it is not currently indexed in Scopus, Web of Science or DOAJ.
Ready to submit a carefully prepared paper?
Sending a paper to JMVRA is free of charge. Reviewers with subject expertise assess it, and an email and a WhatsApp message tell you whether it was accepted or rejected. Only after acceptance does the ₹971 INR (Indian) or $41 USD (foreign) fee apply.
Checklist for Your AI Use
- Human contributionThe ideas, analysis and conclusions are yours
- References verifiedEach DOI opened and each claim checked
- No fabricated materialNo data, images or results from AI
- Disclosure writtenTool, purpose, stage and responsibility all stated
- Right placeWherever the journal requires; at JMVRA, ahead of the references
- Policy readThe journal’s present AI policy has been followed
Everyday Safeguards for Using AI Tools in Research Writing
AI tools can speed up routine tasks, but responsibility for the content always stays with the human authors. A few everyday safeguards keep your use transparent and your paper trustworthy.
- Check the journal's policy first. Policies on AI assistance differ between journals and publishers. Read the author guidelines before you use any tool, and follow them when preparing both the manuscript and the declaration.
- Keep confidential material out of prompts. Do not paste unpublished data, participant information or manuscripts under review into public tools. Anything you enter may be stored or reused outside your control.
- Log what you used and why. Keep a simple note of the tool, its purpose and which parts of the text it touched. When a disclosure is requested, you can write it accurately without guessing.
- Verify every claim and citation. Generated text can sound confident while being wrong or invented. Confirm each statement, number and reference against the original source before it appears in your paper.
- Keep your own reasoning visible. Use tools for language support or brainstorming, but write the argument, interpretation and conclusions yourself. Your expertise is what readers and reviewers are really evaluating.
Follow the journal's rules, protect confidential material, disclose honestly, and check everything a tool produces before it carries your name.
Questions Researchers Often Ask
In Summary
AI tools are now a part of academic research in 2026. Researchers who use them well do so openly and specifically, as real aids to their own thinking. The ones who run into trouble use AI to generate content they present as theirs, to build reference lists they never verify, or to avoid the intellectual effort that peer review is there to assess.
The rules are simple: declare what you used and why, never allow AI to take the place of your own intellectual contribution, check everything it produces, and never name AI as an author. If you apply these four principles consistently, you can gain from AI without exposing your work or reputation to risk.
Ready to send in your paper?
You can submit to JMVRA online at no cost. Specialist reviewers examine your paper, with the acceptance or rejection decision reaching you by email and WhatsApp. The ₹971 INR (Indian authors) or $41 USD (foreign authors) fee is payable only after acceptance.
References and Sources
Summary of Responsible AI Use
- Assist, not replaceLanguage, translation, search, code help
- VerifyEvery reference, claim and output
- No fabricationNo AI data, images or results
- DeclareTool, purpose and responsibility
- Human authorsAI is never an author
Related: JMVRA AI policy · Plagiarism guidelines · Author guidelines.
Ready to Publish? Submit to JMVRA
JMVRA is aligned with UGC's Suggested Parameters for Choosing Peer-Reviewed Journals.
JMVRA is a Transparent, Peer-Reviewed, Open Access Journal (Refereed Journal). Following the discontinuation of the UGC and UGC-CARE journal listing, and in alignment with the latest UGC-CARE recommendations and guidelines, JMVRA is aligned with UGC’s Suggested Parameters for Choosing Peer-Reviewed Journals.
Further reading on jmvra.org
Key Terms
Subjects and search phrases this article addresses.
Related Reading
Further guides on research integrity, plagiarism and responsible publishing.
Plagiarism Guidelines
Read the guide RelatedResearch Paper Plagiarism: Its Types, How It Is Detected and Keeping Below the Threshold
Read the guide RelatedLiterature Reviews That Synthesise, Not List
Read the guide RelatedThe Full Predatory Journal Red-Flag Checklist
Read the guide RelatedOpen Access, Peer Review, Impact Factor & UGC Parameters Norms — A Complete Guide
Read the guide RelatedDOI Explained
Read the guideKeep Reading on the JMVRA Blog
This article sits on the JMVRA journal blog, where practical guidance covering publishing, indexing, metrics and research integrity is written to be verifiable, not promotional.