How to Use AI for a Research Proposal, Section by Section
By Samuel de Bruyn 11 min read
Use AI on a research proposal for the reading work: finding papers, summarising and comparing them, and checking your claims against the passages you cite. Keep the research gap, the research questions, the argument and the wording your own, because several universities and funders do not accept AI-written proposal text as your original work, and many ask you to declare substantive AI use.
What “using AI for a research proposal” means: using an AI tool to find, read, compare or check the sources behind a research proposal, while the student decides the gap, frames the questions and writes the text. Having AI generate the proposal text is a different use, and many universities restrict or ban it.
What can AI do on each section of a research proposal?
AI helps most on the research proposal sections that depend on reading (background, literature review, gap) and is least acceptable on the sections that carry your judgement (research questions, contribution, final wording). The sections follow how to write a research proposal.
| Proposal section | What AI can do with your papers | What stays yours | Where the line usually falls |
|---|---|---|---|
| Title and abstract | List the key terms your papers use for the topic | The title, the proposal’s central claim, the abstract’s wording | Checking terms is fine; generated wording is grey |
| Introduction and background | Summarise what your papers report about the problem, with citations | Which facts frame the problem, and the paragraphs themselves | Reading help is fine; ask before using generated text |
| Literature review | Line studies up by population, method and finding, and flag where they conflict | The themes, the synthesis and the argument | Comparing is fine; the synthesis is your writing |
| Research gap and rationale | Show which populations, settings or methods none of your papers cover | Which gap is real (after a wider search) and why it matters | Finding candidates is fine; the gap claim is yours to verify |
| Research questions and objectives | Check whether a candidate question is already answered in your papers | The questions themselves | Generated questions are grey at best; ask your supervisor |
| Methodology | Pull out the methods and samples that comparable studies used | The design, and the justification for it | Reading methods is fine; the justification is yours |
| References | Format citations in your department’s style and export them | Checking that each reference exists and says what you cite it for | Fine, if every reference is checked |
The last column is a default. The binding rule is whatever your university, department, supervisor or funder sets for that proposal.
Can you use AI to write a research proposal?
Many universities allow some AI help with a research proposal (UCL’s “assistive role” category is one example), but many also restrict or ban AI-written proposal text, including in PhD applications. The University of Warwick’s Department of Sociology (UK) says it expects “the research proposal and application materials to be your own original writing” and states: “We strongly discourage students from getting assistance from AI in the development of their research proposal” (Warwick Sociology, research proposal guidance).
Warwick Sociology applicants who used any AI should say so, in what ways and which aspects of the work it relates to, and give full references to the AI sources used, in a statement that must be “clearly located on the front page of their proposal”.
Funders draw the line more sharply: the US National Institutes of Health (NIH) does not treat applications “substantially developed by AI” as the applicants’ original ideas (NIH notice NOT-OD-25-132, 17 July 2025).
Many universities do not allow AI-written text in submitted work even after you edit it. The defensible default for a research proposal: use AI to read, compare and check; write the words yourself; declare what you used.
What do universities say about AI in a research proposal?
Universities set AI rules per institution and often per programme or assessment, so the rule for your research proposal is the one your department publishes. These examples were read in November 2026.
| University (country) | Published rule | What it means for a research proposal |
|---|---|---|
| University of Warwick, Sociology (UK) | Proposal must be “your own original writing”; AI assistance is strongly discouraged; any AI use is declared on the front page | The text is yours, and any AI use is declared |
| UCL (UK) | Staff guidance (“not a formal policy”) places each assessment in one of three categories: “GenAI tools cannot be used”, “GenAI tools can be used in an assistive role” or “GenAI has an integral role”; “Students must acknowledge GenAI use, but do not need to cite it” | Find your proposal’s category before using any AI tool; the person setting the assessment makes the final decision |
| University of Georgia Graduate School (US) | Generative AI in theses and dissertations is “prohibited unless specifically authorized by members of the advisory committee for use within the approved scope”; approved use “should be disclosed in a statement within the thesis or dissertation” | The policy does not name proposals, so ask your committee whether it covers your prospectus (in the US a “thesis” is usually a master’s, a “dissertation” a PhD) |
| Stellenbosch University (South Africa) | “All postgraduate students at Stellenbosch University are required to declare any use of Artificial Intelligence (AI) tools”; the declaration discussion should ideally happen “during the proposal preparation” | Agree AI use with your supervisor at proposal stage; you both sign an internal form confirming that discussion |
Sources: UCL; University of Georgia (18 January 2024); Stellenbosch University.
What do research funders say about AI in grant proposals?
NIH, NSF and UKRI each set rules on AI in research proposals: NIH does not treat AI-developed applications as original, NSF encourages and UKRI expects applicants to be open about AI use, and both NSF and UKRI restrict reviewers from putting proposals into AI tools. A PhD student applying for a studentship or first grant meets these rules directly.
NIH (US). Since the 25 September 2025 receipt date, NIH does not treat applications, or sections, “substantially developed by AI” as the applicants’ original ideas. AI use detected after an award can be referred to the Office of Research Integrity and can cost the grant (NOT-OD-25-132).
NSF (US). The National Science Foundation’s notice of 14 December 2023 encourages proposers to state in the project description the extent of any generative AI use, and says “Proposers are responsible for the accuracy and authenticity of their proposal submission”. NSF reviewers “are prohibited from uploading any content from proposals” into generative AI tools that NSF has not approved (NSF notice on generative AI in merit review).
UKRI (UK). Under UK Research and Innovation’s policy of 23 September 2024, applicants are “expected to be transparent where they have used generative AI tools in the development of an application”. Its list of “Examples of substantive use that could be declared” includes “comparing literature sources” and “interpreting, analysing or summarising data (including audio, image files) and text”. Minimal use, such as “improving the standard of English” or “formatting or reducing word count”, need not be disclosed, and the declaration “will not affect the assessment process” (UKRI policy).
How do you check AI output before it goes into your research proposal?
Check every AI summary, comparison and citation against the source passage before it goes into a research proposal. Duke University Libraries’ guide to AI in literature reviews puts it plainly: “Generative AI tools can produce inaccurate citations and summaries”, and users must “verify all information independently” (Duke University Libraries, updated 28 September 2026).
Run three checks on every AI output before it enters a research proposal:
- The passage check. Open the cited page and read the sentence; if it does not say what the summary says, drop the claim.
- The reference check. Confirm each reference exists and has the right authors, year and DOI, using the free reference checker and the steps in how to spot fake AI citations before you submit.
- The coverage check. Search beyond your reading list before you call something a research gap, as how to use AI to find a research gap explains.
How do I use an AI assistant for writing backed by sources on my research proposal?
An AI assistant for writing backed by sources runs the reading side of a research proposal on your own papers: add them, ask them questions, compare them in a cited table and check your paragraphs against them.
Litrevu is an AI assistant for writing backed by sources. It reads every paper, report and PDF you add, answers your questions with a citation to the exact passage behind each claim, compares sources side by side, and checks your own draft against them.
Here is the proposal workflow in Litrevu, step by step as on its how it works page.
- Add the papers behind your research proposal. Upload PDFs, paste a PDF link, a DOI, an arXiv link or a paper’s text, or let Litrevu find open-access papers on your topic and pick the ones you want.
- Ask the papers your proposal questions. “Which populations have these studies sampled?” “Which methods did they use?” “Where do their findings disagree?” Answers cite only your papers, and Litrevu says so when your papers do not cover a question.
- Open every citation. Each citation opens its passage highlighted on the PDF page it came from, so you read the evidence before your proposal relies on it.
- Compare the key studies in a table. Line the studies up side by side, with citations, and see where they agree and where they differ. Rows where findings conflict, or where no paper covers a population, are candidates for your gap and rationale.
- Check the paragraphs you wrote. Paste your problem statement, gap paragraph or rationale. Each claim comes back Supported, Partly supported, Contradicted or Not in your papers.
- Cite and export. Switch between APA, MLA, Chicago, IEEE, Harvard and Vancouver, and export references to BibTeX or RIS.
A claim check on a gap sentence looks like this (a format illustration, not real output):
“No study has examined this intervention in rural schools (Author, Year; Author, Year).” Not in your papers
“Not in your papers” means your library does not cover that claim. It is a cue to search wider, not proof that a gap exists in the literature. Litrevu finds and lines up the evidence; you decide which gap is real, frame the questions and write the proposal.
Which AI tools help with which part of a research proposal?
AI tools for a research proposal fall into two groups: discovery tools that search the published literature for papers, and tools that read, compare and check the papers you have chosen. A proposal needs both.
| Tool | What the vendor says it does | Where it fits in a research proposal |
|---|---|---|
| Elicit | Search across a corpus of more than 138 million papers, structured data extraction into tables, and a systematic review workflow that screens thousands of papers; it also works with PDFs you upload (Elicit help centre) | Finding papers at scale, extracting data into tables, and screening for a systematic-review proposal |
| Consensus | An AI-powered academic search engine over peer-reviewed papers; its Consensus Meter shows the share of studies supporting or contradicting a yes/no claim (consensus.app) | Quick evidence answers to a specific question while you scope the topic |
| ResearchRabbit | Paper discovery through citation graphs: start from one paper and expand through related works, authors and topics, with visualisations of how topics relate (researchrabbit.ai) | Mapping how the field’s papers connect and spotting related work your search missed |
| Litrevu | Answers questions about the papers in your library with cited passages, compares papers in a cited table and checks your paragraphs against them (how it works) | Reading and comparing your chosen papers, finding where they conflict and what they do not cover, and checking your gap and rationale |
Elicit, Consensus and ResearchRabbit are stronger than Litrevu at discovery across the published literature, and Elicit is built for systematic-review screening at scale. Litrevu’s part is reading and checking once the papers are chosen; for a large pile, see how to read research papers quickly for a research proposal.
How do you declare AI use in a research proposal?
Declare AI use in a research proposal with a short statement naming the tool, what you used it for and which parts it touched, placed where your university or funder asks. Warwick Sociology wants the statement on the proposal’s front page; NSF encourages it in the project description; at Stellenbosch, student and supervisor sign an internal form confirming they discussed AI use, and a separate examination declaration goes with the thesis.
A disclosure statement for a research proposal might read like this (a format illustration; adapt it to your actual use):
I used [tool name] to summarise and compare the papers on my reading list and to check the claims in my own draft against those papers. I did not use it to write the text of this proposal. I checked every citation against its source.
The free AI disclosure statement generator builds the wording from your answers, and how to write an AI-use disclosure statement covers placement and referencing. For where the line sits, see the ethics of AI in research writing.
Frequently asked questions
Can I use ChatGPT for my research proposal?
Whether you can use ChatGPT for a research proposal depends on your university, department and assessment, not on the tool. UCL's guidance, for example, places each assessment in one of three categories, from "GenAI tools cannot be used" to "GenAI has an integral role". Where AI is allowed, use it to read and check sources, write the proposal yourself, and declare the use.
Do I have to declare AI use in a PhD proposal?
Many universities and funders expect you to declare substantive AI use in a PhD proposal. Stellenbosch University requires all postgraduate students to declare any use of AI tools, Warwick Sociology wants the statement on the proposal's front page, and UKRI expects applicants to be transparent about AI use and lists comparing literature sources among the examples of substantive use that could be declared.
Can I use AI to improve the English in my research proposal?
Many institutions treat language polishing as minimal AI use. UKRI, for example, says applicants need not disclose minimal uses such as "improving the standard of English" or "formatting or reducing word count". A university may set a stricter rule for an assessed proposal, so check your department's policy before you rely on the UKRI line.
Can my research proposal be rejected for using AI?
A research proposal can be rejected or investigated if its AI use breaks the rules that apply to it. NIH may refer AI use detected after an award to the Office of Research Integrity, and the University of Georgia counts generative AI in theses and dissertations as "unauthorized assistance per the Academic Code of Honesty" unless the advisory committee approves it; ask your committee whether that rule reaches your prospectus. Declared AI use within the rules is a different matter.
What is an AI research proposal generator?
An AI research proposal generator is a tool that produces the text of a proposal from a prompt. Generated text is the use that Warwick Sociology discourages and NIH will not treat as original. For an academic proposal, use AI to read, compare and check your sources, and keep the writing yours.
Where should you start with AI on your research proposal?
Start with your department’s AI policy and the papers you already have. Put the reading list into a tool that answers only from those papers and shows the passage behind every answer, then write the gap and rationale yourself. Add your papers and ask them your first proposal question. Free to start, no credit card required.