Research proposal for PhD admission

In the university admission notices we reviewed, mostly for 2025-26, length limits ran from a 1,000-word summary to 3,000 words or more, and most asked for the same parts: a working title, background, research questions or objectives, a literature review ending in a gap, the methodology, expected outcomes, a timeline and references. The interview panel reads it before it meets you, so write a proposal you can defend line by line.

What is a research proposal for PhD admission?

The UGC’s PhD Regulations, 2022 define a research proposal as “a brief write-up giving an outline of the proposed research work” that you submit with your application for registration. In practice, departments use it to screen applicants and to steer the interview. Where a university runs its own entrance test, the regulations give the test 70% of the weight and the interview 30%; JNU’s 2025-26 prospectus applies the same 70:30 split to NET percentiles and the viva-voce, while JRF holders are selected on the viva-voce alone.

After admission, the regulations make your Research Advisory Committee responsible for reviewing the proposal and finalising the topic. Many universities call that fuller, post-admission document the synopsis. This guide covers the admission-stage proposal; for the registration document, see our PhD synopsis format guide.

What format and word limit do universities ask for?

Each department sets its own rules, and they differ a lot. These are the proposal rules for eight programmes at four institutions, taken from their own admission documents.

Summarised on 2026-09-29 from each institution’s own document, linked in the first column. The Delhi, Jamia and JNU documents are for 2025-26; the TISS rule is from its current admissions page, which does not state a year. We are not affiliated with any of them. Rules change every cycle, so read your department’s current notice.

Across these notices the same core keeps appearing. This table lists each part, what it must answer and which of the notices above name it; your department’s own template always comes first.

DU Anthropology also asks for a hypothesis, the universe of study and a conceptual or theoretical framework. If a notice gives no word split, spend most of your words on the literature review and the method.

How do you write a research proposal for PhD, step by step?

1. Read the notice and the faculty pages

   Note the word limit, the sections, and whether the proposal is uploaded or brought to the interview. Read recent papers by two or three faculty members whose work is close to yours.

2. Choose a question, not an area

   “AI in healthcare” is an area. “Can a chest X-ray model tell when to refer a case to a radiologist?” is a question. Our PhD topics in computer science are written in this form.

3. Read 15 to 25 recent papers and find the gap

   Start from one or two surveys, put the papers in a small table (method, data, result, limitation), and write the gap in two sentences that point to that evidence.

4. Plan the method question by question

   Name the data you can access, the method, the baseline you will compare against and the metric that decides success.

5. Add outcomes, a timeline and references, then cut to the limit

   A shorter proposal that fits the limit beats a longer one that ignores it. Cite every source and run a similarity check.

What does a sample PhD research proposal outline look like?

• Working title: Faithful cross-lingual question answering over Indian public documents with retrieval-augmented generation.
• Background: People ask about public schemes in Hindi or Marathi, while many source documents are in English. Language models answer fluently but can state things the documents do not support.
• Research questions: (1) How much accuracy does cross-lingual retrieval lose against same-language retrieval for Hindi and Marathi questions? (2) Which grounding methods, such as citation-constrained generation or answer verification, cut unsupported claims without cutting useful answers? (3) Can faithfulness be scored automatically in these languages in a way that agrees with human raters?
• Literature and gap: three strands, compared in one table: retrieval-augmented generation, multilingual language models, and hallucination detection. The gap, as the review might find it: few studies measure faithfulness when the question and the documents are in different Indian languages.
• Methodology: a bilingual test set of questions over public documents, each answer checked by two annotators; open 7⁠–⁠8B models with a multilingual retriever; same-language and English-only retrieval as baselines; answer accuracy, citation precision, unsupported-claim rate and agreement with human ratings as metrics.
• Expected outcomes: an evaluation set released if licences allow, a grounded answering method with measured faithfulness, and two or three papers.

References the outline would start from, in IEEE style:

1. Y. Huang and J. Huang, “A survey on retrieval-augmented text generation for large language models,” ACM Comput. Surv., vol. 58, 2026, doi:10.1145/3805774.
2. L. Qin et al., “A survey of multilingual large language models,” Patterns, vol. 6, 2025, Art. no. 101118, doi:10.1016/j.patter.2024.101118.
3. L. Huang et al., “A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions,” ACM Trans. Inf. Syst., vol. 43, 2025, doi:10.1145/3703155.

How is a research proposal different from a PhD synopsis?

Some universities use the two words for the same document, so check your ordinance. Where they differ, this is how.

How do you present a research proposal at a PhD interview?

Plan about ten slides for a ten-minute talk, one idea per slide, then cut to whatever limit the panel sets.

1. Title: the working title and the research area.
2. The problem: why it matters now, with one fact or example.
3. What is known: three or four key papers, grouped by approach.
4. The gap: one sentence, with the evidence.
5. Research questions or objectives.
6. Method: data, approach and how you will measure success.
7. Framework figure of the proposed approach.
8. Timeline for three years.
9. Expected outcomes and planned papers.
10. Why you, why here: your preparation and how the topic fits the department.

Limits can be tight. Delhi University’s Computer Science department allowed at most five slides and four minutes for its 2025-26 Phase II interviews. For a five-slide version, merge slides 2 and 3, 4 and 5, 6 and 7, and 8 to 10, and keep the title on the first. Bring printed copies if the notice asks: Jamia asked for 10 and JNU’s Centre for Historical Studies for six.

Expect questions such as: Why this problem? What if your main idea does not work? Where will the data come from? What is the smallest result that would still be worth a paper? Why this department?

Why are PhD research proposals marked down?

Interview boards say what they look for. Delhi University’s Faculty of Education marks the understanding of theory, awareness of methodological alternatives, relevant field and policy developments, and the overall quality of the proposal. Its Anthropology department assesses theoretical understanding, methodological rigour, conceptual clarity and research aptitude. Proposals usually lose marks for the opposite:

• An area instead of a question, so nobody can tell what you will actually find out.
• A gap nobody can check, such as “little work has been done”, with no papers cited.
• A method with no data, baseline or metric.
• No fit with the department: a topic none of the faculty works near.
• Ignoring the notice: over the word limit, or missing a required section.
• Copied or uncited text.
• A plan too big for three years, or one you cannot explain when questioned.

What should you check before you submit?

• Right document, right format: the department’s own sections, word limit and file or print format.
• A title that names the problem, the approach and the setting.
• Two to four research questions, each one testable.
• A gap backed by recent, peer-reviewed papers, all cited in one style.
• A method per question: data you can access, baseline and metric.
• A timeline that fits three years, including coursework.
• A fit with at least one faculty member’s research.
• A similarity check done, and every sentence one you can explain at the interview.

Once admitted, the proposal grows into the synopsis; our PhD guidance covers that stage, from the literature review to the committee presentation.

Frequently asked questions