A bad research question wastes months. These 15 prompts walk a topic through the standard refinement moves used in graduate methods training — narrowing, variable mapping, PICO or PEO framing, feasibility checks — until you have a single FINER question you can defend in a committee meeting. Paste them into any frontier chatbot; the model-specific notes below tell you where each step pays off.
TL;DR
- FINER (Feasible, Interesting, Novel, Ethical, Relevant) is the audit checklist from Hulley et al.’s Designing Clinical Research; PICO structures quantitative/clinical questions, PEO structures qualitative ones.
- Use the prompts in order: narrow (1) → frame as PICO/PEO (2-3) → map variables (4) → audit on FINER (5) → check it is not already answered (11) → finalize one sentence (15).
- For the “already-answered” and operationalization steps, a deep-research tool with live citations (ChatGPT Deep Research on Plus at $20/mo, Gemini Deep Research, or Perplexity) beats a plain chat answer. For paper discovery, Elicit and Semantic Scholar are stronger than a general chatbot.
- AI generates candidates and audits feasibility; you choose, ground the question in literature, and defend it. Expect 3-5 revisions before fieldwork.
Who this is for
Honors thesis writers, MPhil and PhD candidates picking a project, capstone students, and clinicians drafting quality-improvement (QI) projects.
When not to use these prompts
Skip these when your question is already locked by a grant or supervisor — refining beyond your constraints wastes time. Skip them too if you have not yet skimmed at least 10 papers on the topic; the model cannot judge novelty against a literature you have not read.
Which framework fits your study
Pick the frame before you write the question, not after. The three below cover most thesis and clinical work.
| Framework | Best for | Components | Comparator? |
|---|---|---|---|
| FINER | Auditing any question (all fields) | Feasible, Interesting, Novel, Ethical, Relevant | n/a (it is a checklist, not a structure) |
| PICO | Quantitative / clinical, prospective intervention studies | Population, Intervention, Comparator, Outcome | Yes |
| PEO | Qualitative or observational, naturally occurring exposures | Population, Exposure/Experience, Outcome | No |
Rule of thumb: if a researcher introduces or tests the factor, use PICO; if the factor already exists in the world or your dataset, use PEO. FINER is not an alternative to those two — it is the quality gate you run a finished question through.
Prompt anatomy
A refinement prompt should carry six elements. Drop any one and the output drifts.
- Role: who the AI plays — research-methods professor, peer reviewer, thesis committee member, librarian.
- Context: your level, field, deadline, how many papers you have read, target citation style, course or program.
- Goal: one concrete deliverable — 5 candidate questions, a variable table, a 200-word specific-aims paragraph.
- Constraints: word count, depth, which study designs are off the table, what to never claim.
- Output format: numbered list, table, or graded blocks so you can paste straight into Notion, Word, or a proposal template.
- Signal: a reference question you like or an anti-example (
not a topic, an answerable question).
15 copy-ready prompt templates
Each template uses [bracketed placeholders] — replace them with your own text before sending. Swap any model name in the role line if you prefer a different tutor persona.
1. Topic to 5 narrower questions
First-pass narrowing; produces working candidates.
You are a research methods professor. My topic is "[topic]" in [field]. Generate 5 candidate research questions that narrow this topic, each more specific along a different dimension (population, time, setting, mechanism, comparator). For each: 1-sentence question plus 1-sentence rationale.Replace: [topic], [field]
Optimization: If candidates stay vague, add: “Each question must specify at least one of: population, time, setting, comparator. No abstract framings.”
2. PICO formatter (clinical / health)
Convert my research interest "[interest]" into a PICO-formatted question: Population, Intervention, Comparator, Outcome. Output 3 candidate PICO formulations and note which would be most feasible given my constraint of [constraint].3. PEO formatter (qualitative)
For my qualitative interest "[interest]", produce 3 PEO-formatted questions: Population, Exposure / Experience, Outcome / theme of interest. Note the methodological tradition (phenomenology, ethnography, grounded theory) that best matches each.4. Variable map
For my candidate research question "[question]", identify: independent variable(s), dependent variable(s), key moderators, potential confounders, and the unit of analysis. Output as a table.5. FINER feasibility audit
Audit my research question "[question]" on FINER criteria: Feasible (time, sample access, instruments), Interesting (to whom), Novel (vs prior literature), Ethical, Relevant (to whom). Score each 1-5 with a 1-line justification. End with one revision suggestion.6. Operationalization probe
For my question "[question]", how would I operationalize each construct? Suggest 2 candidate measures per construct, citing one prior study that used each.Run this one in a deep-research tool so the cited studies are real and clickable, not invented.
7. Scope-creep detector
Below is my current research question draft and a list of "things I also want to study". Identify which extras would blow scope; suggest which to defer to a follow-up study and which can be folded into the primary question.
Question: [paste]
Extras: [paste list]8. Stakeholder framing
My question "[question]" matters to which stakeholders (researchers, practitioners, policymakers, patients, students)? For each, give the 1-sentence answer they would want from my study. Mark which framing fits a thesis vs a paper vs a policy brief.9. Comparison-pair refinement
My current question compares [A] vs [B] in [setting]. Suggest 3 alternative comparator framings (different baseline, different intervention magnitude, different population) and the trade-off each implies.10. Hypothesis derivation
For the research question "[question]", state 2-3 specific hypotheses that follow. For each: directional or non-directional, what data would test it, what would constitute a "null" result.11. “Already-answered” check
Has my question "[question]" likely been answered already? List 3 search strings I should run, the kinds of sources I should check, and 2 indicators that the question is settled vs still open.For the real verdict, run the search strings inside Elicit or Semantic Scholar, or hand the question to a deep-research mode that browses live and cites sources.
12. One-paragraph specific aims
Convert my research question "[question]" into a 200-word specific-aims paragraph: long-term goal, overall objective, central hypothesis, 2-3 specific aims, expected impact. Voice: NIH-style if applicable.13. Mentor-pitch script
I have 2 minutes with my mentor. Write a 200-word script pitching the research question "[question]": why it matters (1 sentence), what is known (2 sentences), the gap (1 sentence), the question, the method (1 sentence), feasibility (1 sentence). End with the 1 ask I should make.14. Refinement diff
Below is my old question and my new question after refinement. Identify the moves I made (narrowed population, added comparator, swapped outcome) and any remaining weaknesses.
Old: [paste]
New: [paste]15. Final FINER one-liner
Refine my draft question into a single-sentence FINER-compliant research question. Include population, comparator (if any), outcome, and time-frame. Max 30 words. Topic: "[topic]". Draft: "[draft]".Which AI tool for which step (June 2026)
The prompts work in any chatbot, but a few steps reward specific tools:
| Step | Best tool | Why |
|---|---|---|
| Generating candidates (1-3), reframing (9, 14) | Any frontier chat — Claude Opus 4.7, GPT-5.5, or Gemini 3.1 Pro | Reasoning quality is high across all three; pick the one you already pay for. |
| Reading 10+ papers before you start | Claude (Opus 4.7 / Sonnet 4.6) — 1M-token context, or Gemini 3.1 Pro (1M) | Drop a stack of PDFs in one chat; ChatGPT Plus holds far less in-app (full 1M only on the $200 Pro tier). |
| Operationalization (6) and “already-answered” (11) | ChatGPT Deep Research (Plus, $20/mo), Gemini Deep Research, or Perplexity | These browse live and attach citations, so the studies they name are real. |
| Paper discovery and extraction | Elicit, Semantic Scholar | Purpose-built for the literature, not general chat. |
A plain chatbot will happily invent a citation for prompt 6 or 11. Always verify any study a non-browsing model names before you put it in a proposal.
Common mistakes
- Stopping at “I want to study X” — that is a topic, not a question.
- Skipping PICO / PEO when a structured frame would have saved weeks of confusion.
- Leaving feasibility until late — questions that pass FINER on paper still need real-world sample access and instruments.
- Not naming a comparator; without a contrast you have description, not investigation.
- Trusting an AI-named citation without checking it exists. Non-browsing models fabricate references.
- Letting the question expand to cover everything you find interesting; scope creep kills projects.
- Treating the refined question as final; expect to revise it 3-5 times before fieldwork.
How to push results further
- Write at least 5 candidate questions before picking one (template 1).
- Pair refinement with feasibility (template 5); the best question you cannot execute is a bad question.
- Operationalize early (template 6); if you cannot measure it, your question is still fuzzy.
- Run the “already-answered” check (template 11) in a tool that cites real papers before you commit.
- Keep a “deferred questions” file for the extras you cut; they often become your next study.
- Bring 3 candidate questions to the mentor meeting, not one; you will leave with better options.
- Revise the question after the first 5 papers, after IRB submission, and after data collection starts.
FAQ
- How specific should a research question be?: Specific enough that the data needed to answer it is identifiable in one sentence. If you cannot name that data, refine further.
- Does every field need PICO?: No. PICO is for quantitative and clinical intervention questions. PEO fits qualitative or observational work; other fields use independent-variable / dependent-variable framings or theoretical questions. FINER applies to all of them as an audit step.
- Which AI model is best for this in 2026?: For generating and auditing candidates, any of Claude Opus 4.7, GPT-5.5, or Gemini 3.1 Pro works well. For reading many papers at once, Claude and Gemini 3.1 Pro offer 1M-token context. For checking whether a question is already answered, use a deep-research mode that browses and cites (ChatGPT Deep Research on Plus, Gemini Deep Research, or Perplexity).
- How long does refinement take?: For a thesis, 2-4 weeks of iteration is normal. Less than a week usually means the question is under-refined.
- Can AI write my research question?: No. It generates candidates, audits feasibility, and reframes. Choosing, grounding in the literature, and defending the question are yours to do.
- What if my mentor wants a different question?: Bring 3 refined candidates and the trade-offs. Mentors push back more usefully when you have done the work.