How to Write a Research Paper (Without Losing a Week to Panic)
A research paper answers a focused question using sources, and its real difficulty is sequencing: students who struggle usually drafted too early or researched without a question. The working order that holds up — question → sources → working thesis → outline → draft (middle first) → citations → revision — turns a monster assignment into seven finishable jobs.
- Introduction — research question, why it matters, thesis
- Background / literature — what is already known, where sources disagree
- Argument section 1 — sub-claim + evidence + analysis
- Argument section 2 — sub-claim + evidence + analysis
- Counter-evidence — findings that cut against you, addressed honestly
- Conclusion — answer restated, limits, what future work should ask
- References — every in-text citation, formatted, verified real
1. Turn the topic into a research question
"The opioid crisis" is a library section. "Why did opioid prescriptions keep rising for a decade after the risks were documented?" is answerable. A good research question is specific enough that you can imagine the evidence that would settle it.
2. Skim wide, then read three sources deep
First pass: abstracts and conclusions of 10–15 sources to map the terrain. Then pick the three most load-bearing and read them properly, taking notes with page numbers. Record full citation details the moment you touch a source — chasing them at 2 a.m. before the deadline is how fabricated references happen.
3. Write a working thesis from the evidence
After reading — not before — draft the one-sentence answer to your research question. It will sharpen as you write; its job now is to give every section something to prove.
4. Outline by sections with evidence attached
For a standard paper: introduction, background/literature, two or three argument sections, counter-evidence, conclusion. Under each heading, list the specific sources and data points that section will use. An outline with evidence attached is half the paper.
5. Draft the middle first
Start with your strongest argument section, where you know the material best. Introductions written before the paper exists get rewritten anyway. Keep placeholder citations as you go — "(Smith 2023, p. 41)" — never a bare claim you plan to source later.
6. Build the reference list from your notes
Format every source in the required style (APA, MLA, Chicago) from the details you recorded, and verify each one is real and says what you claim — especially any source an AI tool suggested along the way. A single unverifiable reference can trigger an academic-integrity review.
7. Revise in two separate passes
Structure pass: do the sections answer the research question in order, and does any section argue against the thesis without acknowledgment? Then a line pass for clarity, transitions, and format. Reading the paper aloud catches what silent rereading misses.
8. Run the integrity checks before submission
Verify citations resolve, check quotes against sources, confirm paraphrases are actually in your words. If AI assisted anywhere in the pipeline, disclose per your instructor’s policy and make sure the prose reads as yours.
Question: "Did remote learning during 2020–21 widen or narrow achievement gaps?" Thesis: "District-level data show remote learning widened achievement gaps primarily through attendance variance rather than instruction quality — which is why the districts that recovered fastest invested in re-engagement, not curriculum."
Mistakes that cost the most points
Researching forever, drafting never
Endless collection is procrastination in a productive costume. Cap the initial pass at 10–15 sources; you can always add one to fill a specific hole in the outline.
The quote quilt
A paper that is 40% quotation reads as a compilation, not an argument. Default to paraphrase with citation; save direct quotes for wording that genuinely cannot be restated.
Citing sources you never opened
Citing an abstract as if you read the study — or trusting an AI-suggested reference without opening it — is how hallucinated citations end up in bibliographies. Every entry in the reference list should be something you actually saw.
Writing this with AI in the mix
Research papers are where AI assistance is most tempting and most dangerous. Language models are useful for mapping unfamiliar territory and summarizing dense sources you have already opened — but they invent studies, authors, page numbers, and DOIs with total confidence. Fabricated references are the #1 way AI-assisted papers get flagged, independent of any detector.
Two checks before submission: verify every reference with the fake-citation checker (it confirms DOIs and looks up whether sources exist), and run your prose through the AI detector to see what reads as generated. Your instructor is likely doing both.
Sentence-level scan — see what reads as AI before your grader does.
Rewrite flagged passages in a natural, human register.
Verify every reference is real — AI tools invent sources.
Frequently asked questions
What are the parts of a research paper?
Introduction with a research question and thesis, background or literature review, argument sections built on evidence, treatment of counter-evidence, conclusion, and a verified reference list. Empirical papers in the sciences follow the stricter IMRaD format: Introduction, Methods, Results, Discussion.
How many sources should a research paper have?
A common baseline is one to two sources per page of finished text — so 8–15 for a typical undergraduate paper — weighted toward peer-reviewed material. Follow the assignment sheet over any rule of thumb.
Can I use AI for a research paper?
For finding directions and summarizing sources you have opened, usually yes with disclosure; for writing submitted text, usually no. Whatever the policy, verify every AI-suggested source independently — models fabricate convincing citations routinely.