What a literature review is actually for
A literature review answers three questions for the reader: what is known about this topic, where does the field disagree or fall silent, and why does your study therefore need to exist. That third question is the one weak reviews miss — they end after "what is known" and leave the gap implicit. Every paragraph should be quietly building toward "and that is why this study".
It helps to know which kind you are writing. A narrative review surveys and interprets a field with expert judgment about what matters. A systematic review follows a documented protocol — predefined search strings, inclusion criteria, and screening — and is a methodology in its own right. A scoping review maps the breadth of a literature without appraising every study. Most thesis chapters and paper introductions are narrative reviews with systematic habits: transparent about how sources were found, but selective by argument rather than by protocol.
1. Search wider than Google Scholar’s first page
Start from two or three recent, highly relevant papers and mine them in both directions: their reference lists take you backward to the foundations; citation search ("cited by") takes you forward to the current conversation. Add a keyword search in at least one scholarly index — OpenAlex, Scopus, or Web of Science — using synonyms for each core concept, because fields rarely agree on terminology.
Know when to stop: saturation. When new searches keep returning papers you have already seen, and new papers cite the same core set you have already read, the map is complete enough to write from. Log roughly how you searched (databases, key terms, rough date range) — even a narrative review earns credibility from one sentence of search transparency.
- Backward chaining: read the reference lists of your best papers.
- Forward chaining: check who cited them since.
- Keyword search: one scholarly index minimum, with synonyms per concept.
- Recency pass: search the last two years separately, so a 2019 seed paper does not blind you to the current debate.
2. Build a synthesis matrix before writing
The single highest-leverage tool in review writing is a synthesis matrix: a table with one row per paper and columns for the things your argument needs — research question, method, sample, key finding, limitation, and which of your themes it speaks to. Filling it in forces real reading; scanning a finished column shows you instantly where the field agrees, conflicts, and goes quiet.
The matrix is also your defense against the most common structural failure: organizing the review paper-by-paper ("Smith found… Jones found… Lee found…"). Once the matrix exists, the columns — not the rows — become your section headings.
3. Structure by theme, not by paper
Strong reviews are organized around ideas: themes, debates, methodological traditions, or a chronology when the field genuinely developed in stages. Under each theme, papers appear together because they speak to the same question — which is what makes synthesis possible. A serviceable skeleton: an opening that frames the territory and announces the organization; two to four thematic sections; a section on gaps and limitations of the existing work; and a closing paragraph that lands on your research question.
Within each thematic section, the unit of composition is the claim, not the paper: open with a topic sentence stating what the literature shows, support it with grouped citations, then complicate it with the studies that disagree and your assessment of why (different populations? weaker designs? older measures?). That assessment sentence — the one only you could write — is the difference between synthesis and annotated bibliography.
4. Write with an authorial voice
Markers and reviewers consistently reward the same signals: verbs that position studies against each other (extends, challenges, replicates, fails to account for), explicit weighing of evidence quality rather than treating every citation as equal, and paragraphs that end by advancing the argument rather than trailing off after a citation. Tense convention: present tense for what the literature says (Smith argues), past tense for what a study did (participants completed three trials).
Keep direct quotation rare. In most sciences a review with more than a handful of quotes reads as unprocessed; paraphrase demonstrates understanding, and your citation style covers attribution.
The five failures markers see most
If you fix only five things in a draft review, fix these:
- The catalogue: one paragraph per paper, no synthesis. Reorganize by theme using the matrix.
- The missing gap: pages of "what is known" with no argument for what is not. The gap is the point.
- Uncritical citation: reporting every finding as fact. Weigh designs, samples, and measures.
- Stale coverage: nothing from the last two years. Reviewers notice immediately.
- Scope sprawl: reviewing everything adjacent instead of what the research question needs. Every section should be traceable to the question.
Where Kahubi fits
Kahubi runs this workflow in one place: the agent searches OpenAlex and Semantic Scholar and imports papers with open-access PDFs into your library; library chat answers "where do these papers disagree?" with citations back to the sources; and the literature review flow drafts thematic sections grounded in the papers you selected — in your own writing voice, with [CITATION NEEDED] marked wherever your sources cannot support a claim.
Last updated 2026-08-07.