How to write a literature review: structure, synthesis example and template

This guide is for students and researchers writing a review chapter, a review section in a paper, or a standalone review article. It is the pillar of our series on literature reviews and reading, which also covers doing a literature review with AI, how to read a research paper and how to summarize a research paper with AI.

What is a literature review?

A literature review is a structured account of what published research says about a question: which studies exist, what they found, how they differ, and what is still unknown. The UNC Writing Center puts it plainly: a literature review "discusses published information in a particular subject area" [4]. In research, it has three jobs: show that you know the field, justify your own question, and give the reader a fair map of the evidence.

It is not an annotated bibliography, which lists sources one by one. A review is organized around ideas, and each source supports, complicates or contradicts a point.

What are the types of literature review?

The type sets how much method you must report. Grant and Booth identified 14 review types and their methods [1]; the six below cover almost every student and early-career project.

This guide covers the narrative and integrative review that most course and thesis chapters ask for. For the systematic kind, see what is a systematic review and scoping review vs systematic review.

What are the 5 parts of a literature review?

A literature review has five parts: introduction, search approach, body, gaps, and conclusion. Snyder's guidelines for review methodology make the same point: whatever the type, readers need to see what you reviewed, how you chose it, and what it adds up to [2].

1. Introduction. The topic, why it matters, the scope (population, years, settings) and the order of the review.
2. How the literature was found. Databases, search terms, years and inclusion criteria, in one short paragraph.
3. Body. Organized by theme, time or method (see below). Most of the word count goes here.
4. Gaps and debates. What is contested, under-studied or missing.
5. Conclusion. What the literature adds up to, and how it leads to your research question.

Thematic, chronological or methodological: how should you organize it?

Choose the structure that makes your argument easiest to follow [4]:

• Thematic: each section is one theme or sub-question. The default for most reviews.
• Chronological: sections follow how the field developed. Use it only when the order is the point, such as a shift in theory.
• Methodological: sections group studies by method. Use it when results differ by method or your contribution is a method.

Many good reviews are thematic at the top level and compare methods within each theme.

How to write a literature review in 8 steps

Here is how to write a literature review from question to final draft.

1. Write a focused question. "Screen use and sleep" is a topic. "How does evening screen use relate to sleep onset and duration in adolescents aged 12 to 18?" is a question you can review. See how to write a research question.
2. Search at least two sources. Combine a subject database (PubMed, PsycINFO, ERIC, through your library) with a broad index. Record each search string, date and count. See how to search PubMed.
3. Snowball from the best papers. Check their reference lists and who has cited them since. See citation snowballing.
4. Screen with written criteria. Decide in advance what is in (age range, study type, years, language) and apply it to every result.
5. Read in passes and fill a synthesis matrix. Skim to decide, then read the included papers closely and record the same fields for each (template below). See how to read a research paper.
6. Find the themes. Read down the matrix columns, not across the rows. Where do findings cluster or split, and does the split follow method, sample or year?
7. Outline, then draft by theme. Each paragraph opens with a claim about the literature, supported by several studies.
8. Check and revise. Confirm every citation supports its sentence, check for retractions, then reread once only for the argument: does every section lead to your gap?

What is the difference between synthesis and summary?

Summary reports one study at a time; synthesis makes a claim about several studies and uses them as evidence. A review that only summarizes leaves the reader to do the thinking. The UNC Writing Center describes synthesis as reshuffling information into new patterns, while summary restates each source's points [4].

Synthesis is the act of combining findings from several sources to make a claim none of them makes alone, such as a pattern, a contradiction or an explanation of why results differ.

Here is the same material written both ways. The studies below are invented for illustration.

Summary paragraph (weak):

Lee and Park (2020) surveyed 1,200 adolescents in Seoul and found that those who used phones after 10 p.m. slept 34 minutes less on school nights. Okafor et al. (2021) ran a two-week experiment with 96 students in Lagos and found no difference in sleep duration when phones were removed from bedrooms. Martins (2022) used actigraphy with 310 Portuguese teenagers and found that evening screen use delayed sleep onset by 18 minutes on average.

Synthesis paragraph (strong):

Evidence that evening screen use shortens adolescent sleep comes mainly from cross-sectional surveys, and it weakens when sleep is measured objectively or screen access is changed experimentally. A large survey found 34 fewer minutes of school-night sleep among late-night phone users (Lee & Park, 2020), but sleep was self-reported. Studies that measured sleep with actigraphy point to a smaller effect on sleep timing rather than duration, with onset delayed by about 18 minutes (Martins, 2022), and the one experiment that removed phones found no change in duration over two weeks (Okafor et al., 2021). The effect may therefore be real but smaller than surveys suggest, and no study has tested it over a full school term.

The second version leads with a claim, groups the studies by method, explains why results differ and ends on a gap. Useful openers: "Most studies agree that..., although..." and "Results differ by method: surveys report..., whereas experiments...".

Synthesis matrix template

A synthesis matrix is a table with one row per study and one column per theme or variable, used to see patterns across studies before you write. Webster and Watson proposed a concept-centric matrix for exactly this reason: a review should be organized by concepts, not by authors [3].

Copy this blank template into a spreadsheet:

And here it is filled for the example above (invented studies):

Write the page number next to every finding, so you never reread a paper to find a number, and add columns as new themes appear.

A worked literature review outline

This is an example. A master's student writes a 15,000-word dissertation and gives the review 20 percent, or 3,000 words, in line with typical thesis chapter shares (see how to write a thesis outline). Topic: evening screen use and adolescent sleep. Structure: thematic.

Each theme section runs in the same order: the claim, the strongest evidence, the conflicting evidence, the reason for the conflict, and what it means for the question.

How long is a literature review, and how many sources?

Length follows the container: a few paragraphs in a journal article's introduction, around a fifth to a third of a thesis (see the chapter table in how to write a thesis outline), and longest as a standalone review article. There is no correct number of sources; use these decision rules:

• Coverage: every major study, method and position on your question appears.
• Saturation: new searches and reference lists keep returning papers you already have.
• Recency: recent work is well represented, unless your topic is historical.
• Balance: you cite work that disagrees with your position, not only work that supports it.

Common literature review mistakes

How to write a literature review faster in AI for Research

AI for Research helps with the slow parts: finding papers, extracting the same details from each, and a first draft you then rewrite. You still decide the question, the themes and the argument. The full method is in AI for literature review.

• Find and snowball, free. Find Papers searches 250M+ scholarly works with filters for year, study type and journal quality; Citation Map shows references and citing papers. Both cost 0 credits.
• Fill a review matrix from your PDFs. The literature review feature builds a matrix from Library papers: one row per paper, columns such as method, sample, findings and limitations, or your own themes. Each cell shows its quote and page. 1 credit per paper.
• Ask a PDF a question in My Library: answers carry page citations, 1 credit each. Storing PDFs is free.
• Draft in the Writing Studio. The literature review module drafts a cited review organized by theme, time or method: 15 credits for a whole draft, 3 for one section. Without an account, the literature review generator gives 3 free runs a day.
• Check before you submit. Citation check confirms each reference exists and flags retractions (0 credits); claim check shows whether cited studies support your claims (2 credits).

The free plan gives 45 credits a month and 25 Library PDFs: enough for a 20-paper matrix and one full draft. Plus is $12 a month for 750 credits. See pricing.

Limits: the AI draft groups what it found; it does not know your argument or your supervisor's expectations, so treat it as raw material. Matrix cells are AI readings of the text: open the quoted page before you use a number. Scanned PDFs need OCR first. For a systematic review, use the Systematic Review workspace instead, which has screening and PRISMA reporting.

FAQ

What are the 5 parts of a literature review?

Most literature reviews have five parts: an introduction that sets the topic, scope and purpose; a short note on how you found and chose the sources; a body organized by theme, time or method; a section on gaps and debates; and a conclusion that says what the literature adds up to and how it leads to your study. Short reviews inside a paper may merge the first two.

How long is a literature review?

It depends on where it sits. Inside a journal article it may be a few paragraphs of the introduction. In a thesis it is usually a full chapter, often around a fifth to a third of the total word count, so roughly 3,000 to 4,500 words in a 15,000-word master's dissertation. A standalone review article is longer. Your handbook or target journal sets the real limit.

How many sources should a literature review have?

There is no fixed number. The right number is the one that covers the main studies, debates and methods on your question. A practical test is saturation: when new searches and reference lists keep returning papers you already have, you have most of the field. Undergraduate reviews often cite a few dozen sources; doctoral reviews cite far more. Ask your supervisor what is expected.

What is the difference between a literature review and a summary?

A summary reports what one source says. A literature review compares many sources: it groups them, shows where they agree and disagree, explains why results differ, and says what is still unknown. If each paragraph of your review is about one paper, you have written summaries. If each paragraph is about one idea and cites several papers, you are synthesizing.

Can AI write a literature review for me?

AI can find papers, pull the same details from each one and produce a first draft with citations to real papers. It cannot decide what your review should argue, judge which studies are trustworthy or know which gap matters to your project. Use it to save time on search and extraction, then check every citation against the paper and write the argument yourself.

Next steps

• Speed up search, extraction and drafting without losing control: AI for literature review.
• Read faster and more critically: how to read a research paper.
• Turn each paper into a reliable note: how to summarize a research paper with AI.
• Find the gap your review should end on: how to find a research gap.

Sources

1. Grant, M. J., & Booth, A. (2009). A typology of reviews: an analysis of 14 review types and associated methodologies. Health Information & Libraries Journal, 26(2), 91–108. https://doi.org/10.1111/j.1471-1842.2009.00848.x
2. Snyder, H. (2019). Literature review as a research methodology: An overview and guidelines. Journal of Business Research, 104, 333–339. https://doi.org/10.1016/j.jbusres.2019.07.039
3. Webster, J., & Watson, R. T. (2002). Analyzing the past to prepare for the future: Writing a literature review. MIS Quarterly, 26(2), xiii–xxiii. https://www.jstor.org/stable/4132319
4. The Writing Center, University of North Carolina at Chapel Hill. Literature reviews. https://writingcenter.unc.edu/tips-and-tools/literature-reviews/
5. Page, M. J., McKenzie, J. E., Bossuyt, P. M., et al. (2021). The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ, 372, n71. https://doi.org/10.1136/bmj.n71
6. Torraco, R. J. (2005). Writing integrative literature reviews: Guidelines and examples. Human Resource Development Review, 4(3), 356–367. https://doi.org/10.1177/1534484305278283