How to write a literature review (step by step)

A literature review does three jobs. It shows the reader what has been established about a topic, it shows where the evidence is thin, contested or missing, and it explains why your own project (or your conclusion) follows from that picture. A list of summaries does none of these well. The test of a good review is whether a reader comes away knowing what the field thinks and why, not just which papers exist.

Reviews come in a few common types, and the type decides how strict your method has to be:

• Narrative review: a reasoned overview of a topic, typical in coursework and in the background chapter of a thesis. The search can be selective, but it should still be described.
• Systematic review: answers a precisely defined question with a pre-specified, reproducible search and screening process, often with a protocol registered in advance. Reporting usually follows the PRISMA 2020 statement.
• Scoping review: maps the size and nature of the evidence on a broad question without necessarily appraising each study in depth.
• Integrative review: combines different kinds of evidence (for example, experimental and qualitative studies) to build a fuller picture.

If your assignment or supervisor names a type, follow it. If not, a carefully documented narrative review is the usual default for coursework and thesis chapters.

The most common reason a review sprawls is a question that is too broad. "Social media and mental health" will return tens of thousands of records. "What do longitudinal studies say about daily social media use and depressive symptoms in adolescents?" tells you what to search, what to include and what to leave out.

Write your question down in one sentence, then list:

• the population or setting you care about,
• the exposure, intervention or concept,
• the outcomes or aspects of the topic,
• the kinds of study you will accept, and the date range and languages you can realistically cover.

Health and social science students often use the PICO frame (Population, Intervention, Comparison, Outcome) to structure this. Any frame is fine as long as it turns the question into explicit inclusion and exclusion criteria.

Pick two to four databases that index your field. Your library's subject guide will list them; common examples are PubMed for biomedicine, PsycINFO for psychology, ERIC for education, IEEE Xplore for engineering, and multidisciplinary indexes such as Scopus and Web of Science. Google Scholar is useful for following citations but hard to search reproducibly, so do not rely on it alone.

Build the search string#

• List synonyms for each concept in your question (adolescent, teen, youth, young people).
• Combine synonyms with OR and concepts with AND.
• Use truncation where the database supports it (adolescen* finds adolescent and adolescence) and quotation marks for exact phrases.
• Add the database's controlled vocabulary where it has one, such as MeSH terms in PubMed.

Record what you did#

For each database, note the date, the exact search string, any filters and the number of results. This takes minutes and lets you, your supervisor or an examiner reproduce the search. For a systematic review it is required; for a narrative review it is still what separates a defensible method from "I searched around".

Then widen the net: check the reference lists of the most relevant papers (backward citation searching) and see who has cited them since (forward citation searching).

Export the results to a reference manager, remove duplicates, and screen in two passes: titles and abstracts first, then full texts. Apply the criteria you wrote in Step 1 and note the reason each full text was excluded. If you are doing a systematic review, the PRISMA 2020 flow diagram is the standard way to report how many records you found, screened, excluded and included.

While screening, confirm that every source you keep is real and correctly described. Reference details copied from other papers, search snippets or AI tools are sometimes wrong, and occasionally the paper does not exist. Our guide on how to check that a citation is real covers DOI and database checks and how to spot retractions.

A synthesis matrix is a table with one row per source and one column per thing you need to compare. It makes patterns visible that you would miss reading paper by paper. Useful columns include:

Write notes in your own words and mark direct quotations clearly, with page numbers. Mixing the two in your notes is one of the main ways accidental plagiarism happens later.

Once the matrix is full, look down the columns rather than across the rows. Which findings agree? Where do studies disagree, and can the disagreement be explained by design, sample or measurement? Which questions has nobody tested properly?

Then choose an organising structure:

• Thematic: sections built around themes or debates. This is the most common and usually the strongest.
• Methodological: grouped by research approach, useful when the methods themselves explain differing results.
• Chronological: tracing how understanding developed. Use it only when the development is the point; otherwise it drifts into summary.

A literature review usually has three parts.

Introduction#

State the question and why it matters, define key terms, and say how you found and selected the sources (databases, dates, criteria). Tell the reader how the review is organised.

Body#

One section per theme. Open each section with a topic sentence that makes a claim, support it with the studies that bear on it, weigh their quality, and end with what the section establishes. Be explicit about the strength of the evidence: "one small study suggests" and "a consistent finding across large cohorts" are different claims.

Conclusion#

Summarise what is known, what is uncertain and what is missing. In a thesis or proposal, finish by showing how the gap leads to your research question. If you want a starting structure with word budgets, the free paper outline builder produces one for a literature review.

Format every reference in the style your course requires. If you are unsure of the rules, see the APA 7, Harvard and Vancouver pages, or format a single reference with the free reference formatter.

Before you submit, read the review once for argument only. Delete paragraphs that summarise a paper without connecting it to a theme, check that every claim attributed to a source is something that source actually says, and make sure the conclusion follows from the body. The printable literature review checklist walks through these checks.