How to Write a Literature Review: 7 Steps to Synthesis

As of 2026, writing a literature review still means doing more than collecting good sources. Define the question. Decide what belongs in scope. Search in a way you can explain later. Compare the proof, then turn that work into an argument your reader can check.

Use this stage-gate model when you need a repeatable process:

1. Define the research question and boundaries.
2. Find sources with a documented search strategy.
3. Screen sources against inclusion and exclusion criteria.
4. Extract the evidence you will compare.
5. Synthesize patterns, conflicts, gaps, and methods.
6. Write the review around claims that compare sources.
7. Verify citations before the draft leaves your desk.

That sequence works for course papers, thesis chapters, and internal research memos. A systematic review or scoping review needs a stricter plan and fuller reporting. It may also need a framework such as PRISMA. Do not call a class paper a systematic review unless you used that method.

What a literature review is

A literature review is a structured account of what credible sources say about a research question. Do not stop at single paper summaries. The finished review should show where sources agree and clash. It should also show which methods they use, what proof is missing, and what the pattern means for your topic.

Purdue OWL frames a literature review as putting sources in conversation with each other. Start by writing the review job in one sentence, then set three boundaries before searching:

• Population or subject: Who or what does the review cover?
• Concepts: Which theories, methods, measures, or events matter?
• Evidence type: Which sources count as evidence, and which sources are background only?

For a class paper, these limits can be short. For a thesis chapter, they need enough detail that another reader can see why a source was kept or cut. For a systematic or scoping review, set the rules before the search begins. The NCBI literature-review methods chapter is a useful source for those method boundaries.

Extract evidence into a review matrix

The biggest mistake is writing before the source set is stable. Drafting too early makes the review sound like a list of paper notes because each new paper changes the shape. Work through the stages below before turning notes into paragraphs.

1. Define the research question

Use a question narrow enough to guide search terms and broad enough to support synthesis. "Social media and mental health" is too broad for most reviews. "How do short-form video platforms affect sleep quality among undergraduate students?" gives you a population, topic, and outcome.

Write down the choices that will shape the search and the draft:

• the main research question.
• two or three subquestions.
• the time period you will cover.
• disciplines or databases you will search.
• source types you will include or exclude.

If the assignment already gives you a prompt, translate it into search terms. If it asks for a theory, include that theory and its common synonyms. If it asks for a policy or intervention, include the population and outcome.

2. Build a search strategy

Search in places that fit the field. Start with your library search tools and Google Scholar. Use PubMed or PMC for health topics. Use ERIC for education, PsycINFO, and IEEE Xplore for engineering. Follow citation trails from strong papers.

Use keyword blocks rather than one long query so you can test each part:

• core concept: literature review topic terms.
• group or setting: student, patient, workplace, region, age group.
• method or study type: survey, trial, interview, model, program.
• outcome: learning, retention, anxiety, cost, accuracy, safety.

Keep a small search log with database, query, filters, date, and useful hits. Most course papers do not require a formal PRISMA diagram. They still need enough traceability to explain how you found the evidence. University guides from Texas Libraries, the University of Arizona Libraries, and Concordia Library all stress that a review needs a traceable search and source process.

3. Screen sources before taking deep notes

Screen sources in layers so you do not spend an hour annotating a paper that later falls outside scope. First check title and abstract. Then check methods, sample, source quality, and relevance to the question. Finally, read the most relevant papers closely.

Use inclusion criteria that tell you why a source belongs in the review:

• directly addresses the research question.
• uses a method appropriate to the claim.
• includes enough context to interpret the finding.
• is current enough for the field.
• is cited by, or in conversation with, other credible sources.

Use exclusion criteria that tell you why a source should stay out:

• only mentions the topic in passing.
• lacks the data or method needed for your claim.
• duplicates a stronger source.
• is outside the population, setting, or time period.
• is background reading rather than evidence for the review.

This step protects the draft from becoming a pile of interesting but unfocused sources.

4. Extract evidence into a review matrix

A synthesis matrix turns reading notes into rows you can compare. Do this before writing body paragraphs. The matrix forces each source to answer the same questions, which makes patterns easier to see.

Table 1: Keep the table accurate as you read. A weak source can still be useful for background, but it should not carry a central claim. A strong source can support a claim only if the method and sample match the scope of the review. If your main problem is storing and finding papers before synthesis, compare options in our guide to research paper organizers.

5. Draft and verify the review

Draft only after the matrix shows enough patterns to support the section plan. Start each body section with a claim about the literature, then use the sources to prove, limit, or challenge that claim. Before the draft is done, check that every key sentence still points to a source you have opened.

Use this order when you move from notes to prose:

1. Group rows that support the same theme or debate.
2. Write one claim for the group.
3. Add the strongest source support.
4. Name the limits or conflict.
5. Check the source link before the claim reaches the final draft.

How to synthesize sources

Synthesis starts when you stop asking, "What did this paper say?" Ask instead, "What can I claim after comparing these papers?"

Use four passes through the matrix so the notes turn into claims:

1. Pattern pass: Which findings recur across sources?
2. Conflict pass: Which sources disagree, and is the disagreement caused by method, population, measure, or theory?
3. Gap pass: Which important population, outcome, or method is missing?
4. Claim pass: Which claim can you support with more than one source?

Atlas fits at this stage if your sources are already collected. Add the relevant papers to one project. Ask for themes, clashes, methods, and gaps across the set. Then inspect the citations behind the claims that matter.

Atlas can search and add papers by DOI, arXiv ID, title, author, or topic. It can also compare proof across project sources in cited answers. Treat those citations as check links. Open the source passage before you move a claim into the draft.

Example: turn notes into synthesis

Here is the handoff from rough paper notes to prose you can defend:

• Paper note: Study A found a correlation between evening short-form video use and worse self-reported sleep.
• Paper note: Study B found that notifications delayed sleep preparation in a lab setting.
• Paper note: Study C found that app limits helped some students but broke down during exam weeks.
• Final claim: The proof points less to screen use in general and more to timing, alerts, and context. Evening use may disrupt sleep routines. App limits can help, but they depend on student workload and follow-through.

That synthesized claim is stronger than a one-paper summary because it explains how findings relate. It also keeps the limits visible. The first study shows a link but cannot prove cause. The second is a short lab test. The third depends on behavior during high-stress weeks.

Prompts to move from notes to synthesis

Use prompts that force source separation before you trust the output:

• "Compare these papers by question, method, sample, main finding, and limit."
• "Where do the sources agree and disagree about the mechanism?"
• "Which claims are supported by more than one source, and which rely on a single paper?"
• "Show the strongest citation for each claim and name any conflicting evidence."

The output is not the literature review. It is a draftable evidence map. Your job is to decide which claims belong in the argument and which sources need another read.

For a fuller tool-assisted workflow, see our guide to AI for literature review. If you want step-by-step review stages without tool selection, use the literature review process guide.

How to structure the review

Most reviews work better when they are grouped by ideas, methods, debates, or stages of the problem. Do not group by author unless that helps the reader.

Use a thematic structure when sources cluster around recurring concepts:

1. Define the topic and scope.
2. Explain the major themes.
3. Compare evidence within each theme.
4. Identify gaps or tensions.
5. Explain how the review supports the next section of the paper.

Use a method-based structure when the field disagrees because studies use different methods:

1. Introduce the question.
2. Group studies by method or evidence type.
3. Compare strengths and limits.
4. Explain what each method can and cannot show.
5. State which evidence is strongest for your claim.

Use a time-based structure only when the review explains change over time. This works for the growth of a theory, policy, program, or tool. A timeline can become a source list if each paragraph only says what happened next.

For each body paragraph, use a pattern that keeps claims tied to proof:

1. Make a claim about the literature.
2. Name the sources that support or complicate the claim.
3. Explain the method or evidence behind the claim.
4. Show the implication for the research question.
5. Close by naming the next comparison.

This weak paragraph plan only stacks paper summaries:

• Study A says X.
• Study B says Y.
• Study C says Z.

A stronger paragraph plan compares the sources inside one claim:

• Two survey studies and one lab study suggest that timing matters more than total screen time.
• The survey data shows a link, while the lab study shows one likely cause.
• The app-limit study adds a caveat because workload affects follow-through.
• The review should split timing, alerts, duration, and workload instead of treating "screen use" as one thing.

Where AI helps and where it cannot

AI can speed up parts of a literature review. It can help you sort many sources, compare methods, draft table columns, or find possible themes. It should not replace your source checks, inclusion choices, citation checks, or final judgment.

Use AI for tasks where a draft map or source sort saves time:

• turning paper notes into a comparable matrix.
• finding repeated themes and disagreements.
• drafting search-query variants.
• checking whether a paragraph has claim-evidence structure.
• summing up one source before you read the key parts closely.

Do not use AI for decisions that require your own source judgment:

• inventing sources or filling citation gaps.
• deciding that a weak study proves a strong claim.
• treating a source link as correct without opening it.
• writing final claims you have not checked against the paper.
• calling a narrative review a systematic review.

Atlas is useful when the review depends on cited comparison across a collected set of papers. Ask for cited comparisons. Then check each key claim against the original passage. If a citation is weak, missing, or tied to the wrong source, narrow the question or revise the claim before using it.

For tool choice, compare AI research assistants, AI tools that summarize research papers, and AI with references. For the concept-level view, read what literature review AI is.

Next step: verify your review

Before you submit or publish the literature review, read it once as a skeptic. Look for claims that sound stronger than the proof. Find paragraphs that summarize one source at a time. Open citations you have not checked since drafting.

Use this final checklist to catch problems before the review leaves your desk:

• Every body section answers the research question.
• Each major claim is backed by source proof.
• Conflicting proof is named rather than hidden.
• The review separates background sources from core proof.
• The structure follows themes, methods, or time for a reason.
• Systematic or scoping review language is used only when the method supports it.
• Citations point to sources that directly support the sentence.

A good literature review does not prove that you read a lot. It proves that you can judge how sources relate and explain what that link means for the question.

If you are revising a draft, check the common literature review mistakes before you submit.

Compare the full literature review process