What Is a Literature Review? A Practical Guide to Search, Synthesis, and Verification

Start your literature review now

Before searching, complete this task entry:

Review purpose: I need this review for [course paper / thesis / proposal / research report / evidence synthesis].
Question: I want to understand [specific relationship, experience, intervention, concept, or debate].
Scope: My population/material, context, timeframe, language, and source types are [limits].
Review type: My required approach is [narrative / systematic / scoping / not yet confirmed].
Authority: The method, reporting standard, or assignment guidance I must follow is [source].

Simplified workflow: choose the review type → define the question and scope → build search concepts → search and record → screen and evaluate → extract notes consistently → synthesize patterns → draft by theme → verify every claim and citation.

Short illustrative example: Instead of listing five papers about generative-AI feedback, a synthesis might conclude: “Across the selected studies, feedback usefulness appears to depend on students’ ability to evaluate and apply suggestions; however, differences in course setting, outcome measures, and study duration limit direct comparison.” This is an illustrative sentence, not a real research finding.

Complete Literature Review with Acade

Acade is an AI academic research assistant for students, researchers, and professionals. For a literature review, it can support topic development, literature search, lawful literature access, academic translation, note organization, mind mapping, outlining, literature-review work, academic writing, and polishing. Acade does not guarantee that a source, citation, quotation, classification, or synthesis is correct. It cannot turn an informal search into a compliant systematic review, replace an information specialist or subject expert, conduct ethics or methodological approval, or assume authorship responsibility. The human researcher must verify original sources, make inclusion and interpretation decisions, contribute the analysis, disclose AI use when required, and take responsibility for the submitted work.

What is the purpose of a literature review?

A literature review builds an evidence-based account of a field or focused problem. Depending on the assignment and method, it can:

• define central concepts and competing definitions;

• show how a question has been studied;

• compare theories, methods, populations, and findings;

• identify agreement, inconsistency, uncertainty, and limitations;

• justify a research question or design;

• locate a defensible research gap;

• establish what a new study should—and should not—claim.

A literature review can be a standalone paper, a chapter in a thesis or dissertation, a section of a research article, or a formal evidence-synthesis project. These outputs are not methodologically interchangeable.

Summary is not synthesis

The first column can be useful during note-taking. It is rarely sufficient as the final review.

Choose the right type of literature review

The label controls the rigor, documentation, team needs, and claims you may make.

PRISMA states that PRISMA 2020 is a reporting guideline designed to improve reporting of systematic reviews. It provides checklists and flow diagrams, but following a reporting checklist does not repair an inadequate review method after the fact. See the official PRISMA statement and PRISMA 2020 resources.

The Cochrane Handbook distinguishes systematic reviews through thorough, objective, and reproducible searching across appropriate sources, with prespecified eligibility and transparent study selection. For health intervention reviews, one MEDLINE search alone is not considered adequate. See Cochrane Handbook Chapter 4 and Chapter 3 on eligibility and synthesis grouping.

JBI provides separate evidence-synthesis guidance for scoping reviews. If your goal is a scoping review, follow the current JBI method and applicable reporting guidance rather than relabeling a narrative overview. See the JBI Manual for Evidence Synthesis.

Decision rule: if your instructor, supervisor, institution, funder, journal, or review team specifies a method, that requirement controls. If the type is unclear, resolve it before designing the search.

Step 1: Define the review question and scope

What to do

Write one answerable question and define its boundaries. Include the relevant population or material, phenomenon/intervention/exposure, context, outcomes or concepts, timeframe, language, geography, source type, and study design where appropriate.

Why it matters

An unbounded question produces an unmanageable search and an arbitrary final sample. A scope that is too narrow can hide relevant evidence.

Inputs

• assignment or protocol requirements;

• broad topic and intended audience;

• discipline and review type;

• time, database, language, and access constraints;

• preliminary concepts and known terminology.

Expected output

• one review question;

• inclusion and exclusion boundaries;

• rationale for each important limit;

• unresolved decisions to confirm.

Illustrative example

Broad topic: AI feedback in higher education.

Scoped narrative-review question: “How does recent empirical research describe undergraduate students’ evaluation and use of generative-AI feedback in writing-intensive courses?”

Possible boundaries: undergraduate students; writing-intensive courses; empirical studies; 2022 onward; English-language sources; AI-generated formative feedback; exclude automated grading systems not involving generative AI.

Common error

Choosing date, language, geography, or publication-type limits only because they reduce workload, without explaining what bias they may introduce.

Completion standard

Another informed reader can apply your boundaries consistently and understand what your review will not cover.

Step 2: Build a search strategy

What to do

Break the question into concept groups. Add synonyms, abbreviations, spelling variants, controlled vocabulary where the database supports it, and carefully chosen exclusions. Combine synonyms with OR and distinct concepts with AND.

Why it matters

Authors use different words for the same concept. A natural-language sentence copied into every database usually misses relevant records or retrieves excessive noise.

Inputs

• review question and eligibility boundaries;

• seed papers or authoritative terminology;

• database-specific subject headings and field syntax;

• review-type search requirements.

Expected output

• concept table;

• database-specific search strings;

• search log containing database/platform, exact query, filters, date, and result count;

• plan for citation chaining and supplementary searches.

Illustrative concept table

Example structure:

(“generative AI” OR “large language model”) AND (feedback OR revision) AND (undergraduate OR “college student*”) AND (writing OR composition)

Common error

Using one exact search string unchanged across Google Scholar, PubMed, Scopus, Web of Science, ERIC, PsycINFO, or a library discovery system. Platforms differ in fields, operators, phrase handling, indexing, and result limits.

Completion standard

The strategy can be rerun and the record explains why the selected sources are appropriate. A systematic review requires the more demanding search process specified by its protocol and method, often with an information specialist.

Step 3: Search, document, and manage records

What to do

Run the adapted searches, export records, deduplicate them, preserve the query history, and document supplementary methods such as backward and forward citation searching.

Why it matters

Without a search record, you cannot explain how evidence entered the review, update the search reliably, or distinguish duplicate reports from distinct studies.

Inputs

• approved search strategies;

• relevant academic databases and lawful access routes;

• reference-management or structured record system.

Expected output

• master record set;

• deduplication log;

• search-history table;

• list of records requiring lawful full-text retrieval.

Short example

Record the platform as well as the database: “MEDLINE via [platform], searched [date], exact query [saved], filters [listed], results [count].” Platform details affect reproducibility.

Common error

Treating multiple articles from the same underlying study as independent studies. Cochrane specifically notes that reviews concern studies, even when a study has several reports.

Completion standard

Every record has a traceable source, duplicate decisions are documented, and the search can be reconstructed within the limits of changing databases.

Step 4: Screen and evaluate sources

What to do

Apply the eligibility criteria consistently. For a formal systematic or scoping review, follow the required multi-reviewer, conflict-resolution, exclusion-reason, and flow-reporting procedures. Separately evaluate source credibility and methodological limitations.

Why it matters

Relevance is not the same as quality. A paper can match the topic but provide weak support for a specific claim.

Inputs

• deduplicated records;

• explicit eligibility criteria;

• appraisal method appropriate to the study designs;

• full texts where required.

Expected output

• inclusion/exclusion decisions;

• documented exclusion reasons at the appropriate stage;

• source-quality or critical-appraisal notes;

• final set for synthesis.

Source evaluation questions

• Is this the original study, a review, commentary, preprint, correction, or retraction notice?

• Do the population, setting, method, and outcome match the claim I want to make?

• Are the measures and analysis appropriate?

• What are the limitations, conflicts of interest, and funding disclosures?

• Has the record been corrected, retracted, or superseded?

• Am I relying on an abstract when the full text is needed?

Common error

Excluding a source because its findings are inconvenient, or including it because its conclusion agrees with your expectations.

Completion standard

Each decision is based on a stated rule, and each source’s role in the synthesis is proportionate to what its design and evidence support.

Step 5: Extract notes into a literature matrix

What to do

Use the same fields for each included source. Separate source-reported information, direct quotations, your interpretation, and planned use in the review.

Why it matters

Consistent extraction makes comparison possible and reduces accidental misquotation or unsupported generalization.

Literature matrix template

Illustrative extraction

Instead of “Paper says AI feedback works,” record: “Population: first-year students; setting: one writing course; outcome: self-reported usefulness; design: short survey; cannot establish improvement in writing performance.”

Common error

Copying an AI-generated summary into the matrix without checking the source’s actual method, population, results, and wording.

Completion standard

A second reader can trace every extracted point to the original source and distinguish evidence from your interpretation.

Step 6: Synthesize, do not merely stack summaries

What to do

Group evidence by themes, concepts, methods, populations, chronology, theoretical positions, or competing explanations. Within each group, compare agreement, disagreement, strength, context, and uncertainty.

Why it matters

The literature review’s value lies in the relationship you establish among sources—not the number of summaries you produce.

Inputs

• verified literature matrix;

• theme codes;

• research question;

• quality and limitation notes.

Expected output

• theme map;

• claim-evidence table;

• explanation of convergence and divergence;

• provisional gap statement;

• boundaries on what the literature supports.

Summary vs. synthesis example

Summary-only version:

Study A reports positive student attitudes toward AI feedback. Study B identifies problems with inaccurate suggestions. Study C discusses feedback literacy.

Illustrative synthesis version:

The selected studies frame usefulness as conditional rather than automatic. Positive attitudes are reported when students can interpret and adapt suggestions, while concerns concentrate on inaccurate or context-insensitive feedback. Because the studies use different outcomes and short observation periods, they do not establish whether perceived usefulness produces sustained improvement in writing.

This example demonstrates structure only. It does not summarize real named studies.

Common error

Calling “no paper used my exact wording” a research gap. A defensible gap may concern an underexamined population, unresolved contradiction, inadequate method, missing theory, weak measurement, or limited context—but it must emerge from verified literature.

Completion standard

Each thematic conclusion is supported by multiple appropriate sources where available, exceptions are visible, and uncertainty is not converted into certainty.

Step 7: Build a thematic structure

Avoid organizing the final review as “Author A, Author B, Author C.” Use a structure that answers the review question.

Reusable outline

1. Introduction

   - define the problem and purpose; - state the review question and scope; - identify the review type; - preview the organizing logic.

2. Method or search approach

   - describe sources, search concepts, dates, limits, screening, and appraisal at the level required by the review type.

3. Thematic synthesis

   - Theme 1: what the literature broadly supports; - Theme 2: where results differ and why; - Theme 3: methodological or contextual limitations; - Theme 4: implications and unresolved questions.

4. Conclusion

   - answer the review question proportionately; - state limitations of the review itself; - identify justified next research steps.

Paragraph template

Synthesis claim: State the pattern.
Evidence: Compare multiple sources that support or qualify it.
Difference: Explain conflicting findings, methods, populations, or contexts.
Evaluation: Assess evidence strength and limitations.
Link: Connect the pattern to the review question and next theme.

Step 8: Write and verify the review

Draft from the claim-evidence table, not from memory. Use citations at the point of support, distinguish source findings from your interpretation, and avoid causal or universal language that the evidence cannot justify.

Then run four separate passes:

1. Evidence pass:

   Does each factual or interpretive claim match the cited source?

2. Synthesis pass:

   Does each paragraph compare sources rather than list them?

3. Scope pass:

   Do conclusions stay within population, context, dates, languages, and designs reviewed?

4. Citation and language pass:

   Are metadata, quotations, style, terminology, headings, and transitions accurate and consistent?

For systematic reviews, data extraction must be accurate, complete, transparent, and accessible for later use or updating; see Cochrane Handbook Chapter 5. AI assistance does not remove those requirements.

Complete illustrative worked example

This example is illustrative. It does not cite fictional papers, invent results, or claim to be a systematic review.

Input

• Purpose: a graduate course narrative literature review

• Topic: generative-AI feedback in undergraduate writing

• Audience: instructors and education researchers

• Limits: empirical higher-education studies, 2022 onward, English-language sources available through lawful access

• Time: four weeks

Type decision

The assignment asks for a critical narrative review, not a systematic or scoping review. The student will document the search and selection process for transparency but will not claim comprehensive evidence identification or PRISMA compliance.

Question and search plan

Question: “How does recent empirical research describe undergraduate students’ evaluation and use of generative-AI feedback in writing-intensive courses?”

Concept groups: generative AI; feedback; undergraduate/higher education; writing/revision. Searches are adapted to relevant education, multidisciplinary, and library sources. Citation chaining supplements database searching.

Screening and extraction plan

Include empirical studies addressing students’ use or evaluation of generative-AI feedback in undergraduate writing contexts. Exclude commentary without empirical evidence, automated scoring unrelated to generative AI, and studies outside higher education. Record citation, context, sample, design, feedback task, outcomes, findings, limitations, and claim fit.

Intermediate synthesis

Theme A: students’ ability to evaluate feedback.
Theme B: task and course context.
Theme C: accuracy and trust.
Theme D: limitations in outcomes and study duration.

Illustrative output paragraph

The reviewed evidence would be organized around conditions of use rather than a simple “effective or ineffective” judgment. Studies reporting favorable perceptions would be compared with work identifying inaccurate or context-insensitive suggestions, while differences in writing task, student experience, and outcome measurement would be used to explain variation. Conclusions about durable writing improvement would be withheld unless longitudinal or performance-based evidence supported them.

Quality check

• Review type named accurately: yes.

• Question and limits visible: yes.

• Search process documented: planned.

• Sources and quotations verified: required before drafting.

• Synthesis organized by patterns: yes.

• Completeness or causal claims avoided: yes.

• Gap based on verified evidence: not yet; must be developed after extraction.

Common literature-review errors

• Starting with a conclusion and selecting only supporting papers.

• Searching one platform and implying the search was comprehensive.

• Treating Google Scholar, PubMed, arXiv, and other sources as interchangeable.

• Citing a search-result snippet, AI answer, or secondary citation instead of the original work.

• Confusing reports with underlying studies.

• Mixing narrative, scoping, and systematic methods without naming the change.

• Using “recent” without defining a date boundary.

• Listing papers without cross-source synthesis.

• Treating study quality as a binary label without method-appropriate appraisal.

• Inventing a gap from absence in a small convenience search.

• Omitting contradictory or null findings.

• Presenting correlation as causation.

• Letting polished AI prose hide unverifiable citations or unsupported claims.

Literature review quality checklist

• [ ] The review type matches the question and required method.

• [ ] The question, scope, and exclusions are explicit.

• [ ] Important limits have a rationale and acknowledged consequences.

• [ ] Search concepts include useful synonyms and database-specific adaptation.

• [ ] Search dates, platforms, queries, filters, and results are recorded.

• [ ] Duplicate records and multiple reports of one study are handled explicitly.

• [ ] Eligibility decisions follow stated criteria.

• [ ] Quality or critical appraisal matches the source designs and review type.

• [ ] Every included source has verified identity and metadata.

• [ ] Full texts and quotation contexts are checked where necessary.

• [ ] Notes distinguish source content, direct quotation, and reviewer interpretation.

• [ ] The final structure is thematic, conceptual, methodological, or chronological for a stated reason.

• [ ] Paragraphs compare evidence rather than stack summaries.

• [ ] Contradictions, exceptions, and uncertainty are visible.

• [ ] The gap follows from the reviewed evidence rather than keyword absence.

• [ ] Conclusions stay within the reviewed scope and evidence strength.

• [ ] The review’s own limitations are stated.

• [ ] AI use follows institution, course, journal, and discipline rules.

• [ ] A human researcher has verified every citation and accepts responsibility for the review.

How Acade helps with a literature review

Use Acade as a research assistant within a human-controlled process:

1. Enter the task.

   Tell Acade the review purpose, field, question, required type, audience, deadline, and output format.

2. Provide constraints.

   Supply eligibility boundaries, databases or sources available, language and date limits, institutional rules, and any protocol.

3. Use relevant research capabilities.

   Acade can support topic development, academic literature search from sources that may include Google Scholar, PubMed, arXiv, and other academic sources, lawful literature access, academic translation, note organization, mind mapping, outlining, literature-review work, writing, and polishing.

4. Generate intermediate artifacts.

   Request concept groups, candidate search strings, source-matrix fields, theme codes, claim-evidence tables, and outline options. Do not ask for an unverified “final review” as the first output.

5. Verify evidence.

   Open the original source or reliable record. Check identity, access status, publication type, corrections or retractions, methods, population, results, quotation context, and claim fit.

6. Make human decisions.

   The researcher—and, where applicable, supervisor, review team, information specialist, statistician, or methodologist—decides the review type, protocol, eligibility, appraisal, synthesis, and final wording.

7. Move to the next task.

   After the evidence base is verified, continue to synthesis, research design, proposal development, or an evidence-led draft.

Prompt to give Acade

I am preparing a [narrative/systematic/scoping/not-yet-confirmed] literature review in [field] for [course, thesis, proposal, report, or publication]. My question is [question], and my current scope is [population/material, concept, context, dates, languages, and source types]. Help me create: (1) a review-type decision with assumptions, (2) concept groups and database-adaptation notes, (3) explicit screening criteria, (4) a literature-matrix template, (5) a thematic synthesis plan, and (6) a source-verification checklist. Do not invent citations, DOI records, quotations, findings, or compliance claims. Flag decisions that require my supervisor, information specialist, review team, statistician, ethics body, or journal.

Acade and human responsibility

Acade cannot guarantee a comprehensive search, validate every citation, determine risk of bias without human review, certify PRISMA compliance, perform independent duplicate screening, decide clinical practice, or act as an author. Medical, clinical, legal, and other high-risk reviews require appropriate domain expertise and applicable oversight. Product plans, Credits, Citation features, and Knowledge Base access may change; check the current product interface and plan information before relying on a specific capability.

Conclusion

What is a literature review? It is a structured, critical synthesis of credible research that answers a defined question by comparing evidence, methods, contexts, and uncertainty. Choose the correct review type, document how sources were found and selected, extract information consistently, synthesize across sources, and verify every important claim. Do not call a narrative review systematic unless the method genuinely meets the required standards.

Complete Literature Review with Acade — Tell Acade your review purpose, field, question, required method, evidence constraints, and deadline, then ask it to help you complete and verify the next literature-review task.

FAQ

What is a literature review in simple terms?

A literature review identifies and evaluates relevant research, then explains the patterns, disagreements, methods, limitations, and unanswered questions across that evidence.

How do you write a literature review?

Define a question and scope, choose the correct review type, search appropriate sources, screen and evaluate records, extract consistent notes, organize themes, synthesize across studies, and verify every claim and citation.

How is a literature review different from an annotated bibliography?

An annotated bibliography usually presents separate source entries. A literature review builds a connected argument or evidence account across sources, organized around themes, methods, chronology, or concepts.

Is a literature review the same as a systematic review?

No. A systematic review is a specific form of evidence synthesis using explicit, prespecified, reproducible methods. A narrative literature review may be rigorous and transparent without claiming systematic-review completeness or compliance.

How many sources should a literature review include?

There is no universal number. The appropriate evidence base depends on the question, review type, discipline, assignment, source availability, and required comprehensiveness. Follow the relevant institutional or publication guidance.

Can I use Google Scholar for a literature review?

Google Scholar can help discover sources and support citation chaining, but its coverage, filtering, and reproducibility limitations matter. Use sources appropriate to the field and review type, and document the search.

What should a literature matrix contain?

At minimum, record verified citation details, purpose, context, sample or material, design, methods, relevant findings, limitations, appraisal notes, theme, claim fit, and verification status.

Can Acade write my literature review for me?

Acade can support searching, organizing, outlining, synthesis planning, drafting, and polishing. The researcher must verify sources, make inclusion and interpretation decisions, provide the original synthesis, follow disclosure rules, and take responsibility for the final work.