A reproducible literature review search strategy makes your evidence base easier to defend, update, and explain. This guide shows how to define concepts, choose complementary academic databases, build Boolean searches, record filters and exclusions, export results, and maintain a search log that can support a thesis, article, or systematic review.
Overview
A literature search is more than entering a few keywords into one database. Different platforms describe the same research with different subject terms, indexing practices, coverage, and search syntax. A strong workflow therefore combines a clearly defined question with a documented process for finding peer-reviewed journals, scholarly articles, reports, and other relevant literature.
The goal is not to make every database produce identical results. The goal is to create a transparent search that another researcher could understand, repeat, and reasonably adapt. This matters for systematic reviews, but it is also useful for dissertations, scoping reviews, evidence-informed assignments, and research projects that may need an update later.
Start by separating your work into four layers:
- Question: What population, topic, intervention, exposure, outcome, or setting are you studying?
- Concepts: Which ideas must appear in the search, and which are optional?
- Sources: Which databases, repositories, citation indexes, and specialist collections are appropriate?
- Records: What did you find, remove, screen, include, exclude, and export?
Before searching, decide whether your project requires a broad exploratory search or a formal systematic search. A formal review generally needs more detailed methods documentation, duplicate checking, screening records, and reporting aligned with the guidance required by your department, journal, or review protocol.
What to track
1. Define the concepts before listing keywords
Write the question in ordinary language first. Then identify its main concepts. Depending on the discipline, useful frameworks may include population, concept, context; population, intervention, comparison, outcome; or exposure, outcome, and setting. Do not force every question into a framework that does not fit. The purpose is to expose the parts of the question that need search terms.
For each concept, list synonyms, spelling variants, abbreviations, older terminology, controlled vocabulary terms, and phrases used by relevant authors. Search recent review articles and several relevant research articles for terminology, but treat their reference lists as discovery aids rather than automatic evidence of inclusion.
| Concept | Possible terms | Notes |
|---|---|---|
| Population | Formal name, abbreviation, demographic terms | Check whether age, location, or clinical terms are necessary |
| Topic or intervention | Alternative names, technical terms, spelling variants | Include historical terminology when the time period requires it |
| Outcome | Measurement terms, related conditions, observable effects | Consider leaving narrow outcomes out during the first search |
| Context | Sector, setting, region, or professional group | Use only when context is central to the question |
2. Build Boolean groups
Use OR within a concept group and AND between essential concept groups. Use quotation marks for exact phrases when the database supports them, and use truncation carefully for word variations. Wildcards and truncation symbols differ by platform, so check the database help page before relying on them.
A generic structure might look like this:
(population OR synonym* OR "alternative phrase") AND (topic OR intervention OR related term*) AND (outcome OR measure*)
Use NOT sparingly. It can remove unwanted meanings, but it can also exclude relevant records that mention the unwanted term in a secondary context. Test any exclusion against a small set of known relevant articles.
3. Compare databases by function
Choose databases for complementary coverage rather than collecting a long list without a plan. A general scholarly index may provide broad discovery, a subject database may offer better disciplinary terminology and filters, and a citation index may help you trace articles that cite or are cited by a key study. Library discovery systems and open repositories can add useful material, but their result sets may not be directly comparable to those of curated academic databases.
| Source type | Best use | What to record |
|---|---|---|
| Subject database | Precise searching within a discipline | Subject headings, field codes, filters, coverage notes |
| Multidisciplinary index | Broad discovery across fields | Search string, date range, document types, result count |
| Citation index | Forward and backward citation tracking | Seed articles, citation links, export method |
| Library discovery service | Initial exploration and full-text locating | Scope, duplicate behavior, limits applied |
| Repository or open-access search | Finding accessible versions and supplementary literature | Repository name, query, version type, access date |
Database names, field codes, export limits, subject coverage, and interface instructions can change. Record the platform and access date, and confirm syntax in the database documentation rather than copying instructions from an undated guide. For help evaluating indexing claims about journals, see this guide to verifying Scopus, Web of Science, and DOAJ coverage.
4. Keep a search log
Your search log should capture enough detail to reconstruct the search. A spreadsheet is usually sufficient; a version-controlled document or research notebook can add useful history when the strategy changes.
Use these columns as a reusable literature search log template:
- Date and time: When the search was run.
- Database and platform: Include the database name and host interface.
- Search identifier: The saved-search name or query number, if available.
- Exact search string: Preserve the complete query, including field codes and limits.
- Filters and limits: Date, language, population, document type, or other restrictions.
- Result count: Record the number shown before and after relevant refinements.
- Export details: File format, fields exported, batch number, and destination.
- Notes: Errors, interface changes, unexpected terminology, and decisions to revise the search.
Also maintain a decision log for exclusions. Record the reason at the level needed for your project: duplicate, wrong population, wrong outcome, outside date range, unavailable document type, or not relevant after full-text review. Consistent exclusion reasons are more useful than lengthy notes that cannot be applied uniformly.
Cadence and checkpoints
Use checkpoints to prevent a search from drifting. First, run a small pilot search in one database. Inspect the first several pages of results and identify whether the records reflect the intended topic. If important known articles do not appear, diagnose the problem before adding more filters. The issue may be terminology, field selection, overly narrow phrases, or missing database coverage.
Next, adapt the strategy to each database. Do not assume that a query copied from one interface will work unchanged elsewhere. Map free-text terms to the database's available subject vocabulary where appropriate, then retain free-text terms for newer or inconsistently indexed concepts.
After each major search:
- Save the final query and the result count.
- Export records with abstracts, keywords, identifiers, and citation information when available.
- Store the original export without editing it.
- Create a working copy for deduplication and screening.
- Back up the search log and exported files using consistent file names.
A practical file name might include the project, database, date, and version, such as project_database_2026-11-05_v1. If you revise the query, preserve the earlier version and explain why the change was made. This prevents accidental loss of the original method.
For an ongoing review, consider a scheduled check every month or quarter, depending on how quickly the field changes and what your protocol requires. A scheduled check does not always mean rerunning the full search. It may mean reviewing saved alerts, checking newly published material, or confirming that the original databases and search syntax remain available.
How to interpret changes
A different result count does not automatically mean that one database is better. Counts can change because of indexing updates, revised records, platform changes, changed filters, query translation, or database coverage. Interpret the search as a documented process, not as a single number.
When results are unexpectedly high, inspect a sample rather than immediately adding restrictions. Look for an ambiguous term, a broad truncation symbol, or a concept that is not essential to the question. When results are unexpectedly low, remove optional concepts first, test synonyms separately, and review the language used in relevant records.
When adapting a search between databases, compare whether the same known relevant articles can be found, whether major terminology is represented, and whether the filters have equivalent meanings. A subject heading in one database may not have an exact equivalent in another. Document the judgment used to adapt it.
After exporting, check for duplicate records using DOI, database identifiers, title, author, and publication year. Do not rely on a single field: small differences in punctuation or metadata can hide duplicates. A separate reference audit can help verify identifiers, citation matches, and retraction-related concerns; see the research paper reference audit checklist.
Finally, distinguish search performance from evidence quality. A database can help you locate records, but inclusion decisions still require reading titles, abstracts, and often full texts against predefined criteria. Keep the search strategy, screening decisions, and extracted evidence as separate parts of the workflow.
When to revisit
Revisit the strategy when the review question changes, a key term becomes outdated, a new population or setting is added, or early screening shows that the search is consistently missing relevant research. Also revisit it before submitting a manuscript, completing a thesis methods chapter, or presenting a review as current evidence.
For a continuing project, set a monthly or quarterly reminder to review saved searches, alerts, database access, and the search log. A more substantial update may be appropriate every six to twelve months, or sooner after a major platform change. At each update, verify:
- Database names, coverage, and access routes.
- Search syntax, field codes, subject headings, and truncation rules.
- Filters, export limits, and available metadata fields.
- Search dates, result counts, and any changes from the previous run.
- Whether new terminology or citation links should be added.
- Whether reporting requirements for the target journal, department, or review method have changed.
End each update with a short change note: what changed, why it changed, when it was tested, and which records were added or removed as a result. That habit turns a one-time database search into a maintainable research workflow. Once the search is stable, connect it to a literature review matrix so that discovery, screening, and analysis remain traceable from the original query to the final written synthesis.