03 · What You Need to Know
A Search Strategy Connects Your Research Need to the Literature
It helps to separate three ideas that are often collapsed into the word search: the search strategy, the search query, and the search results.
Search strategy
The broader plan for identifying the literature needed for a particular research purpose, including concepts, terminology, sources, query design, refinement, and appropriate documentation.
Search query
The particular combination of terms and operators entered into a search system at a given point in the search process.
Search results
The records returned by the system in response to that query.
A query is therefore an implementation of part of a strategy. It is not automatically the strategy itself.
This distinction becomes especially important when a search needs to do more than help you discover a few useful papers. If your conclusions depend on which studies you identify, the decisions behind retrieval become methodological decisions.
A Search Strategy Begins With an Information Need
Before choosing keywords, determine what the search is supposed to accomplish.
You may be searching to familiarize yourself with an unfamiliar topic, refine a research question, locate methods used in previous studies, identify literature needed to position an empirical paper, or assemble the evidence set for a systematic review.
Those are not identical information needs.
A researcher exploring a new topic can tolerate considerable uncertainty and may benefit from flexible searching. A systematic review needs a much more explicit connection between its eligibility criteria and the methods used to identify potentially eligible evidence.
This is why clarifying what the literature search is supposed to accomplish should precede arguments about which database or search syntax is best.
The Research Question Must Be Translated Into Searchable Concepts
Research questions are written for humans. Databases retrieve records according to searchable fields, indexing, query syntax, and other characteristics of the retrieval system.
You therefore should not simply paste a complete research question into a database and assume that the resulting records represent the literature relevant to it.
Instead, identify the major concepts that need to be represented in the search.
For intervention reviews in health research, PICO is a familiar framework: Population or Problem, Intervention, Comparison, and Outcome. Cochrane recommends using appropriate elements of PICO and study design to inform bibliographic search structure, while also warning that it may be unnecessary or undesirable to search every element of the question. Comparators and outcomes, for example, may not be consistently mentioned in titles, abstracts, or indexing.
Other research questions may require different conceptual frameworks, and some exploratory searches may not need a formal framework at all. The underlying principle remains useful: identify what concepts the retrieval system actually needs to recognize for a record to have a reasonable chance of being relevant.
Search Concepts Are Not the Same as Individual Keywords
Suppose your question concerns university students' use of generative AI for formative feedback.
The concept generative AI might be represented in the literature by several terms, product names, acronyms, or more general descriptions. The concept formative feedback may likewise appear under related terminology.
If you search only the exact words in your research question, you are assuming that authors and indexers describe the phenomenon exactly as you do.
That assumption is often unsafe.
A developed search strategy therefore identifies alternative ways each important concept might be expressed. Depending on the database and topic, these can include:
- synonyms;
- spelling variants;
- abbreviations and acronyms;
- older or newer terminology;
- broader or narrower terms when conceptually justified;
- controlled vocabulary or subject headings;
- relevant free-text expressions.
The objective is not to create the longest possible list. Each term should represent a defensible route by which relevant records might be described or indexed.
Boolean Operators Express Relationships Between Terms
In many bibliographic databases, Boolean operators help translate the conceptual structure of a search into retrieval logic.
OR is commonly used to combine alternative terms representing the same search concept. If a relevant article might use either term, OR allows either to satisfy that part of the query.
AND is commonly used to combine different concepts. Records must satisfy each connected concept for retrieval.
NOT excludes records containing specified terms and should be used cautiously because a relevant record may contain the excluded term for reasons you did not anticipate.
This structure is common, but it should not become a mechanical recipe. Adding more concepts with AND usually reduces retrieval. If a concept is inconsistently reported in titles and abstracts, requiring it may cause relevant studies to disappear.
A strategy therefore involves reasoning about what not to require as well as what to include.
Controlled Vocabulary and Free-Text Searching Do Different Jobs
Some bibliographic databases assign standardized subject terms to records. MEDLINE, for example, uses Medical Subject Headings, commonly known as MeSH, while Embase uses Emtree.
Controlled vocabulary helps retrieve records that discuss the same indexed concept using different wording. Free-text searching, meanwhile, looks for words or phrases in searchable fields such as titles and abstracts.
Neither approach is sufficient in every circumstance.
New concepts may not yet have suitable indexing terms. Recent records may not yet be fully indexed. Authors may use terminology that provides useful specificity beyond a controlled vocabulary term. Conversely, relying only on free text can miss records that use unexpected language but have been indexed under the relevant concept.
Cochrane consequently recommends an appropriate combination of controlled vocabulary and free-text terms for bibliographic database searches in intervention reviews.
This is a good illustration of why “I used these keywords” does not fully describe a sophisticated database search.
The Database Is Part of the Strategy
A carefully constructed query cannot retrieve records that are not represented in the source being searched.
Different bibliographic databases cover different journals, disciplines, publication types, geographical regions, and periods. Their indexing practices and search functionality also differ.
Choosing where to search is therefore part of search design.
A psychology question might require APA PsycInfo. Nursing research may warrant CINAHL. Biomedical research frequently draws on MEDLINE or PubMed and, depending on the review, Embase. Education researchers may use discipline-specific databases alongside multidisciplinary citation indexes. Cross-disciplinary questions can require several sources because the relevant evidence is dispersed across fields.
Cochrane explicitly recommends selecting databases according to the review topic and notes that specialized, cross-disciplinary, and emerging topics may require additional sources.
No clever keyword combination can compensate for searching a database that does not adequately cover the evidence you need.
A Search Strategy May Include More Than Database Searching
Bibliographic databases are only one way of discovering research.
Depending on the research purpose, a strategy may also include backward citation searching, where you inspect the reference lists of relevant papers, and forward citation searching, where you identify later works that cite them.
Other approaches may involve trial registers, repositories, organizational websites, grey literature sources, handsearching, contacting experts, or searching particular journals.
Formal evidence syntheses may need several complementary approaches because no single database or retrieval method captures every potentially relevant study.
Again, this highlights the difference between a search strategy and a query. A strategy concerns the architecture of evidence discovery, not merely the text entered into one search box.
Why Google Scholar Is Useful
Google Scholar is valuable because it searches broadly across scholarly material and makes discovery relatively easy. It can help you locate known papers, discover related work, follow cited-by links, identify different versions of documents, and explore terminology when you are entering an unfamiliar area.
For exploratory searching, that convenience can be extremely useful.
You might enter a natural-language phrase, find a relevant article, inspect its terminology, follow its references, see who cited it, and use what you learn to improve subsequent searches. That is legitimate scholarly searching.
There is no methodological virtue in avoiding a useful discovery tool merely because it is easy to use.
The important question is what role Google Scholar is being asked to play.
Why Typing Keywords Into Google Scholar Is Not, by Itself, a Search Strategy
Imagine this account of a literature search:
“I searched Google Scholar using the keywords AI, students, learning, university and selected relevant papers.”
Several important decisions remain unexplained.
Why were those concepts chosen? What alternative terminology was considered? Did “AI” include earlier intelligent tutoring systems or only generative AI? What did “learning” mean? How were relevant records distinguished from merely interesting ones? Were other databases needed? Were citation trails followed? Did the search change after terminology was discovered? How was the search bounded? What claims will eventually be made about the resulting literature?
The problem is not that the keywords are necessarily bad. The problem is that a list of words tells us very little about the logic connecting the research need to the evidence that was identified.
A search strategy makes that logic deliberate.
Google Scholar Also Behaves Differently From Traditional Bibliographic Databases
Google Scholar is not simply MEDLINE, Scopus, Web of Science, or another bibliographic database with a different interface.
Research evaluating Google Scholar for systematic-review searching has identified important limitations for reproducible and comprehensive searching. Haddaway and colleagues found that its search functionality and result handling created challenges for systematic reviews, while also recognizing its usefulness as a supplementary search resource.
For a researcher exploring a topic, these limitations may matter relatively little. If your objective is to discover terminology and locate promising papers, ranking can actually be convenient.
If you intend to make a stronger claim that a defined evidence base has been identified systematically, the limitations become much more consequential.
Watch Out
Do not confuse finding many relevant papers with demonstrating that you searched adequately for the literature your study requires. A discovery tool can be excellent for finding useful material without being sufficient, by itself, for every kind of evidence-identification task.
A Search Strategy Should Be Adapted to Each Database
One search string should not necessarily be copied unchanged into every database.
Platforms differ in syntax, controlled vocabulary, field codes, phrase searching, truncation, wildcard characters, proximity operators, and other functionality. Even databases covering similar subject areas may index the same article differently.
Cochrane explicitly states that search strategies need to be customized for each database. MEDLINE and Embase, for example, use different controlled vocabularies and indexing approaches.
The conceptual logic can remain consistent while its technical implementation changes.
This distinction is useful:
Conceptual strategy
What concepts and types of evidence need to be found, what terminology may represent them, and which sources are appropriate.
Database-specific implementation
How that conceptual strategy is translated into the subject headings, free-text fields, operators, syntax, and functions supported by a particular search platform.
A Search Strategy Is Developed Iteratively
A strong search strategy rarely appears fully formed on the first attempt.
You run an initial search. You inspect the results. Relevant papers reveal terminology you missed. Irrelevant papers reveal ambiguous terms. A known relevant study fails to appear, so you investigate why. Adding one concept eliminates important records. A subject heading retrieves literature you had not found with free text.
You revise and search again.
Cochrane explicitly describes search-strategy development as iterative, with terms modified in response to what has already been retrieved.
This is not the same as randomly changing keywords until you like the results. Each revision should respond to information about how effectively the developing strategy represents the intended evidence.
This is where topic knowledge and exploratory searching become part of good search design. You need enough familiarity with the field to interpret what the retrieval is teaching you.
Testing a Search Means More Than Looking at the First Page of Results
Search engines train us to judge a query by the relevance of the first several results. That can be useful for ordinary information seeking, but it is a weak test for a search intended to identify a body of evidence.
One practical approach is to assemble several known relevant papers and determine whether the developing search retrieves them. If it does not, investigate why.
Perhaps the article uses terminology absent from your search. Perhaps the database has indexed it under an unexpected subject heading. Perhaps one of your AND concepts is too restrictive. Or perhaps the record is not contained in the database at all.
Retrieving known relevant papers does not prove that a search is comprehensive. It does, however, provide a useful diagnostic test.
For systematic reviews, search-strategy peer review can add another layer of quality assurance. The PRESS guideline provides an evidence-based framework for peer reviewing electronic search strategies, including the translation of the research question, Boolean and proximity operators, subject headings, text words, spelling and syntax, and limits and filters.
Sensitivity and Precision Are Part of Search Design
Search design involves a trade-off between retrieving relevant evidence and avoiding excessive irrelevant material.
Sensitivity, often discussed alongside recall, concerns the ability of the search to retrieve relevant records. Precision concerns how much of what is retrieved is actually relevant.
Increasing sensitivity commonly reduces precision. Adding synonyms may retrieve additional relevant papers while also producing more irrelevant records. Adding restrictive concepts may improve precision but cause relevant studies to disappear.
For Cochrane intervention reviews, the recommended priority is to maximize sensitivity while striving for reasonable precision.
That recommendation should not simply be copied to every literature search. An exploratory search for a class assignment, a rapid review, and a systematic review operate under different constraints and consequences.
The broader lesson is that search design involves trade-offs. A strategy makes those trade-offs purposeful rather than accidental.
A Search Strategy Should Match the Strength of the Claim You Want to Make
If your goal is simply to find several scholarly sources that help you understand a concept, an elaborate reproducible search protocol would usually be disproportionate.
If you intend to state that you identified the relevant evidence on a question, stronger methods are needed.
If the literature itself becomes the dataset for a systematic review, the search becomes part of the research method and requires correspondingly greater attention to source coverage, sensitivity, reproducibility, documentation, and potential bias.
This is why not every literature search needs to be systematic. Search rigor should be proportionate to the purpose and consequences of the search.
Documentation Turns Search Decisions Into a Record You Can Examine
Even a carefully designed strategy becomes difficult to evaluate if you cannot reconstruct what you did.
For formal evidence synthesis, search documentation may include databases and platforms, dates searched, complete database-specific strategies, limits, filters, supplementary search methods, and other relevant procedures. PRISMA-S provides reporting guidance specifically for literature searches in systematic reviews.
Documentation also serves practical purposes. It lets you rerun a search later, explain changes, update a review, compare strategies, and identify where an unexpected set of results came from.
The required level of documentation depends on the research task. Still, even outside systematic reviews, keeping a useful record of your searches can save considerable reconstruction later.
Few things test methodological memory quite like discovering a folder of excellent papers six months later and having no idea which search produced them.