01 · The Question
Does a Longer Search Strategy Mean a Better Search?
Some database searches are only a few lines long. Others contain hundreds of terms, multiple controlled-vocabulary headings, spelling variants, truncations, proximity operators, field codes, and nested Boolean expressions. Faced with the second version, it is easy to assume that considerable length must indicate considerable rigor.
But length is not a quality criterion. A long strategy can contain excellent vocabulary and careful logic. It can also contain unnecessary concepts, redundant terms, badly chosen synonyms, incorrect Boolean operators, overly restrictive fields, or a single syntax error that changes the result substantially.
The useful question is therefore not "How long is this search?" but "Does this search retrieve the literature required by the research question with an appropriate balance of coverage and relevance?"
03 · What You Need to Know
Why Search Length Tells You So Little About Search Quality
A search strategy is a retrieval model, not a vocabulary collection
A structured database search represents a research question in terms that a particular information system can retrieve. That requires several decisions: which concepts should constrain the search, how each concept might be expressed, which controlled-vocabulary terms are appropriate, how terms should be combined, which fields should be searched, and whether any limits are justified.
Simply accumulating terms addresses only part of that problem.
Consider a search containing 80 synonyms for three concepts. If those three concepts are joined with AND but the third concept is poorly reported in titles and abstracts, the search may systematically miss relevant studies despite its impressive length. Conversely, a shorter two-concept strategy may retrieve the relevant literature more successfully.
This is why adding more search terms can sometimes make retrieval worse rather than better.
Long searches can contain the wrong concepts
One of the most consequential problems occurs before individual keywords are chosen. A researcher can represent the research question using inappropriate searchable concepts.
The PRESS guideline for peer review of electronic search strategies explicitly asks whether the research question has been translated appropriately, whether the search concepts are clear, whether too many or too few elements have been included, and whether concepts are too broad or too narrow. In other words, the conceptual architecture itself requires evaluation.
Suppose a review question contains population, intervention, comparator, outcome, setting, and study-design criteria. Turning all six into mandatory search blocks may appear thorough. Yet every block joined with AND creates another condition that a record must satisfy. Some criteria may be much more reliably assessed during screening.
A carefully constructed search therefore does not necessarily reproduce every element of the research question. Whether every concept belongs in the search string is a retrieval decision, not a formatting decision.
A long concept block can still use poor terminology
Length within a concept does not guarantee good coverage either. Twenty terms may consist mostly of minor lexical variations while omitting a major synonym, historical term, acronym, spelling variant, or controlled-vocabulary heading.
For example, a strategy might devote many lines to variants of "artificial intelligence" yet omit terminology used for a particular technology central to the literature. The block is long, but its lexical coverage is uneven.
The opposite problem also occurs. Researchers sometimes add words that are conceptually adjacent rather than genuinely synonymous. OR then retrieves records containing any of those terms, so one ambiguous expression can introduce a substantial amount of irrelevant literature.
PRESS therefore evaluates free-text terms and subject headings separately, including whether relevant synonyms are missing, whether terms are too broad or narrow, whether truncation is appropriate, and whether relevant subject headings have been used.
Correct terms can still be combined incorrectly
A search can contain excellent vocabulary and still fail because of its logic.
Suppose the intended structure is:
(adolescent* OR teenager*) AND (depression OR "depressive disorder")
If parentheses are omitted or operators are placed incorrectly, the database may interpret the expression differently from what the researcher intended. A strategy that occupies two pages does not compensate for a Boolean error near the beginning.
The same applies to proximity operators, phrase searching, field restrictions, and NOT. PRESS specifically treats Boolean and proximity operators as a separate area for peer review because operator choice and nesting can alter retrieval substantially.
When complex expressions are required, understanding how parentheses control database logic becomes more important, not less.
Complexity creates more opportunities for hidden errors
A long search strategy may include dozens of nested expressions, line references, field tags, controlled-vocabulary commands, truncation symbols, and proximity operators. Each component introduces another opportunity for a transcription, syntax, or translation error.
This does not mean complex strategies should be avoided. Some retrieval problems genuinely require them. It means complexity carries a verification cost.
A strategy developed in MEDLINE, for example, cannot necessarily be pasted unchanged into another database. Search platforms differ in controlled vocabulary, field codes, proximity syntax, truncation rules, phrase handling, and other commands. A lengthy search translated mechanically may therefore look complete while containing platform-specific errors. Search-strategy development literature similarly treats database translation as a distinct stage after initial strategy development and testing.
For multi-database projects, the important skill is translating the search logic into each database's syntax, not preserving the appearance or length of the original query.
Redundancy is not the same as comprehensiveness
Long strategies sometimes contain terms that contribute nothing to retrieval. A term may already be captured by truncation, mapped automatically to a controlled-vocabulary concept, duplicated elsewhere, or rendered redundant by another expression.
Some redundancy can be intentional and defensible. Searchers may retain overlapping terms to protect retrieval across indexing states or database behavior. The important point is that repetition itself does not establish coverage.
Comprehensive
The strategy adequately represents the terminology and searchable concepts needed to identify the intended literature.
Long
The strategy contains many terms, lines, or commands. This describes its size, not its retrieval quality.
A short strategy is not automatically better either
The lesson should not be reversed into "shorter is better." Some concepts genuinely have rich terminology. New and interdisciplinary fields may use competing labels. Older literature may use historical terminology. Controlled vocabulary may need to be combined with free-text expressions, acronyms, spelling variants, and related forms.
Search-strategy development commonly involves dividing the question into searchable concepts, identifying terms within each concept, combining terms appropriately, and then testing and refining the resulting strategy.
If that process produces a long strategy, length may simply reflect the underlying literature. The problem begins when complexity is added without evidence that it contributes useful retrieval.
Search performance matters more than appearance
A strategy should ultimately be judged by how it behaves. Does it retrieve relevant records that should reasonably be discoverable? What kinds of irrelevant records dominate the results? What happens when individual concepts or terms are changed? Are key known studies retrieved? Can the logic be explained and reproduced?
PRESS was developed specifically to support structured peer review of search strategies and includes six broad areas: translation of the research question, Boolean and proximity operators, subject headings, text-word searching, spelling and syntax, and limits and filters.
Notice what is absent from that list: word count.
04 · A Practical Example
When 70 Search Terms Perform Worse Than 25
Hypothetical Example
A systematic search about generative AI and student learning
A researcher develops a database strategy containing approximately 70 terms. It includes blocks for university students, generative AI, academic assessment, learning outcomes, student satisfaction, English-language instruction, and empirical studies. The strategy looks highly detailed and returns 146 records.
Test known relevant records Several clearly relevant papers identified during preliminary exploration are missing from the results.
Inspect the missing records Some do not mention "learning outcomes" in their abstracts. Others describe university participants without using the researcher's chosen student terminology. A few do not identify their empirical design using the design terms included in the query.
Reconsider the concept structure The researcher removes search blocks that are essential for eligibility but unreliable for retrieval, retaining the central searchable concepts of generative AI and higher education. Necessary synonyms and controlled vocabulary remain.
Retest the strategy The revised strategy contains only 25 terms and retrieves 1,080 records. It requires more screening, but the known relevant papers return and inspection reveals additional plausible studies that the longer strategy had excluded.
The 25-term strategy is not better because it is shorter. It is better in this hypothetical example because its searchable concepts align more appropriately with the retrieval task. If testing showed that another concept improved precision without unacceptable loss of relevant material, retaining that concept could be justified.
06 · What This Means for You
Make Every Part of the Search Earn Its Place
Do not set a target number of keywords or lines. Start with the retrieval problem, identify the concepts that genuinely need to constrain the search, and develop terminology for those concepts from the literature and relevant controlled vocabularies.
Then test. If a term adds useful records, it has a reason to remain. If a concept substantially improves precision without unacceptable losses, it may be justified. If removing ten expressions changes nothing meaningful, their presence should not be mistaken for additional rigor.
A simple decision framework
If the strategy is long because a concept has many legitimate names
Keep the necessary terminology and verify that the terms actually represent the concept appropriately.
If the strategy is long because every eligibility criterion became a search block
Reassess which concepts genuinely need to constrain retrieval.
If many terms are included "just in case"
Test their contribution rather than assuming they increase comprehensiveness.
If the strategy contains complex nesting, proximity, field codes, or line combinations
Check the logic and database syntax carefully and consider structured peer review for consequential searches.
If additional terms no longer improve useful retrieval
For systematic reviews and other evidence syntheses, documentation matters as well. A strategy should be reproducible and sufficiently transparent for another researcher or information specialist to understand what was searched and how. Reporting a long strategy faithfully is important, but reporting length does not validate its design.
07 · A Quick Checklist
How to Audit a Long Search Strategy
Before treating a long strategy as finished, check:
Can I justify every concept that is combined with AND?
Do the free-text terms reflect terminology actually used in relevant literature?
Have I checked appropriate database-specific controlled vocabulary where available?
Are OR, AND, NOT, parentheses, proximity operators, and field restrictions doing exactly what I intend?
Are any terms redundant, excessively broad, or included without a clear retrieval purpose?
Does the strategy retrieve known relevant records that it should reasonably find?
Have I examined what is gained or lost when important blocks or terms are changed?
If I translated the strategy between databases, did I adapt the syntax rather than merely copy it?
For a consequential evidence synthesis, would independent search-strategy peer review identify errors I have stopped noticing?
