01 · The Question
When Does a More Detailed Search Become a Worse Search?
You have built a search around the main concepts in your research question. Then another relevant word occurs to you. You add it. Then another. Perhaps you add the outcome, the setting, an age group, a methodological term, and several increasingly specific descriptors. The search looks more sophisticated, but fewer records appear.
Is that an improvement because the results are now more focused, or have you quietly made relevant research harder to find?
The answer depends on what you mean by "adding terms." Adding synonyms with OR can broaden the representation of an existing concept. Adding another concept with AND usually narrows the search. Those two operations have very different consequences, even though both make the search string longer.
03 · What You Need to Know
Why More Terms Can Either Improve or Damage Retrieval
Not all additional search terms do the same thing
The first distinction is structural. Suppose you are searching for studies about artificial intelligence in higher education. You might begin with two concepts:
(artificial intelligence) AND (higher education)
You then discover that authors use several expressions for artificial intelligence. Expanding the first concept might produce something like:
("artificial intelligence" OR "generative AI" OR "machine learning") AND ("higher education" OR universit* OR college*)
You have added terms, but you have not necessarily made the search narrower. Within each concept, OR allows a record to match any of the alternatives. This can increase the likelihood of retrieving records that use different terminology.
Now imagine adding another required concept:
("artificial intelligence" OR "generative AI" OR "machine learning") AND ("higher education" OR universit* OR college*) AND (learning outcomes)
This is fundamentally different. A potentially relevant record must now satisfy the artificial-intelligence concept, the higher-education concept, and the learning-outcomes concept. If an eligible article investigates learning outcomes but its searchable title, abstract, keywords, or indexing do not contain the terms you supplied for that concept, the record may disappear from your results.
Adding terms with OR
Usually expands the ways a concept can be expressed and tends to broaden retrieval.
Adding concepts with AND
Requires records to satisfy another condition and therefore usually narrows retrieval.
This is why the number of words in a search string tells you surprisingly little about its quality. What matters is the logical role each term plays. Understanding how Boolean operators change retrieval is therefore more useful than counting terms.
Every AND creates another condition a record must satisfy
Boolean AND retrieves the intersection between sets. If concept A retrieves records about a population and concept B retrieves records about an intervention, A AND B retrieves records represented in both sets. Adding concept C produces A AND B AND C, so eligible records must now also match C.
That can be exactly what you want. If C represents an indispensable concept that distinguishes your question from a much broader literature, adding it may greatly improve precision. But if C is inconsistently reported or indexed, the same move can reduce sensitivity.
This distinction matters especially in evidence synthesis. The Cochrane Handbook advises against searching every aspect of a review question automatically and recommends avoiding too many different search concepts. Comparators and outcomes, for example, may be poorly represented in titles, abstracts, or controlled vocabulary. A concept can therefore be important for deciding whether a study is eligible without necessarily being a good concept for finding that study.
Eligibility criteria and search criteria are not identical
A common source of over-restriction is translating every inclusion criterion directly into the database query. Suppose your study requires university students aged 18 to 25, use of generative AI, assessment of critical thinking, and a particular study design. It may seem logical to search all of those elements simultaneously.
But database searching and study selection perform different jobs. The search attempts to retrieve plausible candidates. Screening then determines whether those candidates actually meet the eligibility criteria.
If an eligible study describes participants simply as "undergraduate students" in the abstract while reporting their ages only in the full text, an age requirement embedded in the search may prevent you from retrieving it. The study still meets your criterion. The searchable record simply does not express the criterion in the way your query requires.
For this reason, not every concept in a research question necessarily belongs in the search string.
Searchability matters as much as conceptual importance
A concept may be central to your research question but difficult to search reliably. Outcomes are a familiar example. Researchers can describe the same outcome using different constructs, instruments, terminology, or levels of specificity. Some outcomes may not appear in titles or abstracts at all.
The same problem can occur with populations, settings, comparators, and methodological characteristics. Before adding a concept, ask not only, "Is this important to my study?" but also, "Can this concept be represented reliably in the fields and indexing available in this database?"
If the answer to the second question is uncertain, making that concept mandatory deserves testing rather than assumption. In some searches, it may be safer to retrieve a broader set and determine the characteristic during screening. This issue becomes particularly important when deciding whether an outcome should be omitted from the search strategy.
More specificity can improve precision while reducing sensitivity
Search quality cannot be judged simply by whether the result count decreases. A narrower search may contain a larger proportion of relevant records, which means better precision. Yet it may simultaneously miss relevant records, which means poorer sensitivity or recall.
| Change to the search |
Likely effect |
Main risk |
| Add a useful synonym with OR |
Usually broadens retrieval |
May add irrelevant records if the term is ambiguous |
| Add another required concept with AND |
Usually narrows retrieval |
May exclude relevant records that do not express the concept in searchable metadata |
| Add a highly specific phrase |
May improve precision |
May miss alternative wording |
| Add a population, outcome, or design restriction |
May focus results substantially |
May depend on incomplete or inconsistent reporting and indexing |
Neither sensitivity nor precision is inherently "good" in isolation. Their relative importance depends on the purpose of the search. A systematic review generally places considerable emphasis on identifying as much eligible evidence as reasonably possible, whereas a researcher looking for a few papers to understand an unfamiliar topic may tolerate a more focused search. The broader trade-off is addressed when considering search sensitivity versus precision.
A longer query can also add noise rather than restriction
Extra terms do not always make a search too narrow. They can make it unnecessarily noisy. A broad synonym may have several unrelated meanings, an aggressive truncation may capture unintended words, or a loosely specified term may retrieve records from another domain.
For example, adding every remotely related expression with OR can substantially increase retrieval without identifying additional relevant studies. The problem is then not lost sensitivity but declining precision and greater screening burden.
So there are at least two ways extra terms can hurt: a restrictive term can exclude relevant records, while an overly broad term can flood the results with irrelevant ones. This is one reason a very long search strategy can still be a poor search.
The effect of a new term should be tested, not guessed
Search development is iterative. Run the search before and after a meaningful change. Compare the result counts, inspect newly retrieved or newly lost records, and examine whether known relevant studies remain retrievable.
For example, if adding an outcome concept reduces a search from 2,400 records to 380, the reduction itself does not demonstrate improvement. Look at records that disappeared. Are clearly relevant studies among them? Does the narrower version still retrieve key papers that should reasonably be found by the database search?
The Cochrane Handbook describes checking whether a strategy retrieves known key publications as one useful performance test, while cautioning that retrieving known studies alone does not establish that the search is sufficiently sensitive. Known records are diagnostic examples, not proof of completeness.
Watch Out
A dramatic reduction in the number of results can feel like progress because the search becomes easier to screen. That is not evidence that the search became better. Relevant records may have disappeared along with the irrelevant ones.
04 · A Practical Example
How One Extra Concept Can Remove a Relevant Study
Hypothetical Example
Searching for generative AI and critical thinking among university students
Suppose a researcher wants studies examining whether generative AI affects critical thinking among university students. The researcher has already identified one hypothetical article that clearly meets the eligibility criteria and can use it to observe how changes to the search behave.
Search 1: Start with the central searchable concepts ("generative AI" OR ChatGPT OR "large language model*") AND (universit* OR undergraduate* OR "higher education"). The hypothetical relevant article is retrieved.
Search 2: Add the outcome as a required concept The researcher adds AND ("critical thinking" OR "critical reasoning"). The result count falls sharply, but the known relevant article disappears.
Investigate why The article's abstract describes "reasoning performance" and "evaluation of arguments." The phrase "critical thinking" appears only later in the full text. The study fits the research question, but its database record does not express the outcome using the searcher's terminology.
Revise the strategy Rather than assuming that the smaller result set is better, the researcher tests whether the outcome concept can be expanded reliably. If outcome terminology remains unstable, the researcher may omit that concept from the database query and assess the outcome during screening.
The lesson is not that outcomes should always be removed. It is that an additional concept should earn its place in the search by helping identify the intended literature without causing unacceptable loss of relevant records.
06 · What This Means for You
Decide What Each Additional Term Is Supposed to Accomplish
Before adding a term, identify its function. Is it another way authors might describe an existing concept? Is it a genuinely necessary concept? Is it being added only because the current result set feels inconveniently large?
That last reason deserves particular scrutiny. Screening burden is real, but reducing the number of records is not the same as improving retrieval. If your search returns an unmanageable number of results, diagnose why before adding restrictions. The appropriate response may involve improving an ambiguous concept rather than simply adding another AND condition. A more systematic approach is useful when a search returns thousands of results.
A simple decision framework
If the new term is a genuine synonym or alternative expression
Test it within the existing concept using OR and inspect the additional records it retrieves.
If the new term represents another concept joined with AND
Ask whether that concept is essential for retrieval and reliably represented in searchable records.
If adding the concept removes known relevant records
Investigate why they disappeared before retaining the restriction.
If a term adds many records but no apparent relevant material
Check whether the term is ambiguous, overly broad, or unnecessary.
If repeated additions produce little useful improvement
Keep versions of the strategy as you test it. Database search histories are particularly useful because they allow you to compare sets rather than repeatedly rewriting a large query. When a modification changes retrieval substantially, inspect samples from the difference between versions. Search development becomes much more informative when you can explain what a change did rather than merely observe that the result count moved.
07 · A Quick Checklist
Before You Add Another Search Term, Check This
Before keeping an additional term or concept, check:
Can I explain what concept this term represents and why that concept needs to be searched?
Should the term be combined with synonyms using OR, or am I introducing a new requirement with AND?
Is this characteristic consistently present in titles, abstracts, keywords, or controlled vocabulary?
Does the revised strategy still retrieve relevant records I reasonably expect it to find?
Have I examined some records gained or lost after making the change?
Am I adding this restriction because it improves retrieval, or merely because I want fewer results to screen?
Could this criterion be assessed more reliably during screening than during database retrieval?
If the term is joined with OR, have I checked whether ambiguity creates substantial irrelevant retrieval?
