What Concepts From Your Research Question Should Actually Become Search Terms?

Which Parts of Your Research Question Should You Actually Search?

Once you have identified the main concepts in a research question, the next temptation is understandable: turn every concept into a group of keywords and require all of them in the search.

Suppose your research question is:

Does the use of generative artificial intelligence for academic writing improve writing self-efficacy among undergraduate university students compared with students who do not use generative AI?

You could identify at least four concepts: undergraduate students, generative AI, academic writing, and writing self-efficacy. The comparison with students who do not use AI adds another element.

But should every one of those concepts become a mandatory part of the database search?

Not necessarily. The concepts needed to define a research question and the concepts worth requiring in a search strategy are related, but they are not always identical. Some concepts discriminate strongly between relevant and irrelevant records. Others may be inconsistently reported, poorly indexed, implicit in the literature, or so restrictive that requiring them causes useful studies to disappear.

Search the Concepts That Help Retrieve the Evidence You Need

The Research Question Defines the Study, but the Search Has a Different Job

Start with concepts, not individual words

Before deciding what belongs in the search, distinguish a concept from a search term.

A concept is an idea you need to represent. A search term is one expression used to retrieve that idea from a database.

For example, the concept generative artificial intelligence might eventually be represented by terms such as generative AI, GenAI, names of relevant technologies, or database-specific subject headings where suitable ones exist.

This means the first decision is not, “Which words from my question should I type?” It is, “Which ideas from my question should the database be required to find?”

If you have not yet made that conceptual translation, first turn the research question into a searchable question before developing the vocabulary for individual concepts.

Being important to the research question does not automatically make a concept useful for retrieval

A research question may need considerable specificity. You might specify a population, intervention or exposure, comparison, outcome, setting, time period, and other characteristics because those details define what you want to investigate.

A database search operates differently. Each concept you require generally creates another retrieval condition. When concept blocks are combined with AND, a record ordinarily has to satisfy every block to remain in the result set.

That creates an important distinction:

The answers can differ. A characteristic may be essential to eligibility but unnecessary as a mandatory search concept because you can evaluate it during title, abstract, or full-text screening.

Prioritize concepts that distinguish the literature you want

A useful search concept usually contributes meaningfully to retrieval. Imagine searching for literature about generative AI and academic writing among university students. Requiring a concept related to generative AI would ordinarily make sense because it distinguishes the topic from the enormous literature on academic writing generally.

Now consider the phrase undergraduate university students. Whether it should be required is less obvious. Relevant articles might describe their participants as college students, higher education students, first-year students, university learners, or simply students. Some abstracts may not specify undergraduate status at all.

The population remains important to the research question. The retrieval question is whether you can represent it sufficiently well without losing too much relevant literature.

That judgment should be tested rather than assumed.

Comparators often do not need their own search block

In structured questions, particularly those based on PICO, the comparison may be important to the research design without being useful as a search concept.

Suppose you are interested in studies comparing generative AI-assisted writing with conventional writing. Relevant abstracts may describe the intervention prominently but refer to the comparison as a control group, traditional instruction, usual practice, non-AI condition, or something specific to the study design. The comparator might not be described clearly in the searchable fields at all.

Requiring a comparator block can therefore eliminate records that otherwise satisfy the substantive topic.

This principle is well established in systematic review searching. Cochrane advises that it is usually unnecessary, and may sometimes be undesirable, to search every aspect of a review question. Comparators and outcomes can be particularly problematic because they may not be well described in titles and abstracts or consistently represented by controlled vocabulary.

Outcome concepts can be surprisingly restrictive

Outcomes often feel indispensable because they may be central to the research question. Yet outcome terminology can vary substantially across studies.

Suppose you want research on whether an educational technology affects academic engagement. One study may use that phrase. Another may report behavioral engagement. Others might focus on participation, persistence, time on task, interaction, or scores from a named engagement instrument without prominently using your preferred terminology.

If the population and intervention already identify a manageable body of literature, requiring an outcome block may reduce sensitivity without providing enough additional precision to justify the loss.

This does not mean outcomes should never be searched. Sometimes an outcome is precisely what distinguishes the literature you need. The point is that its inclusion should be justified by retrieval performance rather than by its position in the grammatical structure of the research question.

Population concepts may or may not need to be searched

Population terms are often useful, particularly when the population meaningfully distinguishes the target literature. Searching for an intervention in children, for example, may require a population concept when the same intervention is extensively studied in adults.

But population terminology can also be unreliable. Age categories, educational levels, occupational groups, diagnostic categories, and other population characteristics may be described differently across records.

Ask what happens when you require the population concept. Does it remove mostly irrelevant records, or does it also remove studies you know should be found?

The answer may differ by discipline and database.

Setting and context are often better treated cautiously

Research questions frequently specify a setting such as universities, hospitals, rural communities, developing countries, online learning environments, or public schools.

These details may be central to the study. They are not automatically good search concepts.

A study conducted at a university, for example, may not contain the phrase higher education in its title or abstract. A paper about an intervention in a rural hospital may focus its searchable language on the intervention and patient population rather than the rural setting.

If context is indispensable and reliably described, searching it may be useful. If it is inconsistently reported, screening may be safer.

One concept usually needs more than one term

Once you decide that a concept belongs in the search, the next problem is vocabulary. Searching only the wording from your research question assumes that authors and indexers use the same terminology you do.

Often they do not.

Cochrane recommends using a broad range of free-text terms within selected concepts, considering synonyms, acronyms, spelling variants, truncation, wildcards, and other appropriate search features. For databases with controlled vocabulary, subject headings should also be considered alongside free-text terms.

The underlying architecture commonly looks like this:

OR broadens representation within a concept by allowing different ways of expressing the same or closely related idea. AND requires the selected conceptual blocks to intersect.

This is why Cochrane recommends avoiding too many different search concepts while using a wide variety of appropriate terms within each concept for sensitive systematic searches.

Free-text terms and controlled vocabulary solve different retrieval problems

Many bibliographic databases allow you to search words appearing in records as well as standardized subject headings assigned during indexing. In MEDLINE, for example, records are indexed using Medical Subject Headings, or MeSH. The U.S. National Library of Medicine describes MeSH as its controlled vocabulary thesaurus for indexing and searching biomedical literature.

These approaches complement each other.

MeSH, for example, contains preferred descriptors as well as entry terms that provide alternative access points to concepts. Other databases may use different controlled vocabularies, and the vocabulary and indexing practices are not interchangeable across databases.

A concept that belongs in your search therefore does not become one keyword. It becomes a concept block whose exact representation may change when the strategy is adapted to another database.

Search terms should come from more than your own vocabulary

Your first terms will often come from the wording of the research question, but that is only a starting point.

Useful terminology can also be identified by examining relevant articles, especially their titles, abstracts, author keywords, and database indexing. Controlled vocabulary browsers can reveal preferred subject headings and related terms. Preliminary searches may uncover terminology that researchers in the field use but that you did not initially anticipate.

This makes search development iterative. You formulate a concept, identify candidate terms, search, inspect what appears, and revise.

The search itself becomes a source of vocabulary.

Deciding Which Concepts to Require in a Search

The researcher initially identifies five possible elements:

• undergraduate university students;
• generative artificial intelligence;
• academic writing;
• writing self-efficacy;
• a comparison condition without generative AI.

Instead of automatically turning all five into mandatory blocks, the researcher evaluates what each contributes.

The researcher might begin by testing the intersection of generative AI and academic writing. After inspecting the results, the undergraduate concept could be added if the search is dominated by irrelevant populations. Writing self-efficacy could also be tested as an additional block.

If adding that outcome causes several known relevant papers to disappear, the researcher has learned something important: the outcome may be essential to eligibility but unreliable as a mandatory retrieval concept.

If the question contains many more interacting elements than this example, it may be useful to break the complex question into smaller searchable concepts before deciding which blocks to combine.

Common Mistakes When Choosing Search Concepts and Terms

Use Retrieval Value to Decide What Belongs in the Search

When evaluating a candidate concept, do not ask only whether it matters to your research question. Ask what requiring it will do to the search.

This is a balancing problem rather than a mechanical conversion exercise. A concept that seems indispensable on paper may contribute little to retrieval, while a concept that appears broad may be exactly what keeps the search sensitive.

The practical test comes from the results. If your concept choices produce an enormous amount of unrelated material, examine whether the search has become too broad. If relevant studies you expected to retrieve are disappearing, investigate whether the search has become too narrow.

For exploratory searching, you may intentionally begin with fewer required concepts and refine from there. That decision is part of determining how broad the initial search should be.

Before Turning a Concept Into a Required Search Block

Questions About Choosing Concepts and Search Terms

Do Not Search a Concept Merely Because It Appears in the Question

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.