How to Write a Search Strategy for a Literature Review

What a Search Strategy Actually Is

Ask a supervisor what they mean by search strategy and you will get a short answer: show me what you typed, where you typed it, and when. Not a description of your reading habits. The literal strings, the databases, the dates, the counts.

Why so literal? Because a literature review makes a claim about a body of work, not just about the papers you happened to find. When you write that research on a topic has focused mostly on one setting, you are describing a field. The only thing standing behind that description is the method you used to look. A stated search strategy is what converts “here is what I read” into “here is what is out there, and here is how I know”.

How formal it needs to be depends on the review you are writing. A systematic review specifies the full string for every database in an appendix. A narrative review might report a paragraph. Either way the work underneath is the same, and if you have not settled which kind you are producing, decide that first in scoping vs systematic vs narrative review, because it sets how much of this has to be visible.

Step 1: Break the Question Into Concept Blocks

A search string is built from your question, not from your topic. Take the question apart into the two or three ideas that a paper would have to be about to be useful to you. Those are your blocks. If your research question is sharp, the blocks fall out of it almost mechanically.

Take this question: what is known about the effect of remote work on employee wellbeing in the public sector? Three blocks: remote work, wellbeing, and public sector.

Two rules keep this from going wrong. Use no more than three blocks in your first string, because every block you add is another condition a paper must satisfy, and four conditions will usually strangle your results down to nothing. And leave out any block that describes your method rather than your subject: study design, date range, and language are limits, and limits belong in your inclusion and exclusion criteria where you can apply them consciously, not buried inside a string where they silently delete things.

Step 2: Write the Synonym List for Each Block

This is the step that separates a search that works from one that does not, and it is almost always the step people skip. A database does not know what you mean. It matches the characters you gave it. If an author wrote “telework” and you typed “remote work”, that paper does not exist as far as your search is concerned.

So for each block, list every label a researcher might plausibly have chosen. Pull them from four places: the titles of the papers you already have, the keyword lists those papers print under their abstracts, the vocabulary your discipline uses in textbooks, and the subject headings the database itself suggests. Medical databases have a controlled vocabulary called MeSH, and other fields have their own thesauri. If yours does, use it alongside your own words rather than instead of them, since indexing lags behind new terminology by years.

Notice the spelling variants sitting next to each other. British and American spellings, hyphenated and unhyphenated forms, singular and plural. These are not pedantry. They are the difference between forty results and four hundred.

Step 3: Join Them With Boolean Operators

Boolean logic has one rule worth memorising, and everything else is detail: OR inside a block, AND between blocks. OR widens, because a paper only has to match one of the alternatives. AND narrows, because a paper has to match something from each block. Get that pattern right and your string will behave predictably.

Two cautions. Brackets are not optional: without them, a database may read your string left to right and quietly produce something you did not ask for. And NOT is the operator that causes the most invisible damage, because excluding a word removes every paper that mentions it anywhere, including good papers that mention it once in a limitations paragraph. Exclude at the screening stage, where you can see what you are throwing away.

Weak and Strong, Side by Side

The strong version is longer and uglier, and that is the trade you are making. Readability is not the goal. Reproducibility is.

Step 4: Pilot It Against Papers You Already Know

Here is the test that almost nobody runs, and it takes ten minutes. Pick three or four papers you already know are squarely on your topic, the ones you would be embarrassed to have missed. Run your string. Does it return them?

If it does not, you have learned something specific and fixable. Open one of the missing papers, read its title, abstract, and keyword list, and find the word you did not have. Add it to the relevant block and run again. That loop, repeated three or four times, does more for a search than another afternoon of browsing.

Then read the count. There is no correct number, but there are two clear signals. Thousands of hits usually means a block is too broad or you are searching full text rather than title and abstract: restrict the fields, or tighten the vaguest block. Under twenty hits usually means you have too many blocks or too few synonyms: drop the third block and see what happens, since the block you dropped can be applied by eye during screening. Aim for a set you can actually screen, and remember that you are trading precision against recall every time you touch the string.

Step 5: Translate It for Each Database

A string is not portable. Truncation is an asterisk in some databases and a different symbol in others. Field tags differ.Google Scholar ignores most syntax entirely, caps Boolean handling, and has no way to export a clean result count, which is why it is a fine place to start and a poor place to finish. Search two or three subject databases your library subscribes to, and treat Scholar as a supplement rather than the record. Which databases suit your field is covered in how to find academic sources.

Whatever you run, record it as you go, in a table with one row per database: database name, the exact string you used, any field limits, the date you searched, and the number of results. Rebuilding that table three months later from browser history is a special kind of misery, and the date matters because databases keep growing, so a search run in March and a search run in August are genuinely different searches.

Export the results straight into a reference manager rather than bookmarking them. It handles the duplicate removal you will otherwise do by hand across four overlapping databases, and it gives you the deduplicated count your flow diagram needs. If you have not set one up, Zotero, Mendeley, and EndNote compared covers the choice, and organising your research sources covers what to do with everything once it lands.

Step 6: Search Beyond the String

A database search finds papers that used your words. It will still miss work that framed the same idea differently, and the fix is to follow citations rather than keywords.

• Backward chaining. Take the two or three most relevant papers you found and mine their reference lists. This walks you back to the foundational work in your area, which is often too old or too differently worded for your string to catch.
• Forward chaining. Look at who has cited those papers since. This is how you find the most recent work, and it is the fastest route to seeing where the conversation has moved.
• Hand searching key journals. If two journals keep appearing in your results, browse their last few years of contents directly.
• Grey literature. Theses, government reports, and working papers, where your field and your review type expect them. Note where you looked.

Citation chasing is also where research gaps start to become visible, because you begin to see which questions everyone cites and nobody has answered. That is the raw material for finding a research gap.

How to Report It in Your Chapter

The search strategy sits early in your methodology section, just before your eligibility criteria and screening process. A systematic review puts the full string for every database in an appendix. For most dissertation chapters, a paragraph plus a table is the right weight. How much detail your review type demands is the practical difference set out in literature review vs systematic review.

Report the numbers even if nobody demands them. Records identified, duplicates removed, records screened, and full texts assessed are four figures that take seconds to note as you go and are close to unrecoverable later. They are also exactly what a flow diagram needs.

Five Ways This Goes Wrong

• Searching the topic instead of the question. A string built from a broad topic returns a field. A string built from a question returns a conversation you can join.
• One synonym per block. The single most common cause of a thin literature review. If a block has one word in it, you are searching for one research community and missing the others working on the same thing.
• Missing brackets around OR groups. The results look plausible, which is what makes this dangerous. Bracket every block, every time.
• Never writing anything down. If your strategy exists only in your search history, you do not have a strategy. You have an afternoon you cannot repeat.
• Stopping at the first search. A first string is a draft. Two or three rounds of piloting and revising is normal practice, not a sign that you did it wrong. Several of the common literature review mistakes start here.

Do This in the Next Hour

Write your question on one line. Underline the two or three ideas a paper must be about. Give each one its own row and fill that row with every word an author might have used, pulling from the keyword lists of the papers you already have. Bracket each row, put OR between the words and AND between the rows, and run it. Then check whether it finds the papers you already know. That is a search strategy, and it took less time than the browsing session it replaces.

Once the strategy is set and screening is done, the work changes shape entirely. You stop looking for papers and start working out what they mean together, which is where grouping them into themes and synthesising them begins.