How to find sources for a literature review fast

Most slow literature searches are slow for one reason: there is no method, only a search box and a growing pile of open tabs. The same question gets typed six different ways, the same papers keep turning up, and nothing records what has been tried.

A search that is planned takes less time and finds more. What follows is a literature search strategy in the plain sense of the word: a written order of operations you can repeat, hand to a supervisor, and later describe in a methods section.

The order below is the one that saves the most effort: the question, the search string, a review article, the chain of citations, the right databases, the full text, the log. Two jobs along the way are ones EdCitation does free and without an account: Find sources looks for published work by topic or by the claim a single sentence needs, and Cite a source turns each paper you keep into a reference from its DOI or title.

How do I turn my topic into a question I can search?

Write the topic as one question, then split the question into its two to four main concepts. A topic such as "nurse staffing" cannot be searched well, because it has no edges. A question can: Does the number of nurses on a hospital ward affect patient deaths?

That question has three concepts: nurse staffing, mortality, hospitals. Under each, list the other words an author might have used.

Keep that table open. Every review article you read adds a word to it, and the words you add in week two are the ones that find the papers you missed in week one.

How do I build a search string with Boolean operators?

Join the words for one concept with OR, join the concepts with AND, and put phrases in quotation marks. Most academic databases read these operators.

MIT Libraries (n.d.) gives the rule that catches most people out: when a search mixes AND with OR, the words to be joined by OR have to sit inside brackets, or the database will read the string in an order you did not intend.

The worked example becomes:

("nurse staffing" OR "nurse-to-patient ratio" OR "nursing workload") AND (mortality OR death OR "failure to rescue") AND (hospital* OR inpatient*)

If a search returns thousands of results, add a concept with AND or limit the years. If it returns a handful, remove a concept or add synonyms with OR.

Where the syntax differs, and why the help page is worth ten minutes

Databases are not consistent with each other, and a string that works in one can fail silently in another. The National Library of Medicine (n.d.) documents three PubMed behaviours that between them account for a great many wasted afternoons: Boolean operators have to be typed in capitals, a search is processed from left to right unless brackets group the terms, and a word must have at least four characters before the asterisk, so nurs* is accepted but car* is not.

PubMed also rewrites your search behind the scenes. Its automatic term mapping expands a plain word into the matching subject heading and its synonyms, which usually helps and occasionally buries the thing you wanted. Putting a phrase in quotation marks or adding a field tag turns that off. Every large database has an equivalent set of quirks, and they are all written down on a help page that takes ten minutes to read once.

Why should I start from a recent review article?

A recent review article has already done a version of your search, so it gives you the key papers, the vocabulary and the main disagreements in one place. Find one before reading anything else.

In PubMed, the Review and Systematic Review filters sit beside the results. In other databases, add the word review to the search, or filter by document type. Then read the review's introduction for the words the field uses, and add any you missed to your table of synonyms.

Read the studies a review cites before you cite them yourself. A review's summary of a paper is a paraphrase, and paraphrases drift.

How does citation chaining work?

Citation chaining means following citations in both directions from one good paper. The University of Oregon Libraries (n.d.) describes it as taking one relevant source and mining its reference list for more; backwards chaining finds the earlier work it was built on, and forwards chaining finds everyone who has cited it since.

1. Pick the two or three most relevant papers you have, including the review.
2. Backwards: scan each reference list, and mark titles that match your question.
3. Forwards: find each paper in Google Scholar and click "Cited by". Scopus and Web of Science offer the same from a curated index. PubMed shows a "Cited by" list too, and says plainly that it may not be complete.
4. Use "Related articles" in Google Scholar or "Similar articles" in PubMed for papers on the same subject that do not cite each other. For work that makes the same point without citing it, type the paper's central claim, in a sentence, into EdCitation's Find sources: it searches by the claim, not only by the words of a topic.
5. Repeat from the best new papers, until the same titles keep coming back.

Citation chaining is the fastest part of the method and the easiest to do badly. Mark why you kept each paper as you go, or you will be holding forty PDFs and no argument. One way to do both at once: cite each keeper the moment you keep it, by pasting its DOI into Cite a source, and hold the results in a reference list in the dashboard of a free EdCitation account, where each reference can carry a tag saying why it earned its place ("method", "counter-evidence", "UK data") and the list can later be filtered by tag.

Which database should I search?

Search one database built for your subject and one that is broad, because no index holds everything. The figures below come from each service's own pages and data, read on 22 September 2026.

Your library's subject guide will name the specialist databases for your field. Use them: a curated subject index is where the search string earns its keep.

Why Google Scholar should not be your only search

Google Scholar is the fastest way to chain citations and the worst way to run a systematic search, and that is not a matter of opinion. Gusenbauer and Haddaway (2020) tested 28 academic search systems, including Google Scholar, PubMed and Web of Science, on how well each one lets a searcher run and reproduce a Boolean query. They found large differences in precision, recall and reproducibility, concluded that only about half the systems can be recommended for evidence synthesis without serious caveats, and set out why Google Scholar is inappropriate as a principal search system.

For a taught literature review that matters less than it does for a systematic review, and Google Scholar remains excellent for forwards chaining. The rule that follows is simple: find with Scholar, but search in something whose results you can reproduce next month.

How do I get the full text legally?

Try the open copy first, then your library, then the author.

• Filter for it at the search. Much of the chasing can be avoided before it starts: the open-access filter in EdCitation's Find sources limits the results to works marked as openly available, which is worth switching on when you need papers you can read tonight and your library login is not to hand.
• Unpaywall is a free browser extension that looks for a legal open copy of the article you are viewing. Unpaywall (n.d.) says it harvests from more than 50,000 journals and open-access repositories, and only from legal sources: repositories run by universities, governments and scholarly societies, and open copies hosted by publishers themselves.
• PubMed Central is a free full-text archive of biomedical and life sciences articles. The National Center for Biotechnology Information (n.d.) reports 11,327,616 articles in the archive.
• Repositories. University repositories hold accepted manuscripts, and preprint servers such as arXiv hold earlier versions.
• Your library. Log in through the library's website so that subscriptions are recognised. If the library does not hold the article, it can usually request a copy from another library by interlibrary loan.
• The author. Email the corresponding author and ask for a copy.

How do I keep a search log, and when do I stop?

Record every search as you run it, in a plain table like the one below, where the row is an invented example. The log stops you repeating work and becomes the search paragraph of your methods section.

What belongs in that log is not a matter of taste either. Rethlefsen et al. (2021) published PRISMA-S, a 16-item checklist for reporting literature searches, alongside the PRISMA statement. Its items are a good specification for the log even when you are not writing a systematic review: name each database and the platform you reached it through, give the full strategy for each one rather than a tidied version, record the date each search was run, say what limits and filters were applied, and note the other methods used, such as citation chaining and contacting authors.

PubMed's Advanced Search keeps a history of the last 100 searches, and a free My NCBI account will email you when new papers match a saved search. That is a convenience, not a log: the history is capped and the wording of your strategy is not preserved anywhere except where you write it down.

Stop when the search stops teaching you anything: new strings and new chains return papers you already have. Then set an alert, and start writing. How many of those sources belong in the paper is answered in how many references you need.

Is there a tool that searches by the claim I need to support?

Yes. EdCitation's Find sources is the best tool for the stage most searches fail at, the late one, when the structure is written and single sentences still lack a source. The databases above want a Boolean string, with quirks of their own, such as the capitals and four-letter truncation rule PubMed documents; a sentence waiting for its source does not split neatly into concepts. Find sources takes the sentence itself, the claim it needs to support, and searches about 300 million published works for it, with filters for kind of source, years and open access. Free, no account.

Why a looked-up reference beats a written one

Everything EdCitation returns is looked up in the published record, and nothing is written for you: it never writes any part of a review, and it never writes a reference. That is the difference from asking a chatbot for sources, which composes references and cannot look each one up the way this does. Cite a source builds each reference from the publisher's record, the same kind of record the Crossref row above counts at almost 188 million, and screens the source for retraction. In APA 7 and Vancouver it sets a title recorded in title case in sentence case, as both styles ask, and keeps names and initialisms; glance at any name made of ordinary words. Verify references then checks the finished list, marking each entry verified, "check this" or not found.

What is free and what is paid

The three tools in this section need neither payment nor an account. The Pro plan, at $8 a month, includes References from a file: give it the review as a document and it checks each reference and ties each citation in the text to its entry, which catches a paper cited in one section and never added to the list. The Max plan, at $24 a month, includes the Library, where textbooks and scholarly books can be searched by discipline and year among other things, for reviews that lean on monographs as much as on articles. If yours starts as an annotated bibliography, see how to write one.

Quick questions

How many databases should I search for a literature review?

For a thesis chapter or a paper's introduction, two or three is usual: the main database for your subject, one broad index such as Scopus, Web of Science or OpenAlex, and Google Scholar for citation chaining. A systematic review follows its protocol, which names every database in advance.

Is Google Scholar enough for a literature review?

No, not on its own. Gusenbauer and Haddaway (2020) found Google Scholar inappropriate as a principal search system for evidence synthesis, and Google publishes neither a source list nor an index size. Pair it with a curated database for your subject.

Is Unpaywall legal?

Yes. Unpaywall states that it harvests only from legal sources, including repositories run by universities, governments and scholarly societies, and open copies hosted by publishers themselves.

How recent do my sources need to be?

There is no general cut-off. Greenbaum (2021) points out that APA Style has no timeliness requirement at all, though an instructor may set one. Follow any limit in your own brief, and remember that fields age at different speeds.

What should I record about each search?

At least the database, the platform, the full search string, the filters, the date and the number of results kept. Those are among the items PRISMA-S asks for, and they are what a supervisor or a reviewer will ask you to reproduce.

What if one sentence in my review still has no source?

Put the claim, worded as the sentence states it, into EdCitation's free Find sources, and read what comes back before citing any of it. If nothing in the published record supports the claim as written, change the sentence rather than stretch a source to fit it.

References

• Clarivate. (n.d.). Web of Science Core Collection. https://clarivate.com/academia-government/scientific-and-academic-research/research-discovery-and-referencing/web-of-science/web-of-science-core-collection/
• Crossref. (2026). Crossref status. https://www.crossref.org/06members/53status.html
• Elsevier. (n.d.). Scopus content. https://www.elsevier.com/products/scopus/content
• Google. (n.d.). About Google Scholar. https://scholar.google.com/intl/en/scholar/about.html
• Greenbaum, H. (2021, October 18). The "outdated sources" myth. APA Style Blog. https://apastyle.apa.org/blog/outdated-sources-myth
• Gusenbauer, M., & Haddaway, N. R. (2020). Which academic search systems are suitable for systematic reviews or meta-analyses? Evaluating retrieval qualities of Google Scholar, PubMed, and 26 other resources. Research Synthesis Methods, 11(2), 181–217. https://doi.org/10.1002/jrsm.1378
• MIT Libraries. (n.d.). Database search tips: Boolean operators. https://libguides.mit.edu/c.php?g=175963&p=1158594
• National Center for Biotechnology Information. (n.d.). About PMC. https://pmc.ncbi.nlm.nih.gov/about/intro/
• National Library of Medicine. (n.d.). PubMed user guide. https://pubmed.ncbi.nlm.nih.gov/help/
• OpenAlex. (2026). Works [Data set]. https://api.openalex.org/works
• Rethlefsen, M. L., Kirtley, S., Waffenschmidt, S., Ayala, A. P., Moher, D., Page, M. J., & Koffel, J. B. (2021). PRISMA-S: An extension to the PRISMA statement for reporting literature searches in systematic reviews. Systematic Reviews, 10(1), Article 39. https://doi.org/10.1186/s13643-020-01542-z
• University of Oregon Libraries. (n.d.). How to write a literature review: Citation chaining. https://researchguides.uoregon.edu/litreview/citation-chaining
• Unpaywall. (n.d.). Frequently asked questions. https://unpaywall.org/faq