Most essays do not fall apart because the argument was weak. They fall apart because the evidence was thin: three websites, one textbook from 2009, and a news story quoting a report nobody has read. The answer is not to search harder. It is to search in better places, and to know what you are looking at once you find something.

Your university library already pays for access to research you cannot reach through an ordinary search engine. Learning to use those subscriptions properly takes an afternoon. It will save you weeks.

Turn the title into a searchable question

Before you type anything into a search box, strip the essay title back to its concepts. Take something like Evaluate the role of social media in shaping political participation among young voters. There are four ideas in there: social media, political participation, young people, voting. Each one is a search term, and each one has alternatives.

Write them out on a single line, with the synonyms you would accept in place of each:

• Social media: "social networking sites", platform, Twitter/X, TikTok
• Political participation: civic engagement, voter turnout, activism
• Young people: youth, young adults, under-25s, generation Z

This list does two jobs. It gives you search strings that actually work, and it stops you from writing an essay that only answers one narrow version of the question.

Know what your library databases hold

Most UK university libraries give you two routes in. The first is the discovery layer — the big search box on the library homepage, usually branded something like Library Search, Primo or Discover. It sweeps across most of the collection at once, which makes it useful for a scoping search: you want to know whether there is any literature on your topic at all.

The second route is subject-specific databases. These are narrower and far more precise. Depending on your discipline, you will meet JSTOR, Project MUSE, Scopus, Web of Science, ProQuest Central, EBSCO's subject collections, PsycINFO for psychology, Historical Abstracts for history, and the MLA International Bibliography for literature. They offer filters a general search cannot match: peer-reviewed only, publication date range, document type, language, and subject terms.

One caution about that peer-reviewed tick box. It is a useful filter but not a guarantee of quality. Journals that are peer-reviewed still publish book reviews, editorials, corrections and letters, and none of those have been through the same process as a research article. A study can also be peer-reviewed and still be a poor fit for your argument.

Boolean searching without the jargon

The operators below are supported by nearly every database, and they are the difference between 40,000 results and 40. Capitalise them.

• AND narrows: social media AND voting returns only items containing both.
• OR widens, and belongs inside brackets with synonyms: ("social media" OR "social networking").
• Quotation marks keep a phrase together, so "political participation" does not split into two unrelated words.
• Asterisk catches word endings: vot* finds vote, votes, voting, voter.
• NOT removes a concept, but use it sparingly — it can delete the article you needed.

A workable string might read: ("social media" OR "social networking") AND ("political participation" OR "voter turnout") AND (youth OR "young people"). Start broad, then add a concept each time the results look manageable.

When you find one genuinely good article, open its record and look at the subject terms the database has assigned to it. Those are controlled vocabulary, and clicking one will take you to everything else filed the same way — often the fastest route into a field.

Google Scholar is a supplement, not a substitute

Scholar is excellent for two things: finding a known article by title, and moving forwards or backwards through a citation trail. Its limitations matter, though. Coverage is uneven, results are ordered by an algorithm nobody outside the company can inspect, and a great deal of what it returns is not peer-reviewed.

A few habits make it more reliable. Set up library links in Scholar's settings so that "Find it at [your university]" appears next to results. Use Cited by to see who has built on a key article since it was published, and Related articles when you have one good hit and want its neighbours. Check the version you are reading: many hits are author manuscripts rather than the final published article, and page numbers can differ. If a PDF is sitting on an unfamiliar website, track down the original publication before you cite it. Preprints in particular have not been peer-reviewed, so treat them as a signal of where a field is heading, not as settled evidence.

Five questions before you trust a source

1. Who wrote it? Check the affiliation and whether the author has published elsewhere in this area. A name you have never seen is fine; a name with no connection to the field is worth a pause.
2. When was it published? In fast-moving areas such as computer science, a decade is a long time. In philosophy or literary criticism, a 1970s monograph may still be central.
3. Where was it published? A peer-reviewed journal, a university press, an edited collection, a government body, a think tank. A think tank report can be genuinely useful, but it usually has a position to defend.
4. What is the evidence? Data, primary documents, a stated method — or assertion and anecdote? Read the methods section, not just the abstract.
5. Why does it exist? To inform, to persuade, to sell? Look for who funded the research.

A source is not trustworthy because it is old, or because it is new. It is trustworthy because it is the right kind of source for the specific claim you are making.

Read past the abstract

Abstracts are written to get the article read. They compress, round off and occasionally overstate. Before you build a paragraph on a paper, read the results and the limitations. If the sample is thirty undergraduates at one institution, do not describe the findings as a pattern among young people generally. If the article is about television news and you are writing about TikTok, it is not your evidence, however appealing the title.

Two more habits repay you here. Work backwards through the reference list of a strong article, and forwards through its citations. One good source can hand you a dozen better ones, already arranged around your topic.

Record everything from the first search

Save the full citation before you close the tab, and note the database you used and the date you accessed it — some Harvard variants require an access date for certain source types. Add a two-sentence summary in your own words, plus page numbers for anything you might quote directly. A reference manager such as Zotero, Mendeley or EndNote will capture most of this automatically, but check the imported metadata against the article's first page: missing issue numbers and oddly capitalised titles are common.

A ten-minute routine before you cite anything

1. Open the full text and confirm the author, journal, year and volume against your saved record.
2. Skim the introduction and conclusion to check the claim you want to make is genuinely in there.
3. Judge the date against the conventions of your subject.
4. If it is a preprint, a blog post or a news item, find the underlying study and cite that instead.
5. Save the citation in the format your department requires, before it disappears into a browser tab.

Do this for the eight or ten sources that carry your argument, and the essay writes itself with far less anxiety. You will know precisely what you are standing on — and that is most of the work.

Photo: Polina Zimmerman / Pexels