How to write a literature review: a step-by-step guide

Most people learn to write a literature review by doing one badly first. You read fifty papers, summarise each in a paragraph, staple them together, and call it a review. Your supervisor hands it back covered in red ink: this is a summary, not a synthesis. So let's skip that first painful draft and do it properly.

A literature review maps what is already known about a question, where the field agrees, where it argues, and what nobody has answered yet. The point isn't to prove you read a lot. It's to show you understand the conversation well enough to add to it.

Step 1: Narrow the question until it hurts

A vague question gives you ten thousand papers and no way to choose between them. "Social media and mental health" is a topic, not a question. "Is daily Instagram use associated with anxiety symptoms in undergraduates?" is a question you can actually answer. Add the population, the variable, and the outcome. If your search returns millions of results, your question is still too wide.

Step 2: Search like you mean it

One search box and one query won't cut it. Build a small set of search terms, including synonyms and the words other fields use for the same idea, and run them across more than one database. Keep a record of what you searched and when, because you'll need it later and you will not remember.

• List your core concepts, then brainstorm synonyms for each (adolescent, teenager, youth).
• Combine concepts deliberately rather than typing one long sentence.
• Search more than one source so you don't inherit a single database's blind spots.
• Save every query and the date you ran it.
• Track down the studies cited by the best papers you find, and the studies that later cited them.

Step 3: Screen in two passes

First pass, titles and abstracts. Be quick and slightly ruthless. If a paper clearly doesn't fit your question, drop it. Second pass, read the full text of what survived and decide for real. Write down why you excluded the borderline ones. When a reviewer asks how you chose, that note is your answer.

Step 4: Take notes you can actually use

Reading without a system means re-reading everything at writing time. Use a synthesis matrix instead: a simple table with one row per study and columns for the things you care about, such as sample, method, key finding, and limitations. Fill it as you read. By the end you can see patterns down each column without opening a single PDF again.

Step 5: Organise by idea, not by paper

Here's the move that turns a summary into a review. Don't write one paragraph per study. Group studies by theme, by method, or by where they disagree, then write about each group. "Three lines of evidence support X, though two recent trials complicate it" reads like an expert. "Smith found X. Then Jones found Y. Then Lee found Z" reads like a reading list.

Step 6: Write the draft

Open with the question and why it matters. Walk through the themes you found. Be honest about conflicting results instead of averaging them into a fake consensus. End at the gap, the thing the literature hasn't settled, because that gap is usually where your own work begins. A good review doesn't just describe the field. It points somewhere.

Mistakes that cost you marks

• Summarising each paper in turn instead of synthesising across them.
• Citing only the studies that agree with you.
• Leaning on old reviews instead of reading recent primary research.
• Hiding disagreement between studies to make the story tidy.
• Forgetting to name the gap your review is pointing toward.

None of this is hard. It's just slow, which is why the screening and note-taking are the first things people cut when a deadline looms. That's also where an AI assistant earns its keep: it can run the searches, do the first screening pass, and pull structured notes into a matrix, so your time goes to the thinking instead of the spreadsheet.