How do you write a good related work section?

Organise it by idea, not by paper. Group prior work into three to five themes that matter for your contribution, say what each group established and where it falls short, and end each theme by showing how your work differs. Compare papers with each other rather than summarising them one by one, and cite claims you have checked on the page, not from memory or a summary.

By Malek Khannoussi · Updated25 September 2026

1. Find the themes before you write

Read your notes, not the papers, and sort them into groups by the question each paper answers or the approach it takes. Three to five themes usually cover a contribution. Papers can sit in more than one.

2. Write each theme as an argument

Open with what the group of papers established.
Compare them: where they agree, differ, and why.
Close with the limitation that your work addresses.

3. Check every claim against its source

“X showed Y” is a factual claim. Open the passage and confirm it says Y, in the conditions you imply. Notes that link to their passage make this fast; summaries from an AI make it essential, since ahallucination can put a claim in the wrong paper.

4. Don't miss the obvious paper

Papers that many of your sources cite, their co-cited core, are the ones reviewers expect. Check the references of your key papers and the papers citing them.

Common mistakes

One paragraph per paper, in chronological order.
Describing papers without saying how they relate to yours.
Citing a paper for a claim it doesn't make.

Questions people ask

How long should a related work section be?

In a conference paper, often half a page to a page and a half. In a thesis, a full chapter. Length matters less than whether it sets up your contribution.

Should related work come before or after the method?

Both are common. Early placement frames the problem; late placement, common in computer science, lets you compare with your approach once readers know it.

How many papers should I cite?

Enough to cover each theme fairly, including the strongest competing approaches. Missing an obvious key paper hurts more than citing too few minor ones.

Can I use AI to write it?

Check your venue’s policy and disclose use. Models misattribute claims easily, so every sentence about a paper should be checked against that paper.