Skip to content
Home » The Effortless Blog » Finding the most important papers for your literature review

Finding the most important papers for your literature review

During a typical literature review, you will run across 100s of papers. How do you identify what is essential? This article presents four tools and methods to identify relevant literature and save hours by skipping secondary literature. The most modern method utilizes AI to scan abstracts of 100s papers and curate a comprehensive, goal-oriented reading list. Classical methods, however, have their strengths in other domains.

Join the Effortless Newsletter. Receive free tips on note-taking, literature review, AI tools, and other productivity topics specifically tailored for academics and students.

An efficient literature review is, most importantly, goal-oriented. You want to find the most relevant information on a topic and get the scientific opinion on various research avenues related to this topic quickly. The problem is that there are often 100s of papers to choose from. Not long ago, I received this email:

In my literature review webinars, I always urge folks not to collect papers (Collector’s fallacy) but to instead follow a red thread collecting the “intent to collect” rather than the paper. But sometimes, even with the most conservative approach, we still end up with too many relevant papers. Here are four methods to dig through giant collections to identify what is relevant:

  1. Scite Plugin with Zotero to identify controversial and well cited papers
  2. Using Litmaps’ graph feature to visualize a paper’s importance
  3. Using AI to scan through abstracts and asses importance
  4. Identifying key authors or journals to follow them

Let’s get started:

Using Scite and Zotero to identify relevant papers

The easiest way to identify relevant papers in a collection is to look at metrics: number of citations, number of references, and date. If you are using Zotero, you can install the scite plugin, which will help you break down the number of citations into mentioning, supporting, and contrasting citations. This helps identify controversial papers. The year column enables you to determine the most recent papers.

Scite plugin in Zotero installed.

However, this method is very simple, and papers with the most citations are not always the best papers. More problematically, citations always correlate with the age of the paper. An impactful paper is, therefore, recent and has many citations—this is not easy to see immediately. This is where the following method can help: Litmaps.

Using Litmaps to see relevant papers in seconds

Litmaps is at the core of all my literature review workflows because it has a unique graphical feature, making it the most intuitive way to discover relevant literature. First, upload your papers from Zotero (or any other reference manager) to Litmaps. First, you export your collection to BibTeX (every reference manager can do this). The next step is to upload the BibTeX file to a Litmaps collection:

(As of March 2024 Litmaps released a new UI that is currently being tested. If you are reading this at a later date keep in mind that the UI might have changed.)

Next, we are going to click “View as a Map.” This will lay out the papers graphically. Each axis has a meaning (you can configure it as well). The default is:

  • X-axis: Date (old on the left to new on the right)
  • Y-axis: Number of citations (many at the top, few at the bottom)
  • Size of bubbles: Number of references (the more references the more likely it is a review paper)

You can drag and rearrange the map to get a better overview as well:

This helps you to see immediately what is essential. Important papers will be at the top right. They are recent and have many citations. Papers in big bubbles will be review papers, as review papers always have many more references than others. It is not unusual for a reference paper to cite 300+ studies, but it is relatively rare for a method or a results paper.

Now, you can go ahead and cherry-pick the most interesting papers at the top right and discard those on the bottom left. While great in practice, you can still miss work that is relevant only to a small audience, which may be very relevant to you. In New Zealand, for example, Mātauranga Māori describes the body of indigenous knowledge around ecological topics. This is very relevant to New Zealand ecology but less applicable outside of New Zealand. Consequentially these papers will get very few national citations and you might miss them in your review, despite their inherent relevance. Using your intuition is the only way around it.

Using AI to identify relevant literature

AI is becoming increasingly powerful and can save you hours on the literature review. While ChatGPT can read a paper in milliseconds it does not understand it deeply. You can leverage this by letting it read over 100s of papers (or their abstracts) and ask it to sort them semantically (i.e., by content). This allows us to create a reading list that is organized by meaning not citations. This workflow requires you to have all your papers in Zotero or any other reference manager.

First, you need to export your collection from Zotero to a BibTeX file. This is a text file containing the title, abstract, year, and authors. Select “export notes” to export your own notes, too. Here is how it looks in Zotero:

Next, you are going to rename the file from *.bib to *txt. You will notice that you can now open it with any text editor. Double-check that you can see the abstracts included in the file. This will be essential for the next step. If the abstracts are not included, AI will only have access to the paper’s title to decide its contents. Continue with this workflow only if most abstracts are included. This is how this looks:

Now, we are going to upload this file to ChatGPT and ask it to create a reading list. This works because the abstracts of the papers are included in the BibTeX file, and ChatGPT has some context for each paper. It can identify closely related papers and is possibly redundant when generating the reading list. The result is not only a list of documents but also a list of learning goals. This allows you to read the paper with a question in mind and significantly improve your learning.

Here is the prompt I used in this video:

My goal is to learn how {insert your goal here}.
1. Analyze the papers mentioned in the document. Use only them and nothing else. 
2. Create a reading list that takes me from novice to advanced. Starting with broader papers and ending with very specific papers.
3. For each paper identify 2 questions as my "learning goal" based on what you infer in the abstract. 
Choose only 5 papers from the collection, as I do not have more time

If you want to improve this prompt and get better and more reliable results there are a few things you can do:

  • Define a clear output format for your AI using XML and Markdown (Mentioned in this tweet)
  • Describe the input format and which parts are a must-read for the AI (it might otherwise just skip reading).
  • Use a more emotional language, capitalize. Yes, ChatGPT reacts to emotions.
  • Prompt multiple times with a slightly changed goal description to see if there is consensus on the output.

You can learn in-depth techniques on dealing with ChatGPT for academics in the Effortless AI course. My webinar on AI literature review will cover this specific use case in-depth, as well.

Following journals and authors to identify

If you are starting your literature review, following an author is the easiest way to build understanding quickly. Especially one that is closely related to what you are doing. To identify your top authors, sort your Zotero collection by creator and notice recurring authors. Depending on the domain, the order of authors has different meanings. It may be alphabetical or may signify the importance.

Once you have found the relevant authors, use Google Scholar to check what else they have published and whether this aligns with your interests. You can also check the number of citations and decide how “senior” an author is. However, the number of citations is highly domain-dependent, and even related domains may have different ranges for typical citation counts. If you install the scite browser extension small boxes below each paper will tell you how many supporting or contrasting mentions each of their papers got.

Similarly to authors, you can look at the top journals in your collection and follow those. However, journals are often very broad and may not help as well as authors to narrow down what is important.

Summary: Identifying key papers in your literature review

There are many ways to identify relevant papers. Using AI will likely be how we conduct literature reviews in the near future because AI can read abstracts or entire papers and build an understanding of the subject matter to help guide your attention to the most important literature. Until then we can use citation based metrics to identify relevant papers. Litmaps is the best option to do so. A more organic option for your literature review is to follow the paper trail of single authors related to your field.

All techniques presented here are part of my upcoming webinar on Literature Review and Academic Writing with AI:

Workflow for the Literature Review webianr

Leverage semantic and citation search with AI to find the most impactful literature quickly. Uncover reference gaps combining multiple tools. Use ChatGPT assistants to get rid of hallucinations and use AI to aid faithfully and ethically in your lit review and writing process.