How to Organise Research Papers: A Simple System

A good research-paper system should answer three questions quickly: Have I saved this paper before? What did I think about it? Where does it fit in my research?

You do not need a perfect hierarchy or twenty tags. You need one reliable place for bibliographic records, a consistent reading workflow and enough structure to retrieve a source when you need to write.

This guide uses Zotero as the main example because it is free and designed for collecting, organising, annotating and citing research. The principles work with other reference managers too.

The simplest useful research-paper system

Layer	Purpose	Example
Reference library	Store the citation, PDF and metadata	Zotero
Collections	Group papers by project or broad purpose	Thesis / Chapter 2 / Methods
Tags	Track themes or workflow states	to-read / key-study / qualitative
Paper note	Record what the source contributes	Question, method, finding, limitation
Synthesis tool	Compare sources across a question	Literature review matrix

Do not make one tool do every job. A reference manager is excellent at storing and retrieving sources; a literature review matrix is better for comparing studies; your thesis draft is where the argument belongs.

1. Choose one source of truth

Your citation manager should be the place you check before downloading another copy of a paper. Save the bibliographic record there first, attach the PDF where permitted, and correct obvious metadata errors.

Zotero’s documentation describes items as the core bibliographic records that hold the metadata needed for citing and organising research. Working from records rather than a folder full of anonymous PDFs makes later citation much easier.

2. Use collections like playlists, not filing cabinets

One source can be relevant to several parts of your thesis. A strict folder tree forces you to decide where the “real” copy belongs.

Zotero’s collections guidance makes a useful distinction: the same item can appear in multiple collections without being duplicated. Think of collections as playlists.

A simple PhD structure might be:

PhD
  Inbox
  Core literature
  Methods
  Thesis
    Introduction
    Literature review
    Study 1
    Study 2
    Discussion
  Papers to cite
  Published outputs

Keep it broad. If you need a six-level tree to file a paper, you are building a taxonomy instead of doing research.

3. Use a small tag vocabulary

Tags are most useful when they answer a recurring question.

For example:

• workflow: to-read, skimmed, deep-read, cited;
• importance: key-study, background;
• method: RCT, interview, survey, review;
• theme: only the few themes you repeatedly filter.

Avoid automatically creating dozens of near-duplicates such as “AI”, “artificial-intelligence”, “generative AI”, “GenAI” and “AI-tools” unless those distinctions matter to your work.

4. Decide what happens when you save a paper

Use a repeatable intake process:

1. Save the citation and PDF.
2. Check title, authors, year, journal and DOI.
3. Put the item in the relevant collection.
4. Add one workflow status.
5. Write one sentence explaining why you saved it.
6. Decide: discard, skim, deep read or use now.

That last decision prevents the library becoming a guilt archive.

5. Read papers at different depths

Not every source deserves an hour. Use the three-pass guide to reading journal articles to triage relevance before deep reading.

For a paper you keep, record enough that Future You can understand why it matters:

Why I saved this:
Research question:
Data/sample/material:
Method:
Main finding or argument:
Important limitation:
Relationship to other papers:
How I may use it:
Useful page numbers:
Next action:

Write in your own words. Copying an abstract into a note creates text, not understanding.

6. Separate paper notes from synthesis notes

A paper note asks: “What does this source say?”

A synthesis note asks: “What does the literature say about this issue?”

That distinction matters. If five paper notes remain isolated, you still have to do the intellectual work of comparing them when you write.

After reading several related sources, transfer the comparable elements into a literature review matrix or thematic synthesis note. Our guide to writing a literature review shows how to turn those comparisons into paragraphs.

7. Give files boring names

If your reference manager handles attachments reliably, you may rarely need to browse the PDF directory. If you do keep files separately, use a predictable filename such as:

2026_Chen-Smith_Generative-AI-Feedback.pdf

A boring naming system is better than clever abbreviations you will forget.

Avoid:

• download.pdf
• article (17).pdf
• FINAL_new_reallyfinal.pdf

8. Let search do more work than folders

A modern research library is searchable by title, author, tag, note and often PDF text. Use that power. Collections give you orientation; search retrieves specifics.

Zotero also supports saved searches that update automatically when items match specified criteria. A saved search for “to-read” material can be more useful than manually moving papers between folders.

9. Keep an inbox—but empty it

An Inbox collection gives you a frictionless place to save promising material during searching. The rule is that Inbox is temporary.

Once or twice a week:

• remove duplicates;
• correct metadata;
• discard irrelevant sources;
• assign useful papers to collections;
• decide what deserves reading.

If the Inbox contains 800 papers, it is no longer an inbox. It is the library.

10. Build retrieval into your reading notes

When you record a useful claim, add the page or section. When you note a figure, record its number. When you paraphrase an argument, make it clear which source the note belongs to.

This reduces the awful late-stage experience of remembering that “one paper said this somewhere” and searching through forty PDFs.

A worked workflow

Imagine you find a paper while searching for evidence about generative-AI feedback.

1. Save it to your reference manager.
2. Check the metadata.
3. Add it to PhD → Literature review.
4. Tag it to-read and AI-feedback.
5. Write: “Potential evidence on whether automated feedback changes revision behaviour.”
6. Run a quick relevance pass.
7. If useful, deep-read and replace to-read with deep-read.
8. Create a structured paper note.
9. Add the comparable finding to your literature-review matrix.
10. When cited, add cited.

The important part is the flow from discovery to synthesis, not the particular software.

Common organisation mistakes

• Downloading PDFs before checking whether you already have them.
• Creating a new folder for every idea.
• Using hundreds of inconsistent tags.
• Highlighting papers without leaving a summary.
• Taking notes without page numbers.
• Maintaining the same information manually in three apps.
• Saving everything “for later” without a triage decision.
• Spending more time redesigning the system than using it.

A 20-minute research-library reset

1. Choose the reference manager that will be your source of truth.
2. Create an Inbox, Core Literature, Methods and Thesis collection.
3. Choose four workflow tags.
4. Process the newest ten unfiled papers.
5. Delete obvious duplicates.
6. Write a one-sentence purpose note for the papers you keep.

Then return to research. Organisation is infrastructure, not the project itself.

If your digital workspace as a whole has become noisy, the guide to reducing digital clutter for learning can help simplify files and study tools without rebuilding everything from scratch.