03 · What You Need to Know
Think in Terms of Source Records Rather Than PDF Files
A PDF is only one representation of a scholarly source. It contains the article or other document you read, but it does not necessarily provide the organizational structure needed for a literature review.
A useful source record can connect several elements:
- bibliographic metadata;
- a DOI or another persistent identifier;
- the PDF or other full-text attachment;
- PDF annotations;
- paper-level notes;
- tags or collections;
- links to literature-matrix entries or synthesis notes; and
- the citation information used when writing.
Zotero, for example, distinguishes regular bibliographic items from attachments and notes. Its documentation recommends storing a standalone PDF under a regular parent item containing the bibliographic metadata rather than treating the attachment itself as the complete bibliographic record.
That distinction provides a useful general model even if you use another system:
The paper is the scholarly source. The PDF is an attachment to your record of that source.
Why a Folder of PDFs Is Not Yet a Literature-Management System
Folders are useful for storing files. They become less effective when you expect them to represent every relationship among your literature.
A paper may concern several themes. It may use one method, contribute to two arguments, employ a particular theory, and provide evidence for several parts of your review. A conventional folder hierarchy usually asks you to choose one physical location.
You can create duplicate PDFs, but then you have several copies to maintain. Annotations may appear in one copy but not another. Renaming or replacing one file does not update the others.
A source-centered system avoids requiring physical duplication simply because a paper has several analytical roles. One record can be connected to multiple collections, tags, themes, or notes.
Start With Reliable Bibliographic Metadata
Your source record is only as useful as its identity.
Check important metadata such as:
- author names;
- title;
- publication year;
- journal or other source title;
- volume and issue when applicable;
- page range or article number;
- DOI or other persistent identifier; and
- source type.
Reference managers can retrieve metadata automatically, but automatic retrieval should not be treated as infallible. Verify records that you expect to cite, particularly when metadata appear incomplete or unusual.
This bibliographic record gives the rest of your research materials a stable parent. It also prevents the filename from becoming the primary way you identify the source.
Attach the PDF to the Bibliographic Record
When full text is available to you, associate the PDF with the correct bibliographic item.
In Zotero, attachments can be stored copies or linked files. Stored files are copied into Zotero's data directory, while linked files remain in their original location and Zotero stores a link to them. Zotero notes that linked files are not synchronized by Zotero file syncing, which is an important practical difference when choosing between the approaches.
Stored attachment
The reference manager manages a copy of the file as an attachment to the source record and, depending on the software and configuration, may support file synchronization.
Linked attachment
The source record points to a file stored elsewhere, so moving or renaming that external file may break the connection and synchronization may require a separate strategy.
Neither approach is universally correct. The important point is to know which model your system uses so that reorganizing folders does not unexpectedly disconnect your records from their files.
Do Not Let the Filename Carry the Entire Identity of the Source
A filename such as:
article.pdf
is obviously weak.
But even a carefully constructed filename such as:
Garcia_2026_AI_Feedback_Higher_Education.pdf
should not be your only source identifier.
Filenames can change. Two papers may have similar author-year combinations. Titles can be shortened inconsistently. Supplementary files may belong to the same publication.
Use bibliographic metadata and persistent identifiers where available to establish source identity. A readable filename can still make manual browsing easier, but it should support rather than replace the bibliographic record.
Keep PDF Annotations Attached to the Source
Highlights and comments are most useful when they remain connected to the paper from which they came.
Zotero's PDF reader, for example, allows annotations to be added to notes with citations and provides links from those note annotations back to their location in the PDF. This is particularly useful because it preserves a route from your extracted note to the source passage.
The broader principle applies regardless of software:
Annotation → Source → Location
If you extract a highlight into another application, preserve enough source information to reconstruct that relationship.
Paper-Level Notes Should Point to the Same Record
You may want notes that go beyond highlights. These might contain:
- the study purpose;
- methodological details;
- relevant findings;
- limitations;
- important arguments;
- your questions;
- your evaluation; and
- connections to other literature.
Whether these notes live inside your reference manager or elsewhere, they should clearly identify the source.
If you are deciding what information to preserve about each paper, think of the resulting record as part of the same source-centered system rather than an independent summary document.
Separate the Source Record From Your Synthesis Layer
Connecting everything does not mean storing everything in one enormous note.
Your paper-level record answers:
What does this source contain?
Your synthesis layer answers:
What happens when I compare this source with other sources?
A synthesis layer might include a literature matrix, thematic notes, claim-based notes, an evidence map, or a working outline. The same paper may appear in several of these analytical contexts.
The source record should remain stable while the synthesis structure changes.
| Layer |
Main Function |
Connection Needed |
| Bibliographic record |
Identifies the scholarly source |
DOI, metadata, or another stable identifier |
| PDF |
Provides the document you read |
Attached or reliably linked to the bibliographic record |
| Annotations |
Mark important passages in the document |
Connected to the PDF and source location |
| Paper-level notes |
Record what the source contributes |
Connected to the bibliographic record and relevant locators |
| Matrix / synthesis notes |
Compare evidence across sources |
Source identifiers and links back to paper-level evidence |
| Manuscript citations |
Attribute claims in the written review |
Generated or verified against the bibliographic record |
Use Stable Identifiers Between Tools
Many researchers use more than one application. A reference manager may handle citations and PDFs, a spreadsheet may hold a literature matrix, and another application may contain synthesis notes.
This can work well if the systems share a reliable way to identify sources.
Possible identifiers include:
- DOI;
- reference-manager item link;
- citation key;
- another unique record identifier; or
- a consistent author-year identifier when ambiguity is carefully managed.
For example, a literature-matrix row might include the DOI and an item link rather than only the text Garcia et al. (2026). If you later need the original evidence, you have a direct route back to the source record.
Your Matrix Should Not Become a Dead End
Imagine a matrix cell that says:
Students with greater AI literacy were more critical of inaccurate feedback.
That may be useful for comparison. But if the cell contains no source identifier or locator, you may eventually have to search through several papers to determine which study produced the finding.
A well-designed matrix should preserve enough provenance to take you back to the source. This is part of building a literature matrix that helps you think without sacrificing traceability.
For important claims, you may record a page, table, figure, section, or another useful locator alongside the matrix entry or in the linked source note.
Keep Citations Connected to the Records You Actually Used
A reference manager is particularly valuable when the source you read and annotate is the same bibliographic item used to generate the citation in your manuscript.
This reduces several opportunities for mismatch:
- citing a duplicate record with different metadata;
- using the wrong publication year;
- confusing an online-first version with another record;
- citing a similarly titled paper;
- forgetting which version you actually consulted; or
- manually retyping references incorrectly.
Zotero supports word-processor integrations for creating citations and bibliographies from library items. Other reference managers provide comparable functions. The exact interface differs, but the underlying advantage is the same: the citation originates from the same structured source record used elsewhere in the workflow.
Watch for Duplicate Bibliographic Records
A connected workflow can quietly fragment if the same paper enters your reference manager several times.
You might import one copy from a database, another from a publisher page, and a third from a DOI search. Notes become attached to one record, the PDF to another, and the citation used in your manuscript to a third.
Zotero provides duplicate-detection and merging functions for this reason. Its documentation notes that merging duplicate items preserves collections and tags and does not break existing word-processor citations in the way deleting one duplicate and keeping another may.
Watch Out
When you discover duplicate records, do not automatically delete whichever one looks less complete before checking where its notes, attachments, tags, and manuscript citations are connected. Merge or consolidate records carefully so that the source does not become fragmented across your system.
Collections and Tags Can Organize Sources Without Moving the Underlying Record
A paper may belong to several projects, themes, or conceptual categories.
Zotero collections behave more like playlists than ordinary computer folders: an item can belong to multiple collections without creating separate copies of the bibliographic item. Tags provide another way to classify and retrieve sources.
This model is useful for literature reviews because it supports papers that legitimately fit more than one theme.
You can therefore keep one source record while exposing it through several organizational views.
Do Not Confuse Synchronization With Backup
Synchronization is useful when you work across devices, but it should not be your only protection against data loss.
Zotero's documentation explicitly distinguishes syncing from backup and recommends regular, automated backups of the Zotero data directory. This distinction applies more broadly to research workflows.
Synchronization
Keeps versions of your working data coordinated across locations or devices, depending on the service.
Backup
Preserves recoverable copies so that accidental deletion, corruption, or unwanted changes do not become your only available state.
A beautifully connected literature-review system is still fragile if a single storage failure can erase it.
Connections Matter More Than Having Everything in One Application
You do not have to use one application for every stage of research.
For some researchers, an integrated reference manager is sufficient for PDFs, annotations, citations, and basic notes. Others may prefer specialized software for literature matrices, conceptual notes, qualitative coding, or project management.
The problem is not using multiple tools. The problem is losing the relationships among them.
A useful test is whether you can begin anywhere in the system and recover the rest:
From the citation Can you open the correct bibliographic record and locate the paper?
From the PDF Can you identify the correct citation record and your notes?
From a paper-level note Can you return to the source and the relevant passage?
From a matrix entry Can you identify and verify the paper behind the extracted information?
From an idea-level note Can you recover all of the sources contributing to the synthesis?
If those paths work reliably, your tools are functioning as a system rather than a collection of applications.
04 · A Practical Example
What a Connected Paper Looks Like From Download to Draft
Hypothetical Example
A Paper About AI-Generated Feedback
Suppose you find a hypothetical article examining AI-generated feedback in undergraduate writing. The study is relevant to several themes in your literature review. You want to preserve it in a way that remains usable months later.
1. Create or import the source record Store the correct authors, title, publication details, DOI, and other relevant bibliographic metadata in your reference manager. Verify important fields.
2. Connect the PDF Attach or reliably link the full-text PDF to that source record rather than leaving it as an unidentified file in a downloads folder.
3. Read and annotate Highlight important findings, methods, definitions, and qualifications. Add comments where useful while preserving their locations in the document.
4. Create the paper-level note Record the study purpose, method, relevant findings, author interpretations, limitations, and your own observations while keeping those elements distinguishable.
5. Add analytical connections Connect the source to themes such as learning outcomes, student trust, and AI literacy without duplicating the underlying paper record.
6. Add it to the matrix Use the same DOI, citation key, item link, or another stable identifier so the matrix row can be traced back to the source.
7. Develop synthesis notes When the paper contributes to an emerging claim, connect that claim to the specific finding and other papers bearing on the same argument.
8. Cite from the source record When writing the literature review, insert or verify the citation using the same bibliographic record rather than manually recreating the reference from the PDF filename or another secondary list.
Several tools may be involved, but the source retains one identity throughout the process.
The result is a chain:
Citation record → PDF → annotation → paper note → matrix → synthesis → manuscript citation
You can move forward through the workflow and backward whenever verification is needed.