In this article, I'll provide an overview of an efficient literature management and information organization system that combines Zotero, Obsidian, and NotebookLM as a research workflow for the AI era.
By integrating these tools for academic writing and research presentation preparation, I believe you can create outputs that leverage LLMs while maintaining proper source management. I personally use this system, so if you're a fan of these tools, I encourage you to give it a try. If you have other methods you'd recommend, I'd be happy if you could share them in the comments.
For a video demonstration showing the specific screens and workflow, check out (dubbed version available):
Overall Workflow Overview
The workflow I'm introducing today follows this sequence: first, collect, manage, and categorize literature using Zotero; next, integrate notes and organize information using AI in Obsidian; finally, perform cross-cutting information extraction and supplementation with NotebookLM. The key is to leverage each tool's strengths while progressively refining and developing information through stages.
Building the Foundation with Zotero
Basic Literature Management Principles
Zotero serves as a literature management software responsible for accurate source information management and classification. Simply drag and drop PDFs to automatically retrieve literature information, and book information can be easily imported from pages like Amazon. For research-level work where citations are crucial, Zotero as a specialized software is overwhelmingly more convenient than Notion or Obsidian.
For Zotero setup methods and basics, please refer to this article
Other literature management software can generally serve as alternatives. You can check out other software options in this article:
Efficient Organization Methods
Literature in Zotero is primarily classified using the Collections and Tags features. This makes it easier to use when outputting to Obsidian or uploading to NotebookLM. Most other software has similar features, so this approach should be applicable.
For collection management, I strongly recommend project-based management. By classifying according to specific projects like "For ○○ Conference Presentation" or "For △△ Journal Submission," you can create organization that's tied to experience and easier to remember. From experience, overly detailed classifications by disease or theme should be avoided as they can make it unclear where things are located.
For tags, I recommend applying them with how you'll use them later in mind. For more detailed discussion on tag usage and differences from collections, please see this article: https://lab.nounai-librarian.com/tag/
Note-Taking Tips
Previously, I used to add summaries to Zotero notes, but now that large language models make summarization easy, I recommend keeping content to the essential minimum. In my case, I highlight important passages that I'll cite in papers and select "Add to Note" from the right-click menu to efficiently accumulate notes. Rather than detailed summaries, focusing on information necessary for output creates practically usable notes for later work.
▲ Right-clicking highlighted text allows you to add it to notes
Integration with Obsidian
For transferring notes from Zotero to Obsidian, the Better Notes plugin is convenient. After installing the plugin, you can output in markdown format using the "Export Current Note" function, and enabling Auto sync allows bidirectional synchronization to maintain the same content regardless of which tool you edit in.
For more detailed information about plugins, see:
While bidirectional synchronization can sometimes create conflicts about which updates to prioritize when using Zotero and Obsidian simultaneously, making the behavior complex, this is relatively simple among Obsidian integrations, so I use this approach.
Information Integration in Obsidian
AI-Powered Information Summarization, Organization, and Rewriting
Obsidian's major advantage is its rich AI integration capabilities. The simple structure of markdown files in folders makes it easily readable by various AI tools. Options include Gemini CLI, Codex CLI, Claude Code, and Claude Desktop (GUI operation). However, for sensitive content like unpublished papers, using tools with options that prevent use in training data (paid plans or Claude Desktop's incognito mode) is essential.
Detailed setup methods are summarized here:
These integrations enable work such as content overviews, information classification and organization, summary creation, and theme-specific search and extraction for specified folder contents. This allows you to efficiently find necessary content from accumulated information and generate new ideas and questions through interactive idea development.
Information Supplementation with NotebookLM
Integration Timing
While working in Obsidian, it's common to find insufficient information to extract from original notes or papers, or for new questions and challenges to emerge. This is where NotebookLM proves useful. For handling medium-scale literature (around 2-50 documents) cross-sectionally, I think this is the most convenient and manageable method.
A simple explanation is provided in this article:
Efficient Usage Methods
First, filter necessary documents from literature pre-classified in Zotero (this is where the initial classification proves useful). Select all related literature, choose File > Export PDF, and output to an appropriate folder. Then upload everything to NotebookLM at once.
This creates a notebook that enables efficient work such as checking answers to new questions, integrating information across multiple documents, and grasping the overall picture.
The Audio/Video function is particularly convenient when you don't have sufficient reading time. You can grasp the atmosphere of collected literature through audio, enabling efficient information input when reading time is limited and learning during travel time.
New insights and organized information obtained from NotebookLM are re-imported into Obsidian for integration. This completes information organization based on deeper understanding that has evolved from initial notes.
The advantage of NotebookLM here is that sources can be verified. Clarifying sources is essential for researchers when writing papers or articles. While NotebookLM's search function isn't perfect, clicking the source button allows you to verify the quoted sections.
▲ Simply clicking the numbered button displays quoted sections in a popup
You'll sometimes notice that unrelated text is presented, so I recommend always checking. Conversely, this approach helps maintain source accuracy.
Workflow Advantages
The advantage of this workflow is that you can prepare for output while maintaining source accuracy. It can be used for research activities like academic writing, experimental idea generation, and research presentation preparation, as well as blog article creation, YouTube video production, and educational content development (which I actually use it for).
This system leverages the strengths of each tool: Zotero excels at literature information management, Obsidian is nimble and excellent for writing and information integration, and NotebookLM can summarize with LLMs while preserving sources.
Conclusion
I've introduced a simple workflow combining Zotero, Obsidian, and NotebookLM. By clarifying each tool's role and integrating them, you can use the system from literature management to final output while appropriately utilizing AI. Since these are all free and user-friendly tools, I encourage you to try them out.
References
Arango, J., Rheingold, H., 2024. Duly Noted: Extend Your Mind Through Connected Notes. Two Waves Books, Erscheinungsort nicht ermittelbar.
Forte, T., 2022. Building a Second Brain: A Proven Method to Organise Your Digital Life and Unlock Your Creative Potential. Profile Books.
コメント