Draft a dissertation lit review from uploaded sources
A doctoral writer used NotebookLM, which cites only uploaded sources, to handle citations and draft a literature review without invented references. Task: Suggested implementation: A doctoral writer uses NotebookLM to draft literature reviews, restricting citations to uploaded PDFs only. The workflow involves renaming files before upload and applying a standing instruction to format references as author-year. Human roles include vetting sources, organizing them, and synthesizing content, though specific division of labor is not detailed in the source.; tools: custom-built or unnamed agent. Human role: The original post does not say how a person stayed involved.; limitations: Prerequisites include a library of relevant PDFs and strict adherence to file naming conventions. The primary constraint is the risk of citation errors requiring extensive rework, as evidenced by the 60% timeline setback. Integration uncertainty exists regarding the 'broader workflow' mentioned but not fully described. Human approval is critical for verifying every citation against original documents..
Tools
custom-built or unnamed agent
Published on Agent Work Atlas:
Overview
A doctoral writer used NotebookLM, which cites only uploaded sources, to handle citations and draft a literature review without invented references.
Workflow
Suggested implementation: A doctoral writer uses NotebookLM to draft literature reviews, restricting citations to uploaded PDFs only. The workflow involves renaming files before upload and applying a standing instruction to format references as author-year. Human roles include vetting sources, organizing them, and synthesizing content, though specific division of labor is not detailed in the source.
Reported result
The approach reportedly reduced manual cross-checking and prevented invented references in draft chapters. However, it did not support actual data analysis. A significant failure occurred when bad citations slipped through, forcing the writer to rebuild work at 60% of their dissertation timeline, resulting in lost progress.
Human role
The original post does not say how a person stayed involved.
Limitations
Prerequisites include a library of relevant PDFs and strict adherence to file naming conventions. The primary constraint is the risk of citation errors requiring extensive rework, as evidenced by the 60% timeline setback. Integration uncertainty exists regarding the 'broader workflow' mentioned but not fully described. Human approval is critical for verifying every citation against original documents.
Try it
I need to draft a literature review section for my dissertation using NotebookLM. I have uploaded [replaceable: list of PDF filenames] which are renamed to [replaceable: naming convention, e.g., Author_Year_Title]. Please apply the standing instruction: 'Cite sources using author-year format (e.g., Smith, 2020) and only reference the uploaded documents.' First, perform a dry run on [replaceable: 2-3 sample papers] to generate a short synthesis paragraph. I will manually verify these citations against the PDFs. If the accuracy rate is 100%, proceed to draft the full section for [replaceable: specific topic/chapter]. Do not invent any sources. If a citation cannot be verified, flag it for my review. Output the draft in [replaceable: format, e.g., Markdown] with a separate list of all cited sources for my cross-check.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.