NotebookLM Changed How I Do Research – Here’s My Workflow
Google’s NotebookLM is genuinely one of the best free AI tools available and most people haven’t tried it. Here’s how I use it as a researcher.
What it is: You upload documents (PDFs, Google Docs, web pages, YouTube videos) and it creates an AI assistant that ONLY references your uploaded sources. No hallucinations from training data – just analysis of your actual documents.
My research workflow:
1. Upload all relevant papers/articles for a topic (up to 50 sources)
2. Ask it to create a briefing doc summarizing key themes across all sources
3. Ask specific questions and it responds with citations from your documents
4. Use the “Audio Overview” feature to generate a podcast-style discussion of the material (surprisingly good for passive learning)
5. Generate study guides and FAQ documents
What it’s great for: Literature reviews, competitive analysis, studying for exams, preparing for meetings where you need to reference multiple documents.
What it’s not great for: Anything that requires knowledge beyond your uploaded documents. It deliberately avoids using its general knowledge.
Seriously, if you haven’t tried it, spend 10 minutes uploading some documents. It’s free and it’ll change how you process information.
6 Replies
Join the discussion.
Log In to Reply
the citation feature is what makes it actually useful vs other AI tools. every claim is traceable back to a source, which matters a lot when you're prepping something that other people will rely on.
how does it handle contradictions across sources? like if two papers disagree on something, does it flag that or just pick one?
PSA: the 50 source limit sounds generous but there's also a per-source size cap. hit it with a 400 page PDF last week and it only ingested part of it. worth knowing before you build a whole workflow around it
yeah ran into this too. workaround i found was splitting large PDFs into chunks before uploading. annoying extra step but it gets the job done
Windsurf is another one worth looking at. Similar to Cursor but I find the UX a bit cleaner for my workflow. YMMV tho