Taking Good Notes With NotebookLM

NotebookLM (Google, 2023) is a specialized AI tool where you upload sources — PDFs, web pages, notes — and it answers questions only from those sources. The grounding kills most hallucination risks. For a research notebook, it is nearly perfect.

What makes it different

• Every answer is citation-grounded to the specific source lines
• No mixing with the open web — your notebook is a sealed universe
• Auto-summaries, FAQs, timelines generated from your corpus
• An Audio Overview mode that turns papers into a spoken podcast (genuinely wild)

A workflow for reading AI papers

1. 1Create a notebook per topic (e.g., 'agent evaluation')
2. 2Upload 10-30 key papers as PDFs
3. 3Generate an auto-summary to orient yourself
4. 4Ask specific questions — 'How does paper X's method compare to paper Y?'
5. 5Export your Q&A session as your note-taking draft
6. 6Add your own commentary in a separate doc — never let the AI be the only voice

Limits to know

• Cannot pull from the open web — if a source is missing, it will say so (or awkwardly refuse)
• Struggles with math-heavy PDFs where notation matters
• Summaries can blend multiple papers' views into a false consensus
• No multi-user collaboration (yet); inherently single-researcher

“A good note-taking system makes your older self the best teacher of your younger self.”

The big idea: NotebookLM turns a pile of PDFs into a queryable memory. But your own notes are what compound — never skip the write-up step.