NotebookLM answers only from the material you give it. Load paper PDFs and you can ask questions about those papers. Unlike a general chatbot, it points to the passage its answer came from, which is what makes it useful for reading research.
This article covers the steps, and separates the cases it fits from the cases it does not.
What it can do
Once your PDFs are loaded, there are four main things.
- Summarise. It lists the main points of one paper. With several loaded, it can summarise across them.
- Answer questions. You can ask things like “what was the control group in this study”.
- Compare across papers. Load three papers and ask where their conclusions disagree, and it lays out the differences.
- Produce an audio overview. It generates a spoken conversation about your material, which is useful for recalling content while travelling.
The steps
- Open NotebookLM and create a notebook.
- Add your sources. Besides PDFs, it accepts Google Docs, plain text, and web page URLs.
- Wait while the sources are processed. Longer PDFs take longer.
- Type your question.
Paper PDFs downloaded from a publisher’s site work as they are. A scanned paper without embedded text may not be readable, in which case run it through optical character recognition first.
How it shows its sources
Answers come with numbered markers. Clicking a marker opens the passage in your source that the answer came from.
This matters because you can check the answer yourself. When you are reading research, whether a summary is accurate feeds straight into your conclusions. Being able to jump to the passage means you do not have to take the summary on trust.
Whether the passage genuinely supports the answer is still your judgement. A source being shown is not the same as an answer being correct.
What it cannot do
It cannot find papers. NotebookLM reads only what you load. Asking it to name the important papers in a field returns nothing you did not already add. Finding papers happens elsewhere, and you bring the PDFs here.
There are limits on how much you can add. The limits depend on your plan and change over time, so check the official help for current values. It is not built for loading hundreds of papers at once.
It does not format citations. For a reference list, use reference management software.
Where it fits and where it does not
| Situation | Fits | Why |
|---|---|---|
| You have settled on about ten papers | Yes | Within the limits, and cross-paper questions work |
| Finding a passage in a long paper | Yes | You can jump to the source |
| Getting an overview of a field | No | It cannot surface papers you did not add |
| Building a reference list | No | It does not format citations |
| Working with hundreds of papers | No | There is a limit on sources |
Connecting it to the search step
Choosing the ten papers to load is the part that actually takes time. That part belongs to a different tool, and you hand the resulting PDFs to NotebookLM.
In Litlas you can search for papers and follow the citations as a graph. From one paper you can move to what it cites and to what cites it, which is how you reach the papers at the centre of a field. What you pick can be saved with notes attached. The free plan needs no card and works right away.
The search method is in how to find prior research, and the reading method is in reading a paper in three passes.
Summary
NotebookLM is a tool for reading the papers you already have. Because it shows its sources, you can check whether a summary is accurate. It does not find papers and it does not format references. Leave the finding and the writing to other tools, give it the reading, and it fits without friction.