The AI that can’t make things up.
NotebookLM only reads what you give it. That changes everything about how you use it.
Most AI tools answer from the whole internet.
They sound confident. And you have no way to check them.
NotebookLM works in reverse:
It only reads the documents you hand it.
Every answer links to the exact line it came from.
If the answer isn’t in your files, it says so.
If you rather watch how it caught the problem.
Reading about it and seeing it are different things.
The 7-minute video shows every click: the citations opening, the map unfolding, the audit catching the conflict in real time.
…That one constraint turns it from a chatbot into something more useful: a second reader for your life’s paperwork.
I recently fed it an entire vacation. It caught a problem I had missed. More on that in a minute.
First, let me teach you the five features that matter.
1. Grounded answers, with receipts.
The core feature. You upload sources (PDFs, docs, even emails saved as files) into a “notebook“, and then you ask questions in plain English.
The difference from a chatbot is the small number next to every claim.
That number is a citation. Click it, and NotebookLM opens the exact passage in your own document.
You are not trusting the AI. You are checking its receipts.
From my trip: I asked how many checked bags were included and when hotel check-in opened. Two answers, two citations, two clicks to verify both against the actual policies.
2. Reports: the pile becomes one page.
Inside the Studio panel, one click generates a “Briefing document“: a single overview of everything in the notebook.
This is for the moment you don’t even know what to ask yet.
And if the built-in formats don’t fit, “Create your own“ lets you write the report prompt yourself.
From my trip: I asked for a chronological timeline of every date and deadline across all documents. The cancellation deadlines that were scattered across three different policies showed up lined up on one path.
3. Mind Maps: see the shape of it.
One click turns the notebook into a visual map. Branches for each area, details underneath, and you can click any node to ask a focused question about just that piece.
A folder of unrelated files starts looking like one organized thing.
This is the feature where the picture does the explaining, which is exactly why it’s better seen than described.
4. Data Tables: the numbers stop hiding.
Tell it the columns you want: item, provider, base price, taxes, fees, refundable, deadline.
It reads every document and fills the table. Then one click exports it to Google Sheets, citations included.
From my trip: every cost that was buried across confirmations, resort fees, toll passes, deposits, landed in one table. NotebookLM did the extraction. I kept the math.
5. The audit: documents checking each other.
This one is not a button. It’s a way of asking.
Load several documents, then ask where they disagree. NotebookLM reads all of them at once and reports conflicts, citing both sides of each one.
This is the feature that earned my trust.
From my trip: two bookings, each perfectly fine on its own, contradicted each other in the fine print. No single document showed the problem. It only existed between them. The audit pointed at both lines.
It didn’t fix anything. It told me exactly which phone call to make. That’s the point.
One more thing.
You know the person who plans everything from a folder of unread PDFs. The contract they skimmed. The benefits packet they’ve been avoiding.
You can be the reason they stop guessing.
It’s free. One click. Share this with them.
Be that person for someone.
PS: If someone shared this with you, welcome. This is Growsembly, the weekly version of what I’m researching and testing in AI. It’s free: https://letters.guneytopcu.com/