NotebookLM Free: The Complete Guide (and What Gemini Notebook Changes)

You search for NotebookLM free and land on a Google page called "Gemini Notebook". You wonder whether you have the wrong tool, whether the old one is dead, or whether the free tier vanished along the way. Short answer: no, no, and no. But the article you read to check may predate the change, and that is where the confusion starts.

I spent the morning of 7 August 2026 pulling these numbers straight from Google's help centre rather than copying them from somewhere else, because this tool's quotas have moved several times in a year and half the guides in circulation still quote the old ones. Everything below is dated and sourced. Where I could not verify something, I say so.

More importantly, this guide starts from real usage. Not "here are the features", but "here is what happens when you drop in your own documents and start questioning them". That is precisely what this tool does better than a general-purpose assistant, and the only good reason to open it.

What exactly is NotebookLM?

NotebookLM is a research assistant that answers only from the documents you feed it. You create a notebook, drop in PDFs, web pages, YouTube videos or Google Docs, and ask your questions. It reads everything, connects the documents to each other, and answers while citing the exact excerpt behind every claim.

The product was born in 2023, introduced at Google I/O under the code name Project Tailwind. A small, verifiable detail: the Android app identifier is still com.google.android.apps.labs.language.tailwind today, despite two commercial name changes. Google now reports more than 30 million users and 600,000 organisations.

The difference with ChatGPT: it only answers from your sources

This is the distinction that matters, and it is structural. ChatGPT draws on everything it learned during training, plus the web. NotebookLM is grounded: if the information is not in the documents you imported, it tells you it cannot find it rather than manufacturing it. Google puts it bluntly in its documentation: the tool will not be able to answer if there is no answer in the source document.

That is more restrictive, and that is the whole point. On a corpus you assembled yourself — your courses, your contracts, your internal documentation — you move from a tool that might be right to a tool whose every sentence you can check in one click. If you want the opposite comparison, I covered how the general-purpose assistant works in my complete ChatGPT guide.

What "grounded in your documents" means in practice

In practice: every chat answer is dotted with small clickable numbers. You click, the left panel opens on the relevant source and highlights the passage used. So you see not just the answer, but the sentence that produced it.

One nuance Google specifies and nobody mentions: if a source is too short, the tool cites the whole document instead of a precise excerpt. That is not a bug, it is a consequence of internal chunking. Another useful mechanic: sources are always used to answer, and so is conversation history, but your personal notes are only used if you explicitly select them.

NotebookLM becomes Gemini Notebook: what changes

This is the part confusing everyone right now, so let us deal with it head on.

What Google officially says, and since when

On 16 July 2026, Josh Woodward, VP of Google Labs, the Gemini app and AI Studio, published a post on the official blog titled "NotebookLM is now Gemini Notebook". The content fits in one sentence: it is a rename, not a replacement. The product stays standalone, but it extends across the rest of Google's ecosystem.

The product page FAQ is even more explicit: since July 2026, NotebookLM is called Gemini Notebook, it is the same product you already know, and all your existing notebooks remain fully accessible.

The traces of the change are observable everywhere, which is what lets you date it without relying on any third-party article:

• The notebooklm.google page still responds, but its title is now "Gemini Notebook".
• The help centre moved from support.google.com/notebooklm to support.google.com/gemininotebook — the old addresses redirect.
• On the Play Store, the listing is called "Gemini Notebook" and its release note literally reads: NotebookLM is now called Gemini Notebook.
• On the App Store, the app carries the same name, at version 1.51.1 dated 3 August 2026.

A nice detail: the help centre interface is still labelled "NotebookLM Help" on some pages while the content talks about Gemini Notebook. The rename is recent and not finished everywhere. If you see both names side by side, it is not you.

Do you need to do anything? No

No migration, no re-registration, no emergency backup. Your notebooks, sources and notes stay where they were. The old address works. The only visible practical change today is that you can now create and consult your notebooks directly from the Gemini app, with syncing between the two. Google also announces they are coming to AI Mode in Search — announced, not yet available.

Is NotebookLM really free?

Yes. There is a complete free plan, available with an ordinary Gmail account and no credit card. Google now calls it Gemini Notebook Standard. It is not a time-limited trial: it is a permanent tier.

The important point, and one rarely made: the paid plans unlock almost no exclusive features. They multiply counters. You get the same output formats on the free plan as on Pro — you simply get fewer of them per day.

Free versus paid: the table

Recorded on 7 August 2026 from the "Upgrade Gemini Notebook" page in Google's help centre. Google itself notes these limits are subject to change.

The ranges in the Ultra column correspond to its two storage tiers (20 TB and 30 TB), which do not share the same quotas. Also worth knowing: custom chat and analytics are available to all users, free ones included. Only advanced sharing is reserved for subscribers.

The real limits to know about

Beyond the table, four numbers and three behaviours change day-to-day usage.

• 500,000 words per source, or 200 MB for a file uploaded from your disk. No page limit: an 800-page PDF goes through if the word count fits.
• Daily quotas reset after 24 hours, and monthly quotas after 30 days. These are rolling windows, not a midnight reset.
• Sharing a notebook costs your collaborators nothing. The source limit stays the notebook's; it is not duplicated per person.
• Artefacts generated automatically on first import do not count against your quotas. If the tool spontaneously offers you a report or an audio summary on arrival, it is on the house.

On file types, the list is broad: PDF, Word (.docx), plain text, Markdown, CSV, PowerPoint, ePub, images (JPEG, PNG, WebP, HEIC and a dozen more), MP3 or WAV audio files, pasted text, web URLs, public YouTube videos, and Google Drive files. Two particular ceilings to remember: Google Slides decks are capped at 100 slides, and Google Sheets at 100,000 tokens.

One point I like a lot and that is little known: sources imported from Google Drive resync automatically every two to three minutes. Edit your Google Doc and your notebook follows. The flip side: if you lose access to the original file or it gets deleted, the source becomes inaccessible inside the notebook.

Do you need a Google Workspace account?

No, a personal Gmail account is enough and gives you the full Standard plan. Three technical conditions though: your age must be confirmed in your Google account settings, you must be in one of the 180-plus countries and territories where the Gemini app is offered, and — the classic corporate trap — if you use a work or school address, your Workspace administrator must have enabled the service. A lot of "it doesn't work for me" comes from there, not from a bug.

Workspace, Workspace for Education and Google Cloud accounts get extended limits and enterprise-grade data protection, with additional guarantees I come back to below.

Using NotebookLM step by step: a concrete case

Rather than a list of buttons, let us take a real case end to end. Mine, this morning: three documents to reconcile before a decision — a 60-page PDF, a reference web page, and an hour-long YouTube video. This is the typical situation where the tool saves time, because no human wants to read all three.

Step 1: create a notebook and import your sources

Go to notebooklm.google, sign in, click "Create notebook". Then "Add" to import. Three ways to feed the notebook:

• Drag and drop your local files (PDF, Word, PowerPoint, images and so on).
• Paste URLs. A poorly documented tip: you can paste several at once, separated by a space or a line break. Each becomes a separate source.
• Pick from Google Drive, or search the web for sources directly from the search field.

For the YouTube video, pasting the link is enough — it has to be public. Allow a few dozen seconds of indexing per source. You can also paste raw text and give it a title: handy for meeting notes you have nowhere else.

Step 2: ask questions and read the citations

This is where everything happens, and where most people go wrong. Google gives the example in its own documentation: instead of "Summarise this source", ask "What are the main conclusions of document X on this subject?".

Three habits that change answer quality:

• Name your sources in the question when you have several. "According to the annual report, not the video…" — the tool can tell them apart.
• Select a subset of sources in the left panel to narrow a question's scope. It is the best way to stop one document polluting the answer.
• Make contradiction do the work. The question that pays for the tool is "on which points do these three sources contradict each other?". A general-purpose assistant cannot answer that: it does not have your documents.

And every time: click the citation numbers. An answer whose citations you never checked is worth no more than a blog summary.

Step 3: generate a summary, a mind map, a study guide

The "Studio" panel on the right turns the notebook into usable formats: report, mind map, flashcards, quiz, infographic, slide deck, video summary. On the free plan you get 10 generations per day per format — plenty to work with.

The mind map is the most underrated format: it gives you the corpus structure at a glance and reveals the gaps. The quiz is the most honest revision tool there is, because it is built on your documents rather than a generic syllabus.

Step 4: produce an audio summary in your language

Still in the "Studio" panel, select "Audio Overview". Generation happens in the background: you can keep working, asking questions and generating other formats while it runs. Allow a few minutes. I give it its own section below, because it is the feature that made the tool famous and the one that raises the most questions.

Audio summaries, the format that made it famous

Two synthetic voices discuss your documents for about ten minutes, in podcast style. This format is what blew up NotebookLM's profile in 2024, to the point that Andrej Karpathy listed it among the most promising uses of language models.

How to get it in your own language

Output language is set under Settings, then "Output language", and the setting applies to everything: chat answers, generated documents and audio summaries. By default the tool inherits your Google account's preferred language — which is why some people get an unexpected language without understanding why.

Gemini Notebook announces more than 80 languages, with regional variants listed separately: European and Canadian French, European, Latin American and Mexican Spanish, Brazilian and European Portuguese, simplified and traditional Chinese.

When you generate, you also pick one of four formats:

• Deep Dive (default): two hosts analyse and connect the themes, long form.
• Brief: a single speaker, key points in under two minutes. The one I use most.
• Critique: two hosts give a constructive evaluation — useful on an essay or a design document.
• Debate: two hosts formally argue opposing positions on the topic.

You can also enter a free-form instruction to steer the summary toward specific themes or adjust the expertise level. Three generations per day on the free plan, and you need edit access to the notebook.

What it is actually worth, and its flaws

Let us be honest about two limitations Google documents and that enthusiastic articles forget to mention.

The length setting is English-only. The "Shorter", "Default" and "Longer" options only work in English. In any other language, you take the duration the tool decides.

Interactive mode is English-only too. That remarkable feature — joining the conversation by voice and asking the AI hosts a live question — is not available in other languages as of today.

Google also warns that the voices "are generated by AI and may contain inaccuracies or audio artefacts". My take after several months of use: the format is excellent for soaking up a corpus while walking or driving, and mediocre for extracting a precise figure. Never make a decision on the audio alone: go back to the chat and the citations.

Good news on the other hand: you can change playback speed, download the file, listen in the background while you keep questioning your sources, and share it by link. Careful though, public sharing is limited to personal accounts — it is disabled on Workspace Enterprise and Education accounts.

Four concrete use cases by profile

Student: revising from your own course material

The most obvious and most rewarding use case. Drop in your handouts, your lecture recordings (the tool accepts MP3 and WAV), the articles on your reading list, and generate a quiz. You get revision built on your syllabus, not on a generic cheat sheet found online. Fifty sources per notebook comfortably covers a semester of material.

Teacher: building material from your own documents

Educational use is established enough that national education bodies publish their own guidance sheets on it. The appeal for a teacher: produce a slide deck, an infographic or a set of flashcards from your own corpus, without starting from scratch and without risking a general-purpose model's approximations. The Google Drive resync is precious here: fix the source document and the notebook follows.

Professional: working through a tender or a contract

The specification, the technical annexes, the consultation rules, the meeting notes: five to ten PDFs nobody has time to read in full. The questions that pay off here are cross-referencing ones — "which obligations appear in the annexes but not in the specification?", "which deadlines are mentioned, and in which document?". Every answer is cited, so it holds up. If your need is more about extracting structured data from a PDF, I compared a different approach in my article on extracting data from a PDF.

Freelance: querying your own business documentation

A permanent notebook holding your terms and conditions, your quote templates, your technical documentation and your client notes becomes a queryable memory. This is probably the tool's best effort-to-benefit ratio, because the notebook is built once and serves for months. And if you are entirely new to this kind of tool, my step-by-step guide to getting started with AI lays the groundwork before this one.

The limits and the pitfalls

What NotebookLM cannot do

It does not browse the web for you mid-conversation: it works on what you imported. It refuses to answer on certain sensitive subjects, even in a historical context, because of safety filters — Google documents this explicitly. It has no notebook duplication feature as of today. And above all, worth engraving: a deleted note is gone for good, and Google states there is currently no way to recover one.

Finally, the headline addition announced in July 2026 — a "secure cloud computer" per notebook, able to write and execute code to analyse your data — is not in the free plan. It is reserved for Ultra subscribers and certain Workspace customers, with a rollout under way to Pro subscribers. If you read somewhere that NotebookLM does data analysis on the free tier, that is wrong as of 7 August 2026.

Do your documents train the models?

The question deserves a precise answer, because the short answer is "no, except in one case", and the case matters.

Google states it does not use your Gemini Notebook content to train its foundational generative models, unless you submit feedback. And there the consequence is sharp: when you click "like" or "dislike", the full context of the interaction is collected — your prompts, your uploads, your sources, the generated outputs. That data is reviewed by specially trained teams, disassociated from your Google account before review, and retained for up to three years.

In other words: the thumbs-down is not a harmless gesture on a confidential document. Google says as much itself, asking users not to include confidential or sensitive information in their feedback.

Workspace and Workspace for Education accounts escape this mechanism entirely: their uploads, queries and model responses are neither human-reviewed nor used for training, even when feedback is submitted. On Google Cloud, files stay inside the customer's GCP project with data regionalisation applied.

When to prefer another tool

• You want to explore a subject you know nothing about and have no documents: use a general-purpose assistant, NotebookLM would have nothing to read.
• You want to write from scratch: not its purpose, it stays glued to your sources.
• You want to automate a recurring task: look instead at autonomous AI agents.
• Your documents are under a strict confidentiality clause: go through a Workspace or Cloud account, not a personal Gmail.

FAQ: frequently asked questions about NotebookLM

Should you use it? My take

Yes, without hesitation, on one condition: that you have documents to give it. This is a corpus tool, not a curiosity tool. Opened empty, it is useless; fed with your courses, your contracts or your business documentation, it does something no general-purpose assistant can — answer on your material, citing its sources, without inventing.

The free plan is generous enough that the subscription question does not arise before heavy use. Fifty queries and three audio summaries a day is far more than people assume, as long as this is not your full-time job.

As for the name change: it is noise. Gemini Notebook is NotebookLM, with the same account, the same notebooks and the same address. The one real piece of advice I can give you here is to distrust the numbers you read elsewhere — the quotas have moved several times in a year, and many articles still quote the old ones. The figures on this page were recorded on 7 August 2026 in Google's help centre. They will move too.

Official sources consulted

• Google, "NotebookLM is now Gemini Notebook" — official rename announcement, 16 July 2026.
• Gemini Notebook product page — official FAQ on the rename and on privacy.
• Google Help, "Upgrade Gemini Notebook" — full quota table by plan.
• Google Help, supported source types — formats, sizes and Drive syncing.
• Google Help, generate an audio summary — formats, languages and limitations.