NotebookLM (Gemini Notebook) vs. Regular Gemini: What's the Difference and When to Use Each

NotebookLM (Gemini Notebook) vs. Regular Gemini: What's the Difference and When to Use Each

Updated September 27, 2026. Every number below was read on that date from Google's own help pages, developer documentation and product blog, listed at the end. No third-party benchmarks and no reseller prices are used. Where Google publishes no number, this article says so instead of estimating one.

Photo: Dan Dimmock / Unsplash. Used under the Unsplash License.

The name changed, and that is not a trivia point

On July 16, 2026 Google renamed NotebookLM to Gemini Notebook. The help centre moved with it: requests to the old notebooklm paths now resolve under gemininotebook, and the Workspace product page labels it "Gemini Notebook (formerly NotebookLM)". The same announcement put the user base at more than 30 million individuals and more than 600,000 organisations.

Why it matters for a comparison article: there are now three names in circulation for two things. "Gemini Notebook" is the standalone product. "Notebooks in Gemini" is the surface inside the Gemini app, introduced April 8, 2026, which syncs the same notebooks both ways. And "NotebookLM" is the retired name that still appears in older pages. If you search for limits using the old name you will land on pages that redirect, which is exactly how people end up quoting limits from the wrong tier.

One product publishes numbers you can plan around; the other publishes multipliers

This is the real difference, and it is not about model quality. Gemini Notebook publishes a hard per-day counter for every feature it has. The Gemini app publishes no absolute usage number at all — only that a paid tier gives you "2x", "4x", "5x" or "20x" more than the tier below, and that "your limit refreshes every 5 hours until you reach your weekly limit".

So one of these two tools can be scheduled and the other cannot. If you need to know before Monday whether you can get through forty documents by Friday, only one of them answers.

Gemini Notebook, per user	Standard (free)	in Plus	in Pro	in Ultra (20 TB)
Notebooks	100	200	500	500
Sources per notebook	50	100	300	500
Chats per day	50	200	500	2,500
Audio Overviews per day	3	6	20	100
Reports, flashcards, quizzes, mind maps (each) per day	10	20	100	500
Deep Research	10 per month	3 per day	20 per day	75 per day

The reset rule is published as plainly as the limits: "Daily quotas are reset after 24 hours; monthly quotas are reset after 30 days." No clock time is given, so the reset is rolling rather than at midnight.

What a free notebook actually holds, in words and megabytes

Each source can be up to 500,000 words, or 200 MB for an uploaded file. Fifty sources at that ceiling is 25 million words in one free notebook. A 300-page non-fiction book runs roughly 100,000 words, so a single source slot holds about five of them and a free notebook holds around 250. Nobody will hit that. The limit that actually bites is the count of sources, not their size.

Format-specific caps are worth knowing before you plan an import: Google Slides is capped at 100 slides, Google Sheets at 100,000 tokens. Accepted sources include audio files, pasted text, Docs, Slides, Sheets, images, Word, PowerPoint, PDF, CSV, text, Markdown, Play Books ebooks, web URLs, ePub, YouTube links and Gemini chats. YouTube has one condition that catches people out: the video must have captions, user-uploaded or automatic, and "there is no limit for the length of the video unless the caption file contains over 500,000 words". Videos uploaded less than 72 hours earlier may not be available yet.

The daily counters, converted into a working week

Multiply the free tier out and it stops being abstract. Fifty chats a day is 350 chats a week. Three Audio Overviews a day is 21 a week, about 90 a month. Ten reports a day is 70 a week. And ten Deep Research runs a month is the one line that is genuinely tight — the first paid tier converts that same allowance into three per day, which is roughly 90 a month, a ninefold increase, for a subscription Google prices at 7.99 US dollars a month in the United States.

Against that, the Gemini app offers nothing to multiply. You can know that Pro is "4x higher" than free and that the counter refreshes every five hours — which allows at most four full refreshes in a twenty-four-hour day — but not what it is four times of. Two published timings do exist and are useful: an Audio Overview takes "a couple of minutes" to generate from a 2-step procedure, and a Deep Research report in the Gemini app takes 5 to 10 minutes from an 8-step procedure. Creating a notebook is 5 steps, and auto-labelling of sources switches on once you have 5 or more of them.

Where the answers come from, which is the difference that changes your work

Google's one-sentence statement about the notebook is the cleanest thing in any of this documentation: chat responses in Gemini Notebook only use data from your sources. Citations are direct quotes, text and images from those sources, and selecting one jumps you to the quoted passage in context.

Regular Gemini works the opposite way round, and the developer documentation says how. With the search tool enabled, "the model analyzes the prompt and determines if a Google Search can improve the answer", then returns inline annotations mapping each cited span to a source URL by character offset. The decision to go outside your material is the model's, not yours. Grounding with Google Search is priced on the API at 5,000 free search requests a month, then 14 dollars per 1,000 — a rare case where Google puts a number on something the consumer app leaves silent.

So the choice is not "which is smarter". It is whether you want a tool that cannot wander off your reading list, or one that is allowed to.

The upload rules diverge more than the models do

Both sides run on the same 1-million-token class of model, so the models are not where these two products separate. Uploads are.

Limit	Gemini Notebook	Gemini app
Files per request	50 sources per notebook, persistent	10 files per prompt
Size per file	200 MB	100 MB (video 2 GB)
Text ceiling	500,000 words per source	not published
Video length	no limit beyond the caption word count	5 minutes, extending to 1 hour on Pro or Ultra
Audio length	no published limit	10 minutes, extending to 3 hours on Pro or Ultra
Code	not a listed source type	one folder or repo, up to 5,000 files and 100 MB
Context window	not published as a token figure	32k free, 128k Plus, 1M Pro and Ultra

Read that table as a sorting rule. Anything persistent and large goes in a notebook. Anything transient and multi-modal — a long recording, a repository, a folder of screenshots — goes in the app.

Four of Google's own pages disagree, and the gaps are not small

If you are about to pay for a tier, know that the published figures do not line up. These are all current Google pages.

• AI Plus storage: 200 GB or 400 GB. The January 27, 2026 availability post says 200 GB; the plans page and the August 19, 2026 student post say 400 GB. Gap: 200 GB.
• AI Pro storage: 5 TB, or "5 TB or 10 TB". The plans page says 5 TB; the Google One help page says it depends on your specific Pro plan. Gap: up to 5 TB.
• AI Ultra storage: 20 TB or 30 TB. The May 19, 2026 post attaches 20 TB to a 100-dollar Ultra; the Notebook limits table names its two Ultra columns "20 TB Plan" and "30 TB Plan"; the 2025 launch post said 30 TB.
• AI Ultra price: 249.99 or 200 dollars a month. The 2025 launch post says 249.99; the May 2026 post describes reducing the top tier "from $250 to $200" — so Google's own two posts also disagree about what the old price was.
• The Ultra multiplier is one row and two products. The app limits page lists AI Ultra as "5x or 20x higher than AI Pro"; the I/O post splits that across a 100-dollar tier at 5x and a 200-dollar tier at 20x. One plan name, two different published limits.
• The same mid-Ultra tier, two source counts. Workspace "Expanded Level Access" gets 400 sources per notebook, 1,000 chats a day and 40 Audio Overviews; consumer "Ultra (20 TB Plan)" gets 500, 2,500 and 100. Presented as parallel tiers.

Which way to err: buy on the limits table, not the marketing page. The per-day counters are the numbers Google enforces, and they are the only ones stated identically wherever they appear.

What the model specs do and do not tell you

Google's developer documentation states that Gemini 3 models have a knowledge cutoff of January 2025. The current flash model is documented at 1,048,576 input tokens and 65,536 output tokens, with a "latest update" of September 2026 — and, notably, no knowledge-cutoff row of its own. The consumer pages add a naming problem: the app limits page lists "Gemini 3 Pro" on all four tiers, the API model index lists that identifier as shut down and only a 3.1 Pro preview as current, and the plans page calls the benefit "Gemini 3.1 Pro access". Three Google pages, three names for the Pro model.

The practical consequence is small but real: a January 2025 cutoff means anything from the last twenty months has to arrive as a source or a search result. That is an argument for the notebook when your material is recent, and for search grounding when it is public.

Tip: Audio files are an accepted source type at up to 200 MB each, which makes a recorded lecture or interview a first-class notebook input rather than something to transcribe by hand first. A basic digital voice recorder is the cheap end of that workflow. (These are Amazon Associate links — we may earn a small commission on qualifying purchases.)

Privacy terms, which differ by account type rather than by price

For Gemini Notebook, Google publishes that content "will not be used to directly train our foundational AI models, unless you choose to provide feedback", and that on work or school accounts uploads, queries and responses are not reviewed by humans and not used to train models even when you rate them. Reviewed feedback and its content are "retained for up to 3 years", disconnected from the account.

The Gemini app is governed differently: default activity is auto-deleted after 18 months, changeable to 3 or 36 months or off; human reviewers see some collected data; reviewed chats are kept up to three years; and with activity saving off, data is retained for 72 hours. Note the split Google actually draws: it is personal versus work or school, not free versus paid. Paying more does not buy different training terms.

Choosing between them as a scheduling problem

Put the two sets of documentation side by side and the decision resolves on something more boring than capability. If the work has a deadline and a known volume — forty papers, a term of lecture recordings, a due date — use the product that publishes a number you can divide by the number of days you have. Fifty chats and three Audio Overviews a day, reset every twenty-four hours, is a plan. "Four times more, refreshing every five hours" is not.

If the work is open-ended, needs the public web, or involves a long recording or a code folder, the app is the right tool and its unpublished ceilings matter less, because you were not going to schedule against them anyway.

And the one line worth keeping from all of this: the notebook cannot answer from outside your sources, and the app is designed to decide for itself whether to look. That is a property of the product, not a setting, and it is the only difference here that changes what an answer means. The same discipline of reading a vendor's own numbers rather than its claims applies to comparing two assistants on the tasks you actually do, and the wider labour-market version of it is in an earlier piece on this site.

What Google does not publish, stated rather than guessed

Five gaps are worth naming. First, there is no absolute prompt, message, image-generation or video-generation count for the Gemini app at any tier — only multipliers. Second, Audio Overview length is offered as "Shorter, Default, or Longer" with no minute values. Third, the number of supported languages is given as "80+" and regions as "180+", never exactly. Fourth, there is no published retention period for notebook sources themselves; the three-year figure covers reviewed feedback only. Fifth, no clock time is given for when daily quotas reset, and no boundary for the app's weekly limit.

Three checks before committing a term's work to either

1. Count your sources, not your pages. Fifty per notebook on the free tier is the binding constraint; 500,000 words each is not.
2. Divide your deadline by the daily counter. If the answer needs more than 50 chats or 3 Audio Overviews on any single day, you need the paid tier for that week, not for the year.
3. Check the caption on every YouTube source before building a notebook around video, and give anything uploaded in the last 72 hours a day to settle.

Pages every figure above was read from

• Rename announcement, July 16 2026: blog.google. Notebooks in Gemini, April 8 2026: blog.google
• Gemini Notebook tier limits and reset rule: support.google.com. Work and school tiers: support.google.com
• Source types, per-source ceilings, Slides and Sheets caps, YouTube conditions: support.google.com. FAQ: support.google.com
• Grounding in the notebook, and citations: support.google.com. Audio Overview steps and timing: support.google.com
• Gemini app limits, context windows and the 5-hour refresh: support.google.com. Upload limits: support.google.com. Deep Research steps and duration: support.google.com
• Search grounding mechanism and pricing: ai.google.dev. Model specs and the January 2025 cutoff: ai.google.dev
• Plan storage and multipliers: one.google.com. AI Plus price, January 27 2026: blog.google. Subscription changes, May 19 2026: blog.google
• Notebook privacy terms: support.google.com. Gemini Apps privacy and retention: support.google.com. Workspace product page: workspace.google.com

This article summarises Google's published documentation as of September 27, 2026. Limits, tier names and prices are changed by the vendor without notice — the limits table itself is headed "Subject to Change" — so confirm any figure you are about to spend money on against the page linked above it.