Ask Cat › AI Tool Summary › Gemini Notebook
Gemini Notebook Goes Compute-Based on September 2, 2026: Five-Hour Refreshes, Tier Multipliers, and the Old Daily Table Google Never Took Down
Article last updated：2026-08-31
Google announced on 2026-08-28 that starting September 2, 2026, Gemini Notebook (formerly NotebookLM) is replacing its fixed daily generation limits with compute-based usage limits. It lands in days, and it applies to consumer accounts on both web and mobile — not just paying subscribers.
This is not a rewrite of the press release. Three things matter: how the new system actually works, what free-tier users will feel day to day, and one detail almost nobody has flagged — as of 2026-08-31 the old daily-limits table is still live on Google’s own support site, with no notice that it is about to be superseded.
What actually changes
Google’s blog post calls them “flexible, compute-specific usage limits.” Broken down, there are four changes:
| Before (until Sept 1) | After (from Sept 2) | |
|---|---|---|
| How it is counted | A separate daily cap per feature | One shared compute budget |
| Refresh | Daily | Every 5 hours, plus a weekly cap |
| Visibility | You find out when you run out | Remaining usage and reset time shown under the chat |
| When you run out | Wait until tomorrow | Queue a “generate later” job; it runs automatically after the reset |
The support page is specific about what consumes compute: the complexity of your prompt, the models and features you use, the length of your chat, and the number of sources. In other words, the same click on “generate” does not cost the same when your notebook holds 3 sources versus 40.
None of this is new machinery. Google notes that the Gemini app moved to the same compute-based model back in May 2026; Notebook is being folded into it. If you want a preview of how it feels, the Gemini app is already running it.
Plan multipliers: ratios only, no absolute numbers
The support page gives relative multipliers, not figures:
| Subscription | Usage limit |
|---|---|
| No plan (free) | Standard limits |
| Google AI Plus | 2x standard |
| Google AI Pro | 4x standard |
| Google AI Ultra | 5x or 20x higher than AI Pro, depending on your subscription |
Those ratios match the old system — our 2026-08-08 check of gemini.google/subscriptions already recorded Plus and Pro as 2x and 4x the free tier. Per our existing tool-page verification log, the two Ultra tiers correspond to US$99.99/month (5x) and US$199.99/month (20x).
But Google has published no number at all for what “standard” means. That is the real gap in the new system: under the old rules you could at least reason about “three Audio Overviews a day on the free tier.” Under the new rules you get a progress bar. Anyone planning coursework or client deliverables around this should know that up front.
What free-tier users will actually notice
Three concrete changes, all drawn from the official documentation:
1. Shorter lockouts, less predictability. Hitting the cap used to mean waiting until tomorrow. Now the quota refreshes every five hours, which is more generous in principle. The trade-off is that you cannot compute in advance how much you have left, because the cost depends on how many sources you attached and how long the conversation ran.
2. Heavy features get a price tag before you commit. Google says Studio-type generations show an expected AI usage cost with a bar indicating how much it will consume, shown before you trigger it. For free-tier users this is genuinely useful: you can decide whether a Video Overview is worth the spend.
3. “Generate later” is web-only. When you are out of capacity you can defer expensive outputs like Video Overviews or Slide Decks; they generate automatically once your limit resets, and you get a notification. The support page states this is available on web only — if you work mostly on a phone, that escape hatch is closed.
The in-product messages, as quoted by Google and the reporting on the change, read: “You’re almost at your AI usage limit. Limit resets at 3:00 PM.” and “Limit reached. All features are available after 3:00 PM.”
The detail nobody is reporting: the old table is still up
On 2026-08-31 we opened the older official limits page, support.google.com/notebooklm/answer/16213268, directly. The result: the “Usage Limits (Subject to Change)” table is still there and still expressed per day — chats at 50/day on Standard, Deep Research at 10/month on Standard, up to 5,000/day on the Ultra 30 TB tier. There is no notice on that page about the September 2 change.
Meanwhile the new page, support.google.com/gemininotebook/answer/17670842, talks only about compute-based AI interaction limits and never mentions notebook counts or source counts at all.
So two official accounts currently coexist. The reasonable reading is that the new system replaces the daily caps on AI interactions (chats, Audio and Video Overviews, reports, flashcards, quizzes, mind maps, Deep Research), while notebook count (100 per person on the free tier) and sources per notebook (50 on the free tier) are capacity limits rather than compute, and most likely stay as they are.
That is inference, not an official statement, and we are flagging it as unverified. Google has not said which rows of the old table are being retired or when that page will be updated. After September 2, trust the live usage indicator in your own account — it is the only source that cannot go stale.
What we could not verify, and will not guess
- The absolute value of “standard” limits. Multipliers only; no numbers published.
- The size of the weekly cap. Google states one exists but has not published a figure.
- Whether Workspace and education accounts are affected. The documentation addresses consumer accounts and does not spell out enterprise or education handling.
- Whether the change lands simultaneously worldwide. The announcement is not broken down by region.
- When the old table comes down. Not announced.
What to do
- If you have something due this week: run the heavy generations before September 2. Any planning you did off the “three per day” numbers stops applying that morning.
- If you are a heavy free-tier user: spend the first week after the change just watching the usage indicator under the chat box, then decide about upgrading. Upgrading before you have measured your own consumption means buying a quantity whose size you do not yet know.
- If Gemini Notebook sits inside a workflow: build “generate later” into it and make the expensive outputs asynchronous rather than sitting and waiting. Remember that this only exists on the web version.
Official links
- Google’s announcement (flexible usage limits): https://blog.google/innovation-and-ai/products/gemini-notebook/new-flexible-usage-limits/
- Gemini Notebook usage limits support page (new system): https://support.google.com/gemininotebook/answer/17670842
- The older usage limits table (still live as of 2026-08-31): https://support.google.com/notebooklm/answer/16213268
- Google AI subscription plans: https://gemini.google/subscriptions
What Amo and Pimi think
Students, researchers, and data organizers: the free version is sufficient, no need to think twice. Only consider AI Plus for US$4.99 if you need a higher quota. However, don't subscribe to Ultra just for NotebookLM; first confirm that you really need Gemini's other features.
Let's take a look at these
- Gemini Notebook Comprehensive Introduction: Pricing, Features, and Actual Limitations
- Gemini Notebook Is the free quota enough?
- Gemini Notebook Alternatives
- Comprehensive Free Quota List for All Tools
