Short answer: Gemini Notebook (the product Google renamed from NotebookLM) does not use one simple “messages left” counter. Google now describes compute-based usage limits: prompt complexity, model, feature, chat length and other factors affect usage. The short-term allowance refreshes every five hours until the account reaches its weekly limit. The exact remaining amount and next refresh are shown inside Gemini Notebook, not reliably inferred from a plan name alone.
This guide separates the limits people often mix together: the five-hour refresh, the weekly ceiling, notebook/source capacity, feature quotas, and account or region eligibility. It is based on Google’s official help pages and announcement, reviewed September 14, 2026. Limits can change, so use the in-product Usage panel as the final authority.
First, the name changed: NotebookLM is now Gemini Notebook
Google announced on July 16, 2026 that NotebookLM is now called Gemini Notebook. Google describes it as the same standalone research product with closer connections to the Gemini app and Search. That is why older searches, bookmarks and support screenshots still say “NotebookLM,” while the current help center uses “Gemini Notebook.” This article uses both names so you can map older instructions to the current interface.
What the five-hour refresh actually means
Starting September 2, 2026, Google says Gemini Notebook uses compute-based limits. The service considers more than the number of prompts: prompt complexity, the model and feature used, chat length and specific features can all affect how much quota an interaction consumes. A short question about one source and a long Studio generation should not be assumed to cost the same.
When you use up the current allowance, Google says the quota refreshes every five hours until you reach the weekly limit. This is not a promise of a fixed number of prompts every five hours, and it is not a universal clock time shared by every account. If the interface says that your weekly limit has been reached, waiting for the next five-hour refresh will not necessarily restore access immediately. Check the account’s own Usage panel for the displayed remaining amount and refresh time.
| Layer | What it controls | How to reason about it |
|---|---|---|
| Compute usage | How much AI work the account can request across chat and features. | Complexity, model, feature and context matter; there is no safe universal prompt count. |
| Five-hour refresh | The short-term allowance after usage is consumed. | Google says it refreshes every five hours until the weekly ceiling is reached. |
| Weekly limit | A longer-window ceiling on usage. | Do not assume a calendar-day or fixed weekday reset; read the account notice. |
| Notebook/source capacity | How many notebooks or sources a plan can hold. | This is storage/workspace capacity, not proof that more AI generations are available. |
| Feature quota | Daily or monthly counts for chats, overviews, reports and similar outputs. | A feature quota can be available while the compute allowance is temporarily exhausted, or vice versa. |
| Access and eligibility | Whether the account, region, age, Workspace policy or subscription can use a feature. | An upgrade cannot override an unsupported region or an administrator’s policy. |
Where to see your remaining usage and reset time
- Open Gemini Notebook (the old NotebookLM address may still be familiar).
- Open Settings → Usage in the top-right menu.
- In chat, read the usage message at the bottom. Google’s examples show a remaining-limit message and a specific “resets at” time when one is available.
- For Studio generation, look at the expected AI usage cost shown near the generation control. A fuller bar means a higher expected cost; it is a warning about consumption, not a performance score.
On the web, Google also documents Generate later in the Studio panel. When you are out of AI limits, you can schedule a generation for later and receive a notification if notifications are enabled. The help page currently marks this feature as web-only.
Standard, AI Plus, AI Pro and AI Ultra
Google’s usage-limits help page gives relative compute limits: AI Plus is listed as two times the Standard limits, AI Pro as four times Standard, and AI Ultra as five or twenty times AI Pro depending on the Ultra subscription. These are relative plan descriptions, not a promise that every prompt will last a particular number of turns. Google also warns that limits may change based on testing, experimentation and availability.
| Capability | Standard | AI Plus | AI Pro | AI Ultra (20 TB) |
AI Ultra (30 TB) |
|---|---|---|---|---|---|
| Notebooks | 100/user | 200/user | 500/user | 500/user | 500/user |
| Sources | 50/notebook | 100/notebook | 300/notebook | 500/notebook | 600/notebook |
| Chats | 50/day | 200/day | 500/day | 2.5K/day | 5K/day |
| Audio overviews | 3/day | 6/day | 20/day | 100/day | 200/day |
| Video overviews | 3/day | 6/day | 20/day Cinematic: 2/day |
100/day Cinematic: 10/day |
200/day Cinematic: 20/day |
| Reports, flashcards, quizzes, mind maps | 10/day each | 20/day each | 100/day each | 500/day each | 1K/day each |
| Deep Research | 10/month | 3/day | 20/day | 75/day | 200/day |
Google says daily quotas reset after 24 hours and monthly quotas after 30 days in the plan details. Those feature counts should not be read as a replacement for the compute-based five-hour/weekly limit. For example, a plan may still show a daily report allowance while a complex generation is blocked by the current AI usage limit.
Three practical scenarios
1. Light reading and a few source-grounded questions
If you add a small set of sources and ask a few short questions, Standard access may be enough for your workflow. That is a workload description, not a guarantee: a long conversation, a large source set or a heavier feature can consume more compute. If the bottom-of-chat notice gives a reset time, follow that notice instead of counting prompts manually.
2. Frequent Studio outputs
If you regularly create audio or video overviews, reports, slides, quizzes or deep research, inspect both the feature quota and the Studio usage-cost indicator. Several artifact types have separate daily counts, but they still run through the account’s AI usage system. Group source cleanup first, avoid regenerating the same artifact repeatedly, and use Generate later when the web interface offers it. Upgrade only after you know whether the problem is compute usage, a feature quota or notebook/source capacity.
3. A work, school or managed account
A Workspace, Workspace for Education, Google Cloud or other managed account can have a different access path and data-handling policy from a personal Google account. An administrator, supported region or qualifying license may determine whether Gemini Notebook features appear. Confirm the account badge and plan in the UI; do not assume that a consumer Google AI subscription transfers to a managed account.
What to do when Gemini Notebook says you are at the limit
- Read the notice. Record whether it names a five-hour refresh, a weekly limit, a feature quota or a missing entitlement.
- Check Settings → Usage. This is more reliable than advice copied from an older NotebookLM screenshot.
- Reduce repeated work. Stop regenerating the same Studio artifact, shorten an unnecessarily long chat, and consolidate source questions when that preserves the research goal.
- Wait or generate later. If the notice gives a refresh time, wait for it; on web Studio, use Generate later when available.
- Check access before upgrading. Confirm region, account type, Workspace policy and the plan badge. If the issue is a source/notebook cap, a compute upgrade may not solve it.
- Use an upgrade only for a measured need. Compare the published plan table with your actual feature mix, then re-check the current official page because Google can revise the limits.
Common mistakes
- “I have 50 chats, so I have 50 prompts left.” The plan table’s chat row and the compute-based usage system are different layers.
- “The five-hour refresh is my weekly reset.” It is a short-term refresh that stops helping once the weekly limit is reached.
- “More sources always means more usage.” Sources are a capacity limit; the AI work done with them is a separate usage question.
- “The old NotebookLM name means the help page is outdated.” Google’s current product name is Gemini Notebook, but older links and community guides may still use NotebookLM.
- “The plan page proves every account has access.” Google explicitly makes access subject to region, account type, experimentation and availability.
Official sources and last review
Google’s renaming announcement explains the NotebookLM → Gemini Notebook transition and product direction. The official usage-limits help page explains compute-based limits, the five-hour refresh, weekly ceiling, Usage panel and Generate later. The official upgrade page contains the plan capacities and account/data-handling notes. HiseHub last reviewed those pages on September 14, 2026; Google may revise the limits after this date.
Related HiseHub tools: compare AI usage limits across major services, check model and plan access, or review subscription prices by country before choosing a plan.