Google has reworked how usage limits work in its AI note-taking tool, Gemini Notebook. Fixed daily caps on the number of generations are gone, replaced by a system that measures the compute you actually consume. Allowances refresh every five hours, and when you run out, heavy jobs can be queued for later. The change began rolling out to consumer accounts on web and mobile on September 2.

From counting runs to spending a budget

Until now, Gemini Notebook allocated a set number of feature generations per day based on your Google AI plan. The new approach behaves much more like a budget.

Several factors determine how quickly that budget drains: the complexity of your prompt, the models and features you use, the length of your chat, and the number of sources loaded into the notebook. Firing off dozens of short questions is therefore not the same as asking the tool to digest a large source set and produce a video or a slide deck. Because the weighting follows the actual work, lighter users have less reason to watch a counter.

The refresh interval changed too. Instead of a single daily reset, the allowance returns every five hours, so hitting the ceiling no longer ends your day. That said, the supply is not unlimited: the five-hour refresh applies until you reach your weekly limit.

Where the remaining allowance appears, and what happens when it runs out

You can check what is left by opening Settings and selecting Usage at the top right. The information also surfaces naturally while you work.

A message at the bottom of the chat shows your remaining allowance and when it will reset. Google's own examples read along the lines of "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." When you generate an artifact in the Studio panel, a bar indicates the expected compute cost of that operation. The more the bar is filled in, the more expensive the job.

If a request would push you past the limit, Gemini Notebook suggests alternative outputs. For heavy work such as Video Overviews or Slide Decks, you can also choose not to generate now. A Generate later option, currently available only on the web, queues the job and runs it automatically once your allowance recovers. Completion can take a couple of hours, but enabling notifications tells you the moment the result is ready. Rather than stopping dead at the limit, you stack the work in a queue and move on.

How wide the allowance gets by plan

Users without a paid plan receive the standard limits. Paid tiers multiply that baseline.

Google AI Plus is two times the standard limit and AI Pro is four times. AI Ultra goes considerably further, at either 5x or 20x the AI Pro limit depending on the subscription. An upgrade path is built into the screen you see when you hit the ceiling, so waiting and paying are both one click away.

Google notes that these limits are not fixed and may change based on testing, experimentation, or availability. The company says it may at times cap the number of prompts and conversations, or how much you can use certain features within a given timeframe, in order to keep the experience workable for everyone.

Generative AI pricing is converging on the same idea

This is less a quirk of Gemini Notebook than a reflection of where generative AI services are heading. Counting runs is easy to explain, but it does not track server load. Treating a short text reply and a long video render as one unit each drifts from reality for both the provider and the user.

Gemini Notebook is the tool formerly known as NotebookLM, folded into the Gemini brand. As people load sources and ask for summaries, audio explainers, videos, and slide decks, the gap in weight between one request and another has widened. Moving to a compute-based measure is an adjustment that lets pricing and allowances reflect that gap directly.

Summary

Gemini Notebook moved to compute-based usage limits on September 2. Consumption varies with prompt complexity and source count, and the allowance refreshes every five hours. Even at the ceiling, Generate later queues heavy jobs, and the remaining budget is visible from the chat panel and the settings screen. Paid plans start at two times the standard allowance and reach up to 20x the AI Pro limit on AI Ultra. The tradeoff for escaping run counts is that you will think a little more about how heavy your own workflow really is.