Google shifts Gemini Notebook to more granular, compute-based usage limits

Google is adjusting how consumers can use Gemini Notebook, introducing what it describes as flexible, compute-specific usage limits designed to give users more control over their AI workflow. The change comes with a new way of measuring “overall usage” and a set of options for handling tasks when a user hits their limit.

In a product update posted Monday, Google said it will extend access to the features people already use, while changing how limits are managed. Instead of limits refreshing once per day, the company said they will refresh every five hours, allowing users to keep working throughout the day without waiting until the next daily reset.

Google also said the notebook’s usage limit will factor in multiple dimensions of each request. The company cited prompt complexity, chat length, the number of sources used, and which notebook features are invoked as inputs to the overall limit. The intent, as framed by Google, is to make limits more reflective of the compute work different tasks require.

To help users track what they are consuming, Google said Gemini Notebook will keep a record of usage and will suggest alternative outputs if a first option would exceed the user’s limit. For example, the notebook can steer users toward different results rather than forcing them to stop when capacity runs out.

When a user reaches their limit, Google said certain generation tasks can be deferred, including outputs such as Video Overviews and Slide Decks. Rather than blocking the workflow entirely, Google said those items can be queued for later generation and then produced automatically.

The update also introduces an automation layer for these deferred outputs. Google said users can set up notifications so they are informed when tasks like Video Overviews or Slide Decks finish generating, which is aimed at keeping users productive even when immediate compute is unavailable.

Google said the update will roll out to consumer accounts on the web and mobile starting September 2. It did not specify which exact account types will receive the change first beyond that consumer web and mobile availability window, nor did it provide detail on whether existing usage limits will be recalibrated for current users’ plans at the moment the update starts.

The broader context is that AI notebooks and assistants are increasingly managed like compute platforms. As models generate text, summaries, and media, providers must balance performance with infrastructure costs. Google’s decision to tie limits to prompt complexity, source counts, and feature use suggests it is trying to align user constraints more closely with the underlying processing demands of different workflows.

What remains unclear is how Google’s compute-specific limit mechanism will behave under edge cases, such as long back-and-forth chats or unusually heavy source citations. The company did not publish formulas, thresholds, or limit numbers, and it did not disclose how the notebook will classify tasks that mix multiple features in a single session.