> ## Documentation Index
> Fetch the complete documentation index at: https://docs.granola.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Dictation history and privacy

> Find past dictations, retry failed ones, and understand what dictation stores, what it reads from your screen, and where it's sent.

# History

Your recent dictations are in **Settings > Voice typing**, under **History**. The settings page shows your last five. Click **History** to see them all, grouped by day.

* Click a dictation to copy it.
* Open the **...** menu and select **Delete** to remove one.
* Click **Clear all** to remove every dictation. This can't be undone.

Your history is only saved locally. It isn't synced to your other devices or saved to your Granola account. Cancelled dictations aren't saved.

# Local data storage

You can control what is stored locally on your Mac under **Local data storage** in **Settings > Voice typing**:

* **Store data locally**: Keeps your dictation history until you delete it. Your dictation audio is kept only for failed dictations and deleted after 24 hours. This is the default.
* **Delete after 24 hours**: Removes your dictation history and audio from failed dictations after 24 hours.
* **Never store data locally**: Never stores your history or audio from failed dictations locally.

Changing the setting also removes anything already stored that the new setting doesn't allow.

# Retrying a failed dictation

If Granola can't finish a dictation (e.g. if your connection drops), it stores the audio so you don't have to say it again. The pill at the bottom of your screen lets you retry straight away, or you can retry the dictation from **History** later.

Saved audio is encrypted, stays on this Mac, and is deleted after 24 hours. Audio from dictations that worked is never kept. With **Never store data locally**, no audio is saved, so failed dictations can't be retried.

# What leaves your Mac

Granola only listens while you're dictating: from when you press your shortcut until you finish or cancel.

To turn your speech into text, Granola sends these to its transcription and AI providers:

* your audio
* your dictionary and chosen language
* with [screen context](#screen-context) on, limited text from the app you're dictating into

Our transcription and AI providers do not use your dictations to train their models. See [Models & training](https://docs.granola.ai/help-center/consent-security-privacy/model-training.md).

With [Learn from your corrections](https://docs.granola.ai/help-center/voice-typing/spelling-names-and-terms.md#learn-from-your-corrections) on, Granola watches the text field for a moment after pasting. If you fix a word, it sends that word and the text around it to check whether it's a name or term worth learning to improve your experience.

## Screen context

With [screen context](https://docs.granola.ai/help-center/voice-typing/spelling-names-and-terms.md#screen-context) on, Granola reads text from the app you're dictating into when you start and when you finish. This can include:

* the app's name, window title, and the website address if you're in a browser
* the text before and after your cursor in the field you're typing in, and any text you've selected
* other text that's visible nearby, like the email you're replying to or messages in the conversation

Granola reads this text through macOS Accessibility. It doesn't take screenshots, and it skips password fields. The text is sent with your audio for that dictation only. It isn't saved in your history.

To stop Granola reading your screen, turn off **Screen context** in **Settings > Voice typing**. Dictation still works, but [styles](https://docs.granola.ai/help-center/voice-typing/dictation-styles.md) and app-specific formatting are turned off.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.