How to Check Whether Voice Journaling Dictation Stays on Your Device
If you use a keyboard’s microphone to dictate a journal entry, check the exact dictation feature, operating system, language, and settings before relying on local processing. Look for current official documentation that says processing happens on-device for that configuration. If it does not say so, treat the handling as unknown. An offline test can show whether dictation works without a connection, but it cannot prove that audio is never sent elsewhere. Also check whether the journaling app runs its own transcription or uploads recordings after the keyboard has entered text.
First identify which microphone feature you are using
A microphone button can belong to the system keyboard or to the journaling app. These are separate paths. With keyboard dictation, the operating system or keyboard provider recognizes speech and inserts text into the app’s text field. With an app’s own recorder or transcription feature, the app may capture audio and process it itself or send it to a service. A privacy statement about one path does not establish how the other works.
Before dictating, note the app, the keyboard shown, and how you started recording. If you tapped the microphone on the keyboard, investigate that keyboard’s dictation feature. If you tapped a recorder or transcription control inside the app, check the app’s documentation and privacy settings as well. If you cannot tell which component owns the control, pause before speaking sensitive material and look for the feature’s name in the app’s help or settings.
Check the exact device, OS, language, and setting
Start with the manufacturer’s current instructions for the device and software version in use. Confirm that the page describes keyboard dictation, not an assistant, a third-party transcription app, or a different device. Then check whether its local-processing statement covers your spoken language and region. Availability and behavior can vary by language, country, model, and OS release.
On iPhone, Apple’s [Dictate text guide](https://support.apple.com/guide/iphone/dictate-text-iph2c0651d2/ios) explicitly says that many languages are processed on the phone without an internet connection. Enable it under **Settings → General → Keyboard → Enable Dictation**. Check your language in the **On-Device and Modeless Dictation** section of Apple’s [feature availability page](https://www.apple.com/ios/feature-availability/), since general Dictation availability and on-device support are listed separately. The guide also describes a separate on-device Apple Intelligence enhancement for spelling, punctuation and capitalization with its own requirements. Do not confuse that enhancement with the underlying speech recognition. Read **About Dictation & Privacy** in the device settings for the applicable explanation. The guide notes that dictated search-field text may be sent to the search engine; local recognition does not mean the resulting text can never leave the device.
On Android, separate Android’s speech-recognition framework from the keyboard you actually use. Google’s [Gboard voice-typing instructions](https://support.google.com/gboard/answer/2781851?hl=en) explain how to dictate through Gboard and note that support and features vary by language and device. They do not, by themselves, promise that every Gboard dictation is processed locally. Android’s [SpeechRecognizer API reference](https://developer.android.com/reference/android/speech/SpeechRecognizer) warns that its general recognition service is likely to stream audio to remote servers; it separately exposes a check for whether an on-device recognition service is available. This is Android platform guidance for developers, not proof of the routing used by a particular keyboard, phone, or app. Look for documentation that names the exact keyboard feature and says how to select or use, and verify availability of, its on-device mode. If the documents do not settle the question for your model, OS, language, and settings, regard the answer as unknown.
Treat airplane mode as a limited test
You can test the precise dictation path with a harmless sentence: enable airplane mode, make sure Wi‑Fi and mobile data remain off, open the same app and keyboard, and dictate. If it fails, record that it did not work offline in that setup; the failure alone does not identify the cause. If it works, that only shows it can produce text without an active network in that moment. It does not prove that no audio was queued for later upload, that other sessions behave the same way, or that the app does not separately sync text or recordings when connectivity returns.
A successful offline test is therefore useful evidence about dependence on a live connection, not a privacy guarantee. Restore connectivity only after you have considered whether the app might sync pending content. Do not speak a private journal entry just to test the microphone; use a neutral sample phrase.
Check the journaling app separately
After checking keyboard dictation, inspect the app’s settings and privacy documentation for voice notes, transcription, cloud sync, backups, and retention. Look for whether audio is stored, whether transcription is optional, where processing occurs, and what happens when you edit or delete an entry. If the app accepts only the inserted text and never records audio through its own control, that narrows the question, but does not answer whether the keyboard’s speech service transmitted audio or whether the app syncs the resulting text.
If you cannot find a clear statement, ask the app provider which component handles audio, whether it leaves the device, and whether it is retained. Until the answer is specific to the feature and configuration, assume the handling is undocumented rather than inferring privacy from a microphone icon, an offline success, or an operating-system label.
Use a short decision checklist
Before dictating a real entry, confirm:
If any key detail is missing, choose typed text or a feature whose handling is clearly documented until you can verify it. This approach does not guarantee privacy; it helps distinguish documented local processing from an unverified assumption.