Does voice typing save your audio recordings?

Voice typing saves your audio only when the company that recognizes your speech keeps the recording. Many tools do not keep it by default, but some keep audio for a set period, and a few keep it until you delete it. The answer therefore depends on three things: where recognition runs, whether an improvement setting is on, and what the product's policy says about the chat or history that holds the audio.

Apple and Windows keep audio only when you allow it

Apple's rule is an opt-in. Dictation sent to Apple's servers is not stored unless you opt in to Improve Siri and Dictation ¹. That setting lets Apple store and review a sample of your audio ². Leave it off unless you want to help improve Apple's recognition.

Windows voice typing uses online speech recognition. Microsoft says that mode sends voice data to Microsoft only to provide the service and create transcriptions, and that it does not store, sample or listen to recordings without your permission ³. Voice typing therefore sends your speech to Microsoft, but the stored-audio question is answered by Microsoft's permission rule, not by the voice typing itself.

Chat-based dictation keeps audio with the conversation

AI dictation inside a chat app works differently from a system keyboard. When you press the microphone icon in ChatGPT, the recorded audio is sent to OpenAI's models to be transcribed ⁴. The audio is kept for as long as the chat stays in your chat history ⁵. Deleting the chat is the control that matters. The question on how to delete voice recordings lists the steps for each company.

Anthropic's mobile apps work the other way. After converting speech to text, Anthropic says it deletes the audio recording, and does not keep a copy of the speech or use it for training ⁶.

Claude Code is different again. Its voice dictation streams the recorded audio to Anthropic's servers for transcription ⁷. The documentation we read does not say how long Anthropic keeps that audio. Treat that as an open question, not as a no, and do not send confidential material through it until you have an answer in writing.

Google's saving setting is off unless you turn it on

Google's rule is a default-off switch. Google says its voice and audio activity setting is off unless you choose to turn it on ⁸. So for a Google account, the answer depends on whether the setting has been switched on at some point, and it is worth checking in your account settings.

Where recognition runs decides what can be stored at a vendor

Storage and location are separate questions. If recognition runs on your device, the audio does not reach a vendor's server for that step, so there is nothing for the vendor to keep. The airplane test, described in the guide linked below, tells you whether a given tool recognizes speech on the device. If recognition runs on a server, the vendor's storage rules apply.

Table 1. Storage of dictated audio in common tools, as of October 2026

Table 1 shows that the same feature can behave differently in two products. Read the setting or policy for the product you use, because a dictation feature inside a larger app usually inherits that app's rules.

What to do if you need a firm answer

1. Dictate one sentence with Airplane Mode on to confirm where recognition runs.
2. On Apple devices, keep Improve Siri and Dictation off unless you have chosen to share samples.
3. On Windows, remember that voice typing needs online speech recognition. Use it knowing that audio goes to Microsoft, or use a tool that runs on the device.
4. Open the product's history or chats and delete audio you do not need.
5. Read the storage section of the policy. The checklist for a dictation app's privacy policy shows what to look for.

Voice typing is not automatically saved, and it is not automatically private either. The guide to whether voice to text is safe covers the wider picture, and the glossary entry on cloud dictation explains the server case. Run the same tests on Yap, which we make, before you rely on it for anything sensitive.