Recording access or input
Browser or app microphone permission and the operating system's input selection affect recording.
First check that you are using the microphone icon for dictation and allow microphone access when prompted. Dictation records a prompt and returns editable text; Voice is a live conversation. ChatGPT's dictation FAQ does not rank failure causes or provide a dedicated repair checklist. The checks below combine its documented workflow with explicitly identified Voice and operating-system microphone guidance.
Browser or app microphone permission and the operating system's input selection affect recording.
Dictation produces editable text before sending. Voice is a live conversation, so its controls and guidance have a different scope.
The Voice guidance identifies overlapping speech, noise, microphone settings and network conditions as factors.
Use the dictation FAQ to confirm the record, review and send workflow, and the linked Voice guidance for the audio checks it actually documents. There is no dedicated dictation repair procedure in the linked help documentation, so these checks should not be read as a confirmed fix for every failure. If you need another way to compose prompts, WhisperJot can transcribe locally and enter text into the focused app. That avoids ChatGPT's dictation transcription dependency; sending prompts to ChatGPT still needs its service.
WhisperJot is a separate dictation app with one hotkey in any app. Jot Local and Jot Local Pro use local engines by default and work offline after a one-time model download. Choose Jot Local for fast, short utterances or Jot Local Pro for long-form accuracy. This provides an alternative way to enter text; it does not repair built-in dictation. A working microphone and the permissions needed to record are still necessary.
No. Dictation records an audio message and returns text you can edit before sending it as a prompt. Voice provides a live conversation. The help center's audio troubleshooting guidance is for Voice; this guide labels that scope wherever it applies those checks.
The linked documentation does not identify an update-specific dictation failure or fix. Confirm the microphone control, recording permission and operating-system input, then consider noise and network conditions. These are documented workflow and audio checks, not evidence that an update caused your problem.
Its dictation FAQ says recorded audio is sent for transcription and the result is returned as editable text. That workflow needs a connection. WhisperJot's downloaded local engines can compose text offline, but they do not make the ChatGPT service available without internet.
Official documentation checked 2026-09-05:
Microphone not working for dictation? ยท How to use voice typing ยท System requirements