Talk, Don’t Type: How to Use ChatGPT Voice to Check Your Email, Calendar and Slack
Image: OpenAI, from the official ChatGPT Voice feature page (chatgpt.com/features/voice).
ChatGPT Voice just got a lot more useful. On September 23, 2026, OpenAI added plugin support to Voice, which means you can now ask ChatGPT out loud to look at your inbox, check your calendar or search Slack, as long as you’ve connected those apps. Paid users with ChatGPT Work can also start longer jobs by voice, such as drafting a document or building a spreadsheet, and let them keep running after the call ends.
Until now, voice chat was mostly for questions and brainstorming. Now it can reach your actual tools. This guide covers what changed, who gets it, how to set it up on your phone or the web, and the limits you should know about first.
What changed on September 23
OpenAI’s ChatGPT release notes list three changes for September 23, 2026:
- Plugins in Voice. The “Live” voice experience now supports plugins on web, iOS and Android. According to OpenAI, you can use “the plugins and connected apps available to your account during a Voice conversation and follow written responses in the chat.”
- Voice in Work on web and mobile. Voice was already available in the ChatGPT desktop app’s Work and Codex modes (added in July). It now also works in Work on the web and on phones. You can ask it to create documents, presentations and spreadsheets, use connected apps, or work in a browser.
- Tasks that outlive the call. If you end a Voice call in Work while a task is still running, it can continue in text.
The update comes one day after OpenAI released GPT-6 Sol and GPT-6 Luna for ChatGPT Work and Codex (September 22). Earlier in September, OpenAI also let Voice hand harder questions to GPT-5.6 or GPT-6 Astra when it needs to search or reason. In short, the voice you talk to is increasingly backed by the same models and tools as typed ChatGPT.
Who can use it
This is where plans matter, so here is what OpenAI’s help pages say:
| Feature | Who gets it |
|---|---|
| Voice in Chat with plugins | All consumer plans, including Free and Go, “with the plugins their plan supports” |
| Voice in Work (web and mobile) | Plans that include both Voice and Work access (eligible paid plans) |
| Voice in Work and Codex on desktop | ChatGPT desktop app on macOS and Windows, where included in your plan |
| Business, Enterprise, Edu, Healthcare | Available in eligible workspaces, subject to admin settings |
Daily Voice limits for Chat are measured over a rolling 24 hours. According to OpenAI’s Voice help page, Go gets 3 hours with GPT-Live-1 mini, Plus gets 3 hours with GPT-Live-1, Pro ($100/month) gets 15 hours and Pro ($200/month) gets unlimited GPT-Live-1. Free users get “limited access” to GPT-Live-1 mini. Voice in Work on web and mobile uses minutes from the same daily Voice allowance, and the tasks themselves also count toward your normal Work usage.
How to set it up: step by step
Step 1: Update the app and choose "Live"
The update is rolling out through the latest version of ChatGPT, so update the iOS or Android app first. Then open Settings → Voice. You may see three options:
- Live: the new real-time experience (GPT-Live-1 or GPT-Live-1 mini). This is the only mode that uses plugins.
- Advanced: the previous real-time voice. Keep it for video or screen sharing, which Live doesn’t support.
- Standard: turn-by-turn voice that transcribes your speech before answering.
Choose Live. While you’re there, set your Language to the one you speak most. OpenAI says this helps ChatGPT understand you more accurately.
Step 2: Connect the apps you want Voice to use
Voice can only use apps you’ve already connected. It doesn’t gain new access just because you’re speaking. OpenAI’s help page states that “Using Voice does not change your access to plugins or connected apps.”
To connect an app:
- Open Plugins (or Apps, depending on your account) in ChatGPT.
- Select the plugin you want, for example your email, calendar or Slack.
- Select Install plugin or Connect, then sign in to the provider account.
Since September 17, you can also connect multiple accounts to plugins, starting with Gmail, Google Calendar and Google Contacts. That means a personal and a work account can be available in the same conversation.
Image: OpenAI, from the official ChatGPT plugins feature page (chatgpt.com/features/plugins).
Step 3: Check your permission setting before you start
This step is easy to miss, and it matters more once you’re talking instead of reading. OpenAI’s release notes say that if you haven’t changed your plugin settings, the default is “Allow low-risk actions.” ChatGPT can then take actions it considers low-risk without asking each time, but it will still ask before higher-risk actions.
If you’d rather approve everything yourself, review this under Settings → Plugins before you start using Voice with your inbox.
Step 4: Start a Voice conversation
- On iPhone or Android: tap the Voice icon in the message bar, allow microphone access if asked, pick a voice the first time, and start talking.
- On the web: go to chatgpt.com, select the Voice icon in the prompt window, and allow microphone access.
- For bigger tasks (paid plans): switch to Work first, then select the Voice control.
Image: OpenAI, from the official ChatGPT Voice feature page (chatgpt.com/features/voice).
Step 5: Approve actions on screen
If an action needs approval, such as sending something on your behalf, Voice will ask you to review it on screen. Saying “yes” out loud won’t work. OpenAI’s help center states that “spoken approval is not supported.” This is a sensible safeguard, but it means Voice isn’t fully hands-free for actions that change something. You’ll need to glance at your screen and tap.
Things worth trying
These are example requests based on the capabilities OpenAI describes. We haven’t tested them ourselves, and what works depends on your plan and which apps you’ve connected.
For your morning (Chat mode, any plan with the relevant plugins): – “What’s on my calendar today, and do any meetings overlap?” – “Do I have any unread emails from my manager since yesterday?” – “Search Slack for anything about the launch date this week.”
For bigger jobs (Work mode, paid plans): – “Draft a one-page summary of the emails about the Henderson project and put it in a document.” – “Turn the numbers I just read out into a spreadsheet with a monthly total.” – “Build a short slide deck from my notes on the Q4 plan.”
With Work, you can end the call and let the task finish, then come back to the result in the chat. Because Live shows its responses as text while it speaks, you also get a written record to check later.
The limits to know before you rely on it
OpenAI’s documentation is candid about what Live can’t do yet:
- No video or screen sharing in Live. Use Advanced for that (on iOS and Android, for eligible subscribers).
- No ChatGPT Library files. Live “cannot currently find or add files from your ChatGPT Library,” though you may be able to attach a file to the chat manually.
- One conversation at a time. You can only have one Voice conversation running.
- Not built for groups. OpenAI says Live “is not yet optimized for conversations with multiple speakers” and may respond when people are talking to each other.
- Transcripts aren’t verbatim. OpenAI warns that Voice transcripts may not exactly match what was said, especially with background noise or overlapping speech.
- Dates and time zones. Voice uses your device or browser time zone to understand “today” or “tomorrow.” If a calendar answer looks wrong, check your time zone or say the exact date.
Privacy: what happens to your voice
Before you connect your inbox to a voice assistant, it’s worth knowing how the audio is handled. Per OpenAI’s Voice help page:
- Audio clips from Live and Advanced conversations are stored with the chat transcript and retained for 30 days.
- Deleting a chat also deletes its audio within 30 days, with exceptions for security, safety or legal reasons.
- OpenAI says it doesn’t train on your audio clips unless you choose to share them. But if “Improve the model for everyone” is on, transcripts from Voice conversations may be used for training, depending on your plan and settings. You can check this under Settings → Data Controls.
- Business, Enterprise and Edu users can’t share audio clips for training.
If you’re connecting work email or Slack, check your company’s AI policy first. In managed workspaces, admins decide whether Voice, Work and specific plugins are available.
Why this matters
For everyday users, Voice with plugins turns ChatGPT from something you talk to into something that can do small jobs for you while your hands are busy: when you’re commuting (ChatGPT also supports Apple CarPlay), cooking, or walking between meetings. For people on paid plans, being able to start a Work task by voice on your phone, hang up, and find a finished document later is a real change in how the tool fits into a day.
For businesses, this is a reminder that permissions deserve attention. Voice uses the same app connections and action rules as typed ChatGPT, so an agent that can read your inbox by voice is only as safe as the permissions behind it. The “Allow low-risk actions” default and the on-screen-only approval rule are the two settings every team should understand before rolling this out. If you’re planning voice or agent features for your own product, our AI Solutions & Automation team works on these kinds of permission-aware integrations.
What happens next
OpenAI has been adding voice features steadily: GPT-Live-1 in July, Voice in the desktop app’s Work and Codex modes the same month, Lock Screen live activities in August, and model and limit changes on September 9. Plugin support in Live is the latest step. Based on the gaps OpenAI lists itself, likely next steps include support for ChatGPT Library files and video or screen sharing in Live. OpenAI hasn’t announced dates for either.
Final takeaway
If you already use ChatGPT with Gmail, Google Calendar or Slack connected, try this today: update the app, switch Voice to Live, check your plugin permission setting, and ask about your day. Keep in mind that anything that sends or changes data still needs a tap on the screen, and that your transcripts follow your data-control settings. Used with those limits in mind, it’s one of the more practical AI updates this month.
Sources: OpenAI Help Center (ChatGPT Release Notes, ChatGPT Voice, ChatGPT Work and Codex, Plugins in ChatGPT and Codex), OpenAI’s ChatGPT Voice feature page, TechCrunch, 9to5Mac and The Decoder. See sources.md for full links.