ChatGPT's mobile voice mode can now act — invoke connected apps mid-conversation and start agentic ChatGPT Work tasks (draft docs, summarize Slack, build spreadsheets) by voice on iOS and Android.
Last verified · 2026-09-23 · by Moe Ameen
ChatGPT mobile voice agents are the agentic voice capabilities OpenAI added to the ChatGPT phone app on September 23, 2026. Before this, voice mode was a spoken conversation — you talked, it answered. Now the voice layer can act: it can invoke connected apps during a conversation and, on paid plans, kick off ChatGPT Work tasks entirely by voice.
Two things shipped together. First, plugins came to the Live voice experience across web, iOS, and Android, so you can call a connected app — email, calendar, Slack, and others your plan supports — while you're mid-conversation, with written responses shown alongside the audio. Second, Voice arrived in ChatGPT Work on web and mobile, so a spoken request can drive OpenAI's cloud-connected agentic system: finding information in connected services, preparing documents, drafting emails, building spreadsheets and presentations, and handling tasks in the browser.
Access is tiered. Plus and Pro subscribers can open the Work tab on a phone and run those agentic tasks by voice; Free and Go users get plugins and connected apps in voice, without the full Work surface. Actions that need confirmation surface an on-screen approval (voice approval isn't supported), existing app permissions and usage limits apply, and a voice-started Work task that outlasts the call continues in text mode. OpenAI also said voice output now carries richer text, that you can switch between text and voice easily, and start on a phone then resume on desktop. ChatGPT Voice can now run on OpenAI's GPT-6 models — Astra, Sol, and Luna. Plan details and limits shift quickly; confirm what your tier includes in OpenAI's release notes before you build a routine around it.
Think of ChatGPT mobile voice agents as the fastest idea-capture studio you own: on a walk you can talk out ten hooks, a full video script, and a newsletter angle without touching the keyboard, and the agent drafts them cleanly. What it cannot do is turn any of that into content your audience sees — no video gets filmed, no carousel gets designed, nothing gets scheduled or published. [Kompozy](/) is the engine that takes the voice-drafted raw material and manufactures the finished, on-brand output.
The most powerful pairing is video you never have to film. Dictate a script into ChatGPT on your phone, drop it into Kompozy, and it rebuilds the take as a talking-head Persona Short or a longer Persona HeyGen avatar video in your own face and voice — no camera, no studio. From that same spoken source it also spins Carousel Posts, Photo Posts and Quote Graphics, a Blog Article, an Email Newsletter, and native Text Posts, all governed by one Persona Brief so every format sounds like you. A per-post review gate lets you approve each piece, then [Autopilot](/glossary/autopilot) schedules and publishes the batch across the eight social platforms plus blog and email. Voice is how the idea leaves your head; Kompozy is how it reaches every platform. See [AI voice generation](/glossary/ai-voice-generation) for where the spoken layer fits, and [content repurposing](/glossary/content-repurposing) for the one-idea-to-many-formats workflow.
On September 23, 2026, OpenAI brought plugins to the Live voice experience on iOS and Android (so you can invoke connected apps like email, calendar, and Slack mid-conversation) and added Voice to ChatGPT Work on mobile, so Plus and Pro users can start agentic tasks by voice — drafting documents and emails, summarizing Slack, building spreadsheets and presentations, and running browser tasks. Free and Go users get plugins and connected apps in voice, but not the full Work surface.
No. The agentic tasks are office deliverables and connected-app actions — documents, emails, spreadsheets, summaries, browser errands. Voice agents do not generate finished multi-format content or publish to social platforms. To turn a voice-drafted script into video, carousels, a blog, a newsletter, and scheduled posts, you bring the text into a content engine such as Kompozy.
Partly. Plugins and connected apps in voice conversations are available to Free and Go users on the plans that support them. The full ChatGPT Work agentic task surface by voice — creating documents, spreadsheets, presentations, and using the cloud browser — is for Plus and Pro subscribers. Voice-started Work tasks consume the same usage as typed tasks and follow your plan's voice limits.
Not for consequential actions. You can speak requests and have the agent act, but anything requiring confirmation shows an on-screen approval — OpenAI has said voice approval isn't supported there — so you still tap the phone. And a Work task that runs past the end of the voice call continues in text mode rather than staying voice-driven.