Most voice assistants still feel like glorified search boxes. OpenAI is trying to change that. The company has announced an update that lets ChatGPT users kick off real agentic work through voice commands on mobile, including creating documents, presentations, and spreadsheets, controlling connected apps, and running browser-based tasks, all by speaking.
The update ties Voice directly into ChatGPT Work, which is OpenAI’s productivity-focused mode available on both web and mobile. So instead of typing out a prompt to generate a report or trigger a workflow, you can just say it. And if you end the voice conversation before a task finishes, Work keeps running it in text mode. That handoff between voice and text is genuinely useful for anyone bouncing between their phone and desktop throughout the day.
Plugins and connected apps are also now accessible through Voice across web, iOS, and Android. Users can read written responses alongside the spoken conversation, and existing permission settings carry over, so no new security configurations are required on the developer side. That last part matters because it lowers the friction for teams already using ChatGPT integrations in their workflows.
This builds on GPT-Live, the voice model OpenAI introduced in July. GPT-Live was designed around natural conversation: users can interrupt mid-sentence, ask follow-up questions, and get both spoken and streamed text responses at the same time. It’s a more fluid experience than older voice interfaces, which forced you to wait for a complete response before speaking again.
The broader trend here is worth paying attention to. Google has Gemini with voice capabilities baked into Android. Apple is rebuilding Siri with LLM foundations. Amazon is retooling Alexa. Everyone is racing to make voice a serious input method for complex tasks, not just quick lookups. OpenAI’s move to connect voice directly to agentic and productivity features puts it ahead of most competitors on that specific capability right now.
- Voice now works within ChatGPT Work on web, iOS, and Android
- Supported tasks include document creation, spreadsheet work, browser control, and connected app actions
- Tasks continue running in text mode after a voice session ends
- Plugins and app integrations are accessible through voice commands
- Written responses appear alongside spoken output during conversations
For developers building on top of ChatGPT’s plugin ecosystem, this is a meaningful surface area expansion. Voice becomes another entry point into workflows that were previously text-only. And for end users, it’s the closest ChatGPT has come to functioning like a hands-free work assistant rather than a chat window you have to type into.
