ChatGPT is learning to handle more than conversation. OpenAI is adding its most advanced voice system to ChatGPT’s website and mobile app, giving users a way to speak instructions instead of typing them. The feature supports tasks that can stretch across multiple steps, a browser, and separate apps.
OpenAI announced the addition on September 23, 2026. The release allows users to speak to ChatGPT and delegate single tasks or recurring tasks that require work across different places. That distinction matters because the system is not limited to answering a question inside one chat window; it is designed for requests that involve several actions and tools.
The available facts do not describe a narrow voice mode built only for dictation. They describe a system that lets users speak to delegate work involving a browser and disparate apps, with the same feature handling both one-off requests and tasks that repeat. In other words, the microphone is becoming an instruction channel for multi-step activity—not merely a faster keyboard.
OpenAI has been rolling out launches ahead of its big OpenAI DevDay event next week, placing this voice announcement inside a larger run of product releases. The timing gives the feature a clear role in the company’s current push: ChatGPT is being presented as a tool users can instruct across tasks, rather than a service that only responds with text.
The rollout also puts the focus on access. OpenAI is bringing the voice system to both the website and mobile app, covering the two places where ChatGPT users interact with the service. The announcement does not provide figures or a broader breakdown of availability, so the central fact is simple: the same advanced voice capability is moving into ChatGPT’s main user-facing products.
Atty Eleti, OpenAI’s product lead for ChatGPT voice, gave a demo to Business Insider. That demonstration provides the clearest named view of the feature’s intended use, while OpenAI president Greg Brockman is also identified among the people connected to the company behind the rollout.
Voice interfaces have been promised as a more natural way to use software for years, usually followed by a cheerful assistant that can set a timer and little else. OpenAI’s announcement points to a broader target: users speak a request, and ChatGPT handles work that may require multiple steps, a browser, and separate apps. The useful test will be whether that promise holds beyond the demo.
Based on
