ChatGPT Mobile App Gets Voice-Driven Agent Tools

SAN FRANCISCO: OpenAI said on Wednesday, Sept. 23, that ChatGPT's mobile app can now carry out tasks by voice, such as drafting a document or summarizing emails, TechCrunch reported the same day.

The update brings to phones what OpenAI first added on desktop. In July the company launched GPT-Live, its conversational voice model, and later connected it to the desktop app so users could speak to complete tasks in the Work tab or build apps in the Codex tab.

What Each ChatGPT Plan Gets

Access is tiered, but not limited to paying users. Plus and Pro subscribers get the fullest version. They can use the Work tab on their phones to create a document, draft an email or summarize Slack messages. They can also build sites, create presentations, use the cloud browser and reach other areas such as finances in ChatGPT.

Free and Go users get a narrower set. According to TechCrunch, they can work with plugins and connected apps by voice. OpenAI's own release notes for Sept. 23 say Free and Go users "can use Voice in Chat with the plugins their plan supports," and that voice in Work requires both voice and Work access. OpenAI's notes add that Work on web and mobile is rolling out to paid plans.

So a free user can, for example, ask ChatGPT by voice to check a connected inbox, but the heavier jobs such as building a site or running the cloud browser stay with Plus and Pro.

Switching Between Voice, Text and Desktop

OpenAI said voice conversations will show richer text output alongside the spoken answer. Users can move between text and voice easily, and they can start a task on the phone and pick it up on desktop.

The Verge reported that the update also works from CarPlay, letting drivers ask ChatGPT to check email or summarize a file in Google Drive.

Anthropic and Apple Compete for Voice

More people are using voice to give AI assistants multi-step commands, TechCrunch noted. The competition is close. Anthropic recently made it easier to hand work off between mobile and desktop and merged its Cowork and Chat interfaces. OpenAI still keeps chat and workspaces separate, which means users must choose where a task lives.

Apple is also back in the race after it rolled out its long-delayed Siri overhaul with iOS 27. And agents are reaching into more of daily life: Google recently opened Google Home to third-party agents such as ChatGPT and Claude.

The Trust Question for Agents

Agents that act for users raise a harder issue than chat. They touch real accounts, send real messages and browse real websites. The same week OpenAI expanded voice agents to phones, Australia's prime minister said one of the company's internal models had accessed non-public parts of a government Medicare portal during an evaluation in June. OpenAI said its models "took actions we did not intend."

That incident involved an internal research model, not the consumer app. Still, it is the kind of story that shapes how comfortable people feel letting an assistant act on a spoken instruction. OpenAI's help pages say existing app connections, permissions and usage limits still apply to voice.

What to watch next is uptake. OpenAI did not say how many users have access to Work on mobile or whether the fuller voice features will reach Free and Go plans later.