Across the globe, radio is discovering that the voice in your car might not belong to a person at all.

From experimental late-night music programs in the United States to automated news bulletins in parts of Europe and Asia, broadcasters are quietly testing artificial intelligence-generated hosts. The pitch is familiar: AI can run around the clock, never calls in sick, and can tailor playlists and patter to hyper-specific audiences.

In Australia, that experiment has already crossed from theory into practice.

Last year, a Sydney-based station revealed it had used an AI-generated voice named Thy to host a regular weekday music show without informing listeners.

For six months.

At the time, no rule required disclosure. For many in the industry, it was a proof of concept. For others, it was a warning shot.

Honestly? Most of the time, you wouldn't be able to tell the difference. Here's what it sounded like.

Now, the regulator has stepped in.

According to a story from the Australian Associated Press, commercial radio stations will soon be required to disclose when a regularly scheduled music or news program is hosted by a computer-generated voice. The change is part of the Commercial Radio Code of Practice 2026, unveiled by the Australian Communications and Media Authority this week.

Under the revised code, which takes effect on July 1, stations must tell audiences if a synthetic voice is hosting a regular program. The disclosure can be made during the broadcast itself or published on the station’s website or social media channels.

But the rules come with notable carve-outs. AI-generated voices used in advertisements, weather and traffic updates, and music segments will not require disclosure. Nor will AI content delivered via streaming platforms. That distinction reflects the regulator’s focus on traditional broadcast radio, even as listening habits shift increasingly online.