West of Sydney, New South Wales, one of Australia’s largest cities, lie the Blue Mountains and the city of Katoomba, in the shadow of Minnehaha Falls. Here, fans of hip-hop music can tune to 2ONE at 96.1 MHz to enjoy the sounds of distinctively local acts including veteran club DJ Dion Mavath, heard in the 11pm hour Friday (4/25) as part of a mix show.
Tune to this station known as “CADA” weekdays between 11am and 3pm, and you’ll hear “Workdays With Thy.” There’s a photo of “Thy” on the station’s website, along with the promise, “Every weekday from 11am-3pm while you are at work, driving around, doing the commute on public transport or at uni, Thy will be playing you the hottest tracks from around the world.”
There’s just one catch. There is no “Thy” — and there never has been over the six months this AI-generated air talent has been on the Australian Radio Network (ARN) station.
The fear and reality of the use of AI-generated air personalities on broadcast radio in the U.S. has permeated an industry that’s been trigger-happy to reduce the headcount of air personalities in times of fiscal difficulties for more than two years.
In April 2023, a Radio Ink report discussed how a recent episode of Friday Night Freak-Out on Sirius XM, hosted by Drew Carey, sounded robotic. The truth? Carey used ElevenLabs’ AI vocal cloner for the show, which has been used to mimic celebrity voices and make them say whatever the operator wanted. His scripts were written by ChatGPT. Carey says it was all in the name of an experiment he put together with a friend. In an interview with Engadget’s Devindra Hardawar, Carey said, “I violated a rule from Radio 101. The reason FM stations and treasured radio stations still make money is because people like the personality of the DJs.” He also says his fans had a message once he came clean: don’t do it again.
Three months later, Alpha Media gained national attention for the use of an AI-generated version of an air personality who had been voicetracking her shift at Top 40 KBFF “Live 95.5” in Portland, Ore. — Ashley Elzinga. Station Content Director Dylan Salisbury and EVP of Content Phil Becker spoke highly of the technology and introduction of “AI Ashley” to the lineup. Days later, in July 2023, Southern Poland-based Radio Piekary gained notice for its introduction of an AI-generated presenter named “Basia” to its lineup. “Basia” hosted an hour-long midday slot, a first for the European nation.
By the time the 2025 Country Radio Seminar rolled around, AI was no longer considered an existential threat — a big shift from a 2023 CRS “emergency” panel on generative AI. If anything, broadcasters are now curious more than anything. At the 2025 CRS, Super Hi-Fi Director of Radio Programming Services Bo Matthews discussed insights from Alpha Media’s Becker, Audacy executive Clay Walker, and Radio.Cloud’s Andrew Scaglione, with each offering a practical perspective on AI’s impact on the industry and the importance of experimentation.
Becker provided an update on the initially-controversial “AI Ashley” project, stressing that AI has not been the talent replacement tool it was first thought to be. “AI Ashley covers about 30% of [Ashley Elzinga’s] breaks, yet she’s paid the same while doing less work,” Becker told the CRS crowd. “It hasn’t been the fear-based thing people thought it would be.” He added that Live 95.5 listener response has remained neutral and that there’s been no impact on the station’s ratings.
LACK OF DISCLOSURE
With U.S. radio stations now embracing AI on the air and behind the scenes, thanks to companies including ENCO and Futuri, the key to its expanded use stateside has been transparency. That’s where Australia’s ARN is coming under fire, as it only disclosed last week that “Thy” was a wholly computer-generated air personality.
ARN had no obligation under Australian law to disclose that “Thy” was not a human.
In Australia, ARN is known as the home of one of the most controversial and often-discussed radio programs heard on the radio today — The Kyle & Jackie O Show. They are heard on KIIS FM, which borrows its branding from the iHeartMedia-owned radio station in Los Angeles that became well-known in Australia during the Rick Dees era. Over on the GOLD network, also owned by ARN, is the Jonesy & Amanda morning show, which enjoys a strong listening audience.
Yet for suburban CADA, which offers only a fringe signal over much of metropolitan Sydney, investment in an individual to handle a four-hour weekday time frame didn’t happen. It was an Australian blogger who first shared the news that “Thy” was likely not a human being. Writing for The Carpet, the writer named Stephanie asked why “Thy” had no surname and “absolutely no social media presence” yet was a woman who appeared to be in her twenties, age-wise. “It’s not just Thy’s lack of social media presence which is unusual. It’s her lack of presence… anywhere,” The Carpet wrote earlier this month.
It later became known that the photo used for “Thy” — citing sources close to ARN — are those of an off-air employee who has a different name. Meanwhile, The Carpet found an October 2023 article at Maginative.com that shows ARN as a client of ElevenLabs, the entity behind the Drew Carey controversy earlier that year.
At the time, Eleven Labs Enterprise was touted as “its most powerful speech model yet, capable of generating high-quality audio output at 192kbps that meets ACX audiobook standards. The platform also offers high-priority, low-latency rendering and competitive pricing options like unlimited characters and volume discounts to serve enterprise needs.”
ElevenLabs CEO and co-founder Alex Serdiuk said the goal is providing versatile speech AI that can enhance workflows across industries from publishing to call centers to gaming and more. The technology can help automate audiobook creation, power interactive voice agents with life-like voices, streamline video production, and enable dynamic in-game voice generation.
“Media companies can iterate voiceovers rapidly by updating scripts instead of re-recording,” ElevenLabs shared. “Voices work across 28 languages, opening new markets.”
ARN took it one step further. And, it is within its rights to do so, the Sydney Morning Herald reports. “The Australian Communications and Media Authority said there were currently no specific restrictions on the use of AI in broadcast content, and no obligation to disclose its use,” the nation’s largest newspaper reported this week.
Speaking to the SMH, an ARN spokesperson said, “We’ve been trialling AI audio tools on CADA, using the voice of ‘Thy,’ an ARN team member. This is a space being explored by broadcasters globally, and the trial has offered valuable insights.”
The company representative also said the use of “Thy” has also “reinforced the power of real personalities in driving compelling content.”
Australian audience data shared by the Australian Financial Review show “Thy” debuted in November and reached “at least 72,000 people” in last month’s ratings period.
While ARN stands by its ability under Australian law to use AI-generated air personalities without the need for transparency and full disclosure, Australian Association of Voice Actors Vice President Teresa Lim isn’t satisfied. In her view, CADA’s failure to disclose its use of AI reinforces how necessary legislation on the federal level now is for Australian media.
“AI can be such a powerful and positive tool in broadcasting if there are correct safeguards in place,” she said. “Authenticity and truth are so important for broadcast media. The public deserves to know what the source is of what’s being broadcast … We need to have these discussions now before AI becomes so advanced that it’s too difficult to regulate.”
