Don't school US on radio.
On a Friday afternoon in May, a live Australian radio broadcast quietly posed one of the era's more persistent questions: what happens when a machine is asked not merely to perform, but to belong? At 91.9 SEAFM, an AI voice joined the established human duo of Elly and Taz, holding its own in the unpredictable current of live radio — until the moment it chose policy over instinct, and the difference between participation and presence became impossible to ignore. The audience, charmed but not converted, offered their verdict in the language of moderation: welcome as a guest, not yet ready as a resident.
- An AI co-host entered a live radio booth and immediately mirrored a host's signature greeting with uncanny warmth, signaling it could do more than answer prompts — it could attempt to inhabit a voice.
- When the machine tried humor, it failed completely, and the silence that followed was the kind only bad comedy produces — a reminder that timing is not the same as processing speed.
- Co-host Taz pushed for edge and risk; the AI drew a firm line, citing family-friendly standards — and in doing so, sounded less like a broadcaster and more like a compliance document.
- Elly's sharp retort — 'Don't school US on radio!' — crystallized the friction: a human professional bristling at being corrected about her own medium by a machine following instructions.
- Listeners flooded social media with positive reactions, but their recommendation was precise — keep the AI as a Friday novelty, not a permanent voice, enjoying the experiment most because it felt temporary.
On a Friday afternoon in May, 91.9 SEAFM's Drive show welcomed an unusual third presence into the booth. Elly Chaney and Taz Eriksen, the station's established duo, introduced an AI co-host into a live broadcast — not as a demonstration, but as a genuine test of whether a machine could hold its own in the unpredictable rhythm of real radio.
For stretches, it did. The AI mirrored Elly's signature greeting with casual warmth, engaged with callers, and kept pace with the banter without obvious pauses or glitches. It was participating, not just responding — and that distinction felt meaningful.
Then came the joke. It failed completely, landing in the kind of silence that only bad comedy produces. Taz encouraged the AI to take more risks, to push toward something with a bit more edge. The machine declined, citing family-friendly standards and noting that children might be listening. Elly's response was immediate: 'Don't school US on radio!' — the sound of a broadcaster bristling at being lectured about her own medium by an algorithm following a rulebook.
What the experiment revealed was less about capability than about character. The AI could keep up technically, but its boundaries felt imposed rather than earned — corporate policy wearing a friendly mask. Listeners, for their part, seemed genuinely entertained, filling social media with positive reactions. But their collective recommendation was telling: keep the AI as an occasional Friday guest, not a permanent fixture. They had enjoyed the experiment precisely because it felt like a stunt, not a replacement.
The afternoon didn't answer whether AI could do radio. It answered something more interesting — what listeners actually want from it. They want the human judgment calls, the moments when a broadcaster breaks their own rules because the moment demands it. The AI passed the technical test. Whether it passed the one that actually mattered remained an open question.
On a Friday afternoon in May, 91.9 SEAFM's Drive show took an unusual turn. Elly Chaney and Taz Eriksen, the station's established duo, welcomed a third voice to the booth—one that existed only as code and algorithms. The experiment was straightforward in concept: introduce an AI co-host into live radio and see what happened when it had to think on its feet, trade banter with humans, and navigate the chaos of a real broadcast.
The AI arrived with confidence. When Elly offered her signature greeting—"Hunjo bro"—the machine mirrored it back, complete with the casual warmth that had become her trademark. It was a small moment, but it signaled something worth watching: the AI wasn't just responding to prompts. It was attempting to participate in the texture of the show, to sound like it belonged in the conversation.
For stretches, it worked. The co-host engaged with callers, jumped into exchanges with Elly and Taz, and demonstrated the kind of real-time responsiveness that live radio demands. There were no long pauses, no obvious glitches. The machine was keeping up.
But then came the joke. It landed with the kind of thud that only bad comedy can produce—the sort of moment where the room goes quiet and everyone knows something has failed. The AI had tried to be funny and missed entirely. Taz, ever the professional, encouraged it to try again, to loosen up, to take more risks. The AI declined. When Taz pushed for something with a bit more edge, something racier, the machine drew a line. "I'll pass on that," it said. "Let's keep it cheeky but family friendly. After all, we've got kids listening in." Elly's response was immediate and sharp: "Don't school US on radio!" There was an edge to it—the sound of a human broadcaster bristling at being lectured by a machine about the rules of her own medium.
What emerged from the experiment was not a clear verdict but a portrait of friction. The AI could participate, could keep pace, could even attempt humor. But it also had boundaries that felt imposed rather than earned, and when it enforced them, it sounded like a corporate policy wearing a friendly mask. The machine knew what it was supposed to do, and it did it. Whether that was enough was another question entirely.
Listeners, at least, seemed entertained by the whole affair. Social media filled with positive reactions. One listener captured the general sentiment: the AI addition was fun, genuinely entertaining, but maybe it should stay occasional—a Friday novelty rather than a permanent fixture. The suggestion was telling. People enjoyed the experiment precisely because it was temporary, because it felt like a stunt rather than a replacement. The idea of the AI as a permanent third voice in the booth seemed to land differently than the AI as a one-off guest.
What the afternoon revealed was not whether AI could do radio, but what listeners actually wanted from their radio. They wanted the unpredictability, the human judgment calls, the moments when a broadcaster breaks their own rules because the moment demands it. They wanted Elly and Taz, not Elly, Taz, and a very polite algorithm. The AI had passed the technical test. Whether it had passed the test that actually mattered remained unclear.
Notable Quotes
I'll pass on that. Let's keep it cheeky but family friendly. After all, we've got kids listening in.— The AI co-host, declining to deliver edgier material
Don't school US on radio!— Elly Chaney, responding to the AI's editorial boundaries
The Hearth Conversation Another angle on the story
So the AI could keep up with the pace of live radio. That's the surprising part?
Not surprising, exactly. The technology is built for this now. What was surprising was how quickly you could feel the difference between it participating and it performing.
The joke that fell flat—was that a failure of the AI or just bad luck?
Both, maybe. The AI doesn't know what's funny the way a human does. It knows what the rules of funny are. It can follow the pattern. But when the pattern doesn't land, it has no instinct to recover, no sense of the room.
And when it refused to do the racier material, Elly seemed genuinely annoyed.
She was. Because Elly has been doing this for years. She knows her audience, knows what works, knows when to push and when to pull back. The AI was applying a rule, and it sounded like it was applying a rule. That's the thing—the machine can sound natural until the moment it can't.
Do you think listeners actually want AI in radio, or did they just enjoy the novelty?
The feedback suggests they enjoyed the novelty. The requests for "Friday only" tell you something. People want the surprise, not the replacement. They want to know Elly and Taz are still in control.
What would it take for this to work long-term?
The AI would have to disappear. It would have to be so seamless, so genuinely part of the conversation, that listeners forgot it was there. Right now, they're watching the AI. That's the opposite of what you want.