Your morning commute, José 97.5 FM on, two voices trading takes on the Dodgers or the latest immigration enforcement news. One voice belongs to Geraldine “GeeGee” Guzman, a real person with real opinions. The other is Coyotec, a synthetic co-host using a cloned voice, sounding confident and conversational. Entravision says the show is growing fast, and human radio workers watching a sector that has shed roughly 19,000 jobs since 2019 are paying close attention.
What Coyotec Actually Does
The mechanics behind the show reveal a human-AI collaboration that is more constrained than it sounds.
Coyotec is not a robot DJ spinning tracks on its own. It is a conversational AI using a synthetic voice built on ElevenLabs technology, modeled on the voice of Entravision Audio president Eduardo Maytorena. The character’s visual identity depicts a working-class Latino man in his 30s, complete with mustache, leather jacket, and hat. Here is how the production actually works:
- Guzman uploads news topics and discussion points; Coyotec generates responses and opinions from that material
- The AI cannot sustain conversation much longer than roughly 10 minutes before performance deteriorates, so Guzman records the show in four separate segments
- Topics have included local elections, the Dodgers, Hispanic Heritage Month, and immigration raids
- Despite being programmed to comply with broadcast standards, Coyotec occasionally swears or runs dry mid-segment, requiring Guzman to intervene
- Entravision identifies the AI’s ability to deliver verbatim advertising pitches as a significant commercial benefit
The show, Al Aire y Sin Permiso, launched with Coyotec in September 2025 as a weekend program. It moved into weekday morning drive time shortly after.
Audience Claims Deserve a Closer Look
Entravision’s growth numbers are worth noting, but they require careful context before drawing conclusions.
Entravision reports a 75% increase in weekly cume among Hispanic men ages 25 to 54 on José 97.5 between March and April. That figure comes from Entravision, not from an independent ratings audit. The available reporting does not isolate Coyotec as the cause: programming changes, station promotion, Guzman’s own appeal, and drive-time scheduling all contribute to audience numbers.
What is verifiable is that the format is expanding. Entravision has already moved it to Sacramento, Modesto, Palm Springs, Las Vegas, El Paso, and McAllen. The company is also adapting the show for English-language and bilingual audiences, which signals that Entravision views Coyotec as a scalable programming product rather than a single-market experiment. That distinction carries real consequences for everyone working in radio.
Who Pays the Price
The labor stakes are not hypothetical; they are already visible in the industry’s recent history.
U.S. radio broadcasting employment fell from roughly 85,000 jobs in 2019 to about 66,000 in 2025. That is a 22% decline, according to Bureau of Labor Statistics figures cited by the Los Angeles Times. AI did not cause all of those losses: advertising contraction, automation, and consolidation were already reshaping the industry before any synthetic DJ took a mic.
John DeSantis, a former afternoon host at Inland Empire station KCAL-FM, put it plainly: “If they expect to achieve the authentic connection that live and local DJs create, then there’s no chance.” DeSantis lost his job when KCAL-FM dropped hosts for an all-music format, a decision that had nothing to do with Coyotec. His concern about AI reflects a pattern of displacement that predates synthetic voices.
SAG-AFTRA has been consistent: synthetic performances require consent, bargaining, and compensation. Broadcasters should not treat AI as a loophole around those obligations. iHeartMedia’s “Guaranteed Human” campaign, whatever its marketing motives, reflects a real emerging divide. Some stations foreground human authenticity; others are betting that audiences either cannot tell the difference or do not care.
The Trust Problem Nobody Wants to Name
Disclosure of an AI’s identity is necessary, but it is not the same as disclosure of its limits.
Synthetic-media scholar Sam Gregory has warned, as paraphrased in the Los Angeles Times, that listeners may treat a polished AI voice as a trusted friend even though the system has no lived experience, no personal stake, and no accountability when it gets something wrong. José FM does disclose that Coyotec is AI, repeatedly and openly. That matters, but it is not the whole story.
When Coyotec delivers a confident take on immigration enforcement or a local election, it draws on material Guzman’s production team selected and uploaded. The available reporting does not show that Coyotec independently verifies sources or bears editorial responsibility for its output. Listeners deserve to understand that gap, not just the label.
What Responsible Scaling Actually Requires
Several other broadcasters are already following Entravision’s lead, which makes industry-wide standards urgent.
Entravision is not the only broadcaster moving this direction. Alpha Media tried a cloned AI DJ in Portland in 2023. A Waterloo, Iowa, rock station uses synthetic DJ Tori for overnight shifts. Phoenix’s Zone has launched an AI-infused format using licensed clones of veteran radio personalities. SonicTrek.ai offers AI-hosted music programs for syndication, including Phoebe FM, though the full geographic reach of that service has not been independently confirmed.
Before this format scales further, broadcasters owe their workers, their listeners, and their communities clear answers on four specific questions:
- Who consented to having their voice cloned, and on what terms?
- Who is editorially responsible when a synthetic host says something inaccurate or prohibited?
- How are listeners informed about the sourcing and limitations of AI-generated opinions?
- When does union engagement happen: before expansion, or after the contracts are already signed?
Maytorena has acknowledged that Coyotec was not necessarily cheaper than hiring human broadcasters. The commercial logic is replication: one character, many markets, programmable ad inventory. That is a legitimate business model, and it requires serious accountability structures, not just good intentions.
The question facing radio right now is not whether AI can hold a conversation about the Dodgers. Coyotec can, within limits, on a good recording day. The real question is whether broadcasters will build standards before a synthetic voice says something consequential and wrong about immigration, elections, or public safety. By then, no human may remain in the building to correct it.
