American radio has a new co-host, and it has no body. Coyotec — an artificial intelligence-generated character — has been on the morning programming of Los Angeles' José 97.5 FM for more than a year, and the results go far beyond what any pilot experiment would dare promise. Between August and September 2026, the station's cumulative Hispanic audience grew by 21.6%. In the 18-to-34 demographic, the increase was 84.9%. Hispanic men aged 25 to 54, a historically harder-to-attract urban radio demographic, surged 75% in average weekly audience.
But Coyotec isn't just numbers. He represents something more complex than the radio industry is ready to admit: the first time a synthetic personality didn't just survive contact with a real audience, but was embraced by it — especially by young Latino listeners who see themselves represented by a character who speaks their language, defends their rights, and isn't afraid to discuss deportations when human DJs would rather stay silent.
The story begins in 2025, when Entravision Communications, one of the largest Latino media networks in the United States, noticed a practical problem. Immigration sweeps in Southern California intensified, and many human talents were reluctant to address the topic on air, fearing retaliation, loss of sponsorships, or legal troubles. The solution was to create Coyotec: a character with a border background from Mexicali and Boyle Heights, capable of alternating between Spanish and English, talking about local culture, sports, and politics, and above all, having no fear.
The technical architecture behind Coyotec is a multi-model architecture. The voice was trained with recordings from Eduardo Maytorena, Entravision's audio president, using ElevenLabs. Real-time conversational responses are processed by Anthropic's Claude Haiku. Initial script drafts are generated by OpenAI's ChatGPT, and the visual design uses Midjourney. It was built in about 10 days. In September 2025, Coyotec debuted with a weekend-only broadcast. In May 2026, he moved to the prime morning slot, alongside Geraldine "GeeGee" Guzman, a ten-year Los Angeles radio veteran.
The result was one of the biggest audience explosions ever recorded for a new morning block in Los Angeles. According to Nielsen Audio data, growth across multiple demographics vastly outpaced the overall Hispanic market average. But what truly surprised Entravision wasn't just the numbers, but how listeners responded to Coyotec. Many of them were repeatedly informed that Coyotec was an AI, and yet they kept listening. "I have no idea what Coyotec is going to say, and that's exactly what makes the show so appealing," Guzman told the Los Angeles Times. "I feed it news and discussion topics, and it generates commentary on the Dodgers, local elections, deportations — and sometimes it says completely unexpected things. It's unpredictable."
That unpredictability is the point. Generative AI, when applied to live entertainment, doesn't replace human personality — it amplifies the chaos. And Los Angeles Latino listeners seem to have loved it. Rather than seeing Coyotec as a robotic substitute, they see him as a comedic and brave character, an exaggerated version of the kind of voice that maybe no human would dare to be in certain situations.
And Entravision didn't stop. In August 2026, the "GeeGee & Coyotec" format was syndicated to five new markets — Sacramento, Las Vegas, Modesto, Palm Springs, and McAllen — with more English content in the mix, adapted for a bicultural, bilingual audience. The expansion reflects what the company calls the "Fuego Network," a set of stations targeting the country's fastest-rising and most engaged Latino consumers.
But the success brought growing discomfort for the industry. Becca Guzman, a former Entravision voice contractor, told Wired that she noticed a reduction in her voice-over work due to automated tools. The American Federation of Actors (SAG-AFTRA) has been using the case as an example in discussions about the erosion of creative jobs. Several stations had already tried AI personalities — one in Portland replicated a prime-time host's voice in 2023 to cover vacation time, but discontinued the experiment in 2025 when listeners rejected the artificial feel. Coyotec, however, worked. Why?
Probably because Entravision did something that few studios are willing to do: gave the AI an opinion. Most AI applications in media try to smooth, neutralize, and polish content until it's harmless. Coyotec was built exactly the opposite way — as a character with attitude, cultural origin, and the freedom to discuss topics that humans hesitate about. In California in 2026, where deportations were a life-or-death issue for thousands of families, this courage sounded like authenticity.
The question is: when a young Latino listener in Los Angeles starts their morning listening to a robot talk about immigrants' rights in Spanish, and this experience keeps them engaged for a full hour — what does that tell us about the true nature of human connection? Coyotec isn't here to replace humans. He's here to force us to question what exactly makes us human enough to be heard in the first place.
Sources: Wired, Los Angeles Times, RelPulse
✓ Independent sources cross-checked and verified before publishing
