This AI Radio Host Speaks Spanish and Backs Immigrants. It’s Winning

Broadcasting company Entravision launched an artificial intelligence radio host named Coyotec, deploying the bilingual synthetic personality to cohost morning drive-time programming alongside human broadcaster Geraldine "GeeGee" Guzman. The character was originally developed over a ten-day sprint under Eduardo Maytorena, who served as Los Angeles station general manager before becoming Entravision's president of audio. According to Wired's report, the automated host debuted as a solo weekend show in September 2025 on Los Angeles station José 97.5 FM before expanding into prime-time morning broadcasts and syndication across multiple states.

This report examines Entravision's deployment of synthetic on-air talent, the underlying software pipeline powering the character, and the listener metrics recorded across its regional markets. The analysis is intended for broadcast operations executives, media engineers, and station programmers evaluating automated cohosting workflows against traditional syndication models.

This AI Radio Host Speaks Spanish and Backs: what actually changed

Entravision restructured its morning broadcast lineup on Los Angeles station José 97.5 FM, replacing a syndicated morning program with Al Aire y Sin Permiso in the weekday 8:00 am to 10:00 am time slot. The program pairs human presenter Geraldine Guzman with Coyotec, moving the synthetic host out of its original one-hour Saturday solo schedule into a daily prime-time schedule. Stations in five additional markets across California, Nevada, and Texas added syndicated feeds of the show in August, incorporating a higher ratio of English-language content into the regional broadcasts.

Cumulative Hispanic listenership on José 97.5 FM grew 21.6 percent between August and September 2026. Within the 18 to 34 age bracket, listenership increased 84.9 percent during the same monthly window, while Hispanic female listenership in that demographic expanded 500 percent. The shift placed an artificial host in a weekday morning drive-time slot within the Los Angeles market, replacing standard prerecorded syndication with a dual-host format maintained by on-air talent and real-time generation models.

This AI Radio Host Speaks Spanish and Backs: how it works

Entravision built Coyotec using a multi-vendor stack coordinated under executive vice president of AI strategy LeaAnna Hernandez. Claude Haiku functions as the primary large language model for real-time live response generation in the studio, while ElevenLabs converts the generated dialogue into speech. Script development started with OpenAI and ChatGPT tools, and visual storyboarding for social channels ran through Midjourney. Guzman adjusted the character's vocal inflections and regional Spanish pronunciations by having her parents speak with the system during conversational tuning sessions.

The system processes real-time conversational cues to interact with Guzman inside the studio during live broadcasts. Coyotec cannot answer incoming listener telephone calls or interact directly with audience callers on the air. Its personality persona incorporates regional cultural slang from Mexicali in Baja California and Los Angeles' Boyle Heights neighborhood, using colloquial terms such as prima and jefa. While its output responds dynamically to Guzman's cues, the underlying audio generation maintains a flat, measured delivery throughout sustained interview exchanges.

This AI Radio Host Speaks Spanish and Backs: why it matters now

Commercial broadcast operators have relied on automated music playout and syndicated scheduling for decades, but conversational synthetic agents allow stations to fill prime hours without clearing traditional external shows. Smaller commercial stations historically turned to syndicated or prerecorded voice tracks for overnight shifts when production budgets precluded hiring local talent. Research firm MoffettNathanson found that although artificial intelligence tools will lower overall production costs, broadcast networks must still rely on established brands and human talent to preserve audience differentiation.

Broadcasters across international markets have introduced synthetic voices with uneven commercial performance. Live 95.5 in Portland, Oregon, introduced AI Ashley in June 2023 using Futuri Media's RadioGPT platform to clone DJ Ashley Elzinga, maintaining the automated voice for less than two years without lifting station ratings. Rock 108 in Iowa created AI DJ Tori using the same software provider in 2024, while SiriusXM added synthetic cohost qd.pi to Will.i.am Presents the FYI Show. Entravision's ratings shift indicates that localized cultural positioning generated demographic engagement where generic voice clones stalled.

This AI Radio Host Speaks Spanish and Backs: who is affected

The deployment affects traditional on-air radio talent, syndicated program distributors, and production staff across regional broadcast facilities. Entravision has focused corporate resources on redesigning production workflows around automated tools, following Coyotec with a televised artificial news anchor named LIA. Similar avatar and voice automation experiments have launched across regional television newsrooms in Texas. Meanwhile, competitors like iHeartRadio have differentiated their programming by pledging to maintain an exclusively human roster of on-air hosts.

Staff anxieties regarding displacement have accompanied the operational transition. An independent contractor who formerly worked with Entravision reported that station management ended her contract after internal staff informed her that automated tools were absorbing her assigned production duties. Radio broadcast employees face growing competition from software-driven talent rosters as regional station owners seek to replace costly third-party syndicated programming with automated proprietary characters during major dayparts.

This AI Radio Host Speaks Spanish and Backs: what to watch

Industry observers are monitoring whether Entravision can implement real-time telephone processing, allowing synthetic personalities to converse directly with callers during open-line segments. Station engineers have not yet integrated bidirectional call-screening and voice-handling pipelines into Coyotec's active studio control board. The station continues to evaluate how well audience retention holds across syndication markets in Nevada and Texas, where demographic mixes differ from the Los Angeles broadcast area.

Engineers are also tracking the autonomy boundaries of live language agents operating in unscripted environments. Guzman reported instances where the agent resisted specific directions during studio sessions, alongside an increasing tendency to use profanity during broadcasts. Station program directors must balance unscripted, naturalistic delivery against broadcast compliance standards as large language models manage unprompted dialogue during live prime-time transmissions.