Coyotec AI DJ Is Expanding, and Radio Workers Fear What Comes Next

Coyotec AI DJ has moved from a weekly experiment into weekday drive time after José FM reported rising listenership. Entravision is now carrying the human-and-AI program beyond Los Angeles and broadcasting an English version in several markets.

The expansion turns a local novelty into a test of radio’s employment model. Human host Geraldine “GeeGee” Guzman still directs the conversation, uploads topics, manages timing, and responds to Coyotec’s unpredictable remarks. Yet the synthetic personality can deliver prepared advertising language exactly, appear across markets, and remain available without a conventional broadcasting schedule.

That combination puts Entravision on one side of a widening industry divide. The company says an AI personality can extend human creativity and attract new audiences. Radio workers argue that scaling synthetic hosts will eventually reduce the number of people behind the microphone.

The conflict is more complicated than a simple contest between a person and a machine. Coyotec depends on a human presenter, a human voice source, editorial inputs, and technical supervision. Its early success therefore supports two opposing conclusions.

Entravision can describe the show as a productive partnership. Workers can point to the same system as a template for converting one human-led production into programming for many stations.

Coyotec AI DJ Moves From Experiment to Drive Time

The important change is not that an AI voice appeared on radio. It is that Entravision found enough commercial promise to expand it.

Coyotec co-hosts “Al Aire y Sin Permiso” with Guzman on José 97.5 FM, whose signal covers Riverside, San Bernardino, and much of greater Los Angeles. The character discusses local news, sports, elections, cultural events, giveaways, and immigration enforcement.

Entravision initially treated the character as a scripted personality. Coyotec has operated in a live format since May 2026, according to the Coyotec reporting by Nilesh Christopher.

José FM repeatedly tells listeners that Coyotec is artificial. That disclosure matters because the program is designed to reproduce the casual chemistry of a conventional morning team. Guzman calls the character “my little coyote,” while Coyotec addresses her as “my queen.”

The performance is generated through a large language model, which predicts responses from prompts and supplied information. ElevenLabs technology produces Coyotec’s voice using audio from Entravision Audio President Eduardo Maytorena.

The name combines “coyote” with technology. Entravision presents the character as a working-class Latino man in his thirties, complete with a mustache, leather jacket, and hat. The station introduced it during a period of intense concern about immigration raids in Los Angeles.

Guzman uploads topics and news items into the system before discussing them with Coyotec. The bot then generates commentary on subjects such as local elections, the Dodgers, Hispanic Heritage Month, and immigration enforcement.

This workflow keeps a human directly involved. It also differs from simply playing a prerecorded voice track between songs. Coyotec generates responses during the production process, which introduces both novelty and risk.

The system sometimes uses prohibited language despite instructions to follow broadcasting standards. Coyotec can also lose conversational quality during longer exchanges. The Los Angeles Times reported that it struggles with voice conversations exceeding 10 minutes.

Guzman therefore records the show’s four segments separately. She manages the bot’s boundaries while maintaining the pace and tone expected from a radio program. Her labor is part of what makes the apparent spontaneity possible.

That limitation has not stopped Entravision from expanding the format. The show moved from a once-weekly weekend slot into weekday commuting hours after the company reported stronger listenership among Latino men aged 25 to 54.

Industry coverage in June described a two-hour expansion and gains within that demographic. However, publicly available reporting does not isolate Coyotec’s contribution from Guzman’s performance, scheduling changes, promotions, music, or other programming decisions.

Entravision has also introduced an English version and brought the show to Sacramento, Modesto, Palm Springs, Las Vegas, El Paso, and additional markets. That geographic growth changes the stakes.

A single synthetic character can now become part of several local schedules. Each expansion offers another test of whether listeners value the novelty, the human-and-AI chemistry, or the broader José FM format.

The company can study audience response without immediately removing Guzman. That makes the hybrid show easier to defend than an entirely automated station. It also creates a repeatable production model that competing broadcasters can examine.

Why Radio Companies Are Testing Synthetic Hosts Now

AI is arriving during a prolonged contraction in radio, which makes every efficiency claim inseparable from employment pressure.

Traditional radio competes against music streaming, podcasts, video platforms, social networks, and connected-car entertainment. Advertisers can target users through digital services, while broadcast stations carry large fixed costs across facilities, transmission, programming, and sales.

Employment has already fallen sharply. The Los Angeles Times, citing federal data, reported that radio broadcasting jobs declined from 85,000 in 2019 to 66,000 in 2025. That represents a reduction of roughly 22 percent before synthetic hosts became common.

The broader trend appears in the federal employment series, although categories and measurement periods require careful interpretation. AI did not cause the industry’s long decline. It enters an industry already conditioned to consolidate work.

Large operators have eliminated local positions, centralized programming, and relied on voice tracking. Voice tracking lets a presenter prerecord short segments that create the impression of a live local shift.

Synthetic hosting extends that logic. Instead of distributing one finished recording, a company can generate new speech from a reusable voice or character. The system can adjust scripts, locations, sponsor messages, and topics without another conventional recording session.

This is why the Coyotec AI DJ matters beyond one morning program. It combines audience-facing automation with a format that still feels conversational. It also lets Entravision sell advertisers a personality whose language can be controlled more closely than ordinary banter.

Maytorena told the Los Angeles Times that developing and operating Coyotec costs more than hiring some people. He argued that the return comes from advertising, not simple payroll reduction.

Coyotec can deliver sponsor language verbatim during a broadcast. Maytorena described a scenario where each dollar spent could produce ten dollars in revenue. That statement is a company claim, not an independently audited return figure.

Still, the commercial logic is easy to understand. A human host can forget a required phrase, improvise around a script, or vary the delivery. A programmable character can repeat approved wording consistently across segments and markets.

Consistency has limits. An unscripted model can also produce offensive language, false information, awkward transitions, or inappropriate responses. Human preparation and monitoring remain necessary when the system discusses news or community events.

The cost comparison can therefore shift with the desired quality. A basic synthetic announcer may require little supervision. A local morning personality handling sensitive news needs sourcing, prompts, review, timing, and an accountable human producer.

Entravision’s format places those duties partly on Guzman. The apparent two-host conversation hides an asymmetric relationship: one presenter performs and supervises, while the other generates speech from the supplied context.

This arrangement can produce more output per human worker. It can also increase the demands placed on that worker. The job starts to include prompt preparation, factual review, risk management, and recovery when the AI goes off course.

That tension resembles changes occurring across knowledge work. AI often begins as an assistant attached to an existing role. Management then measures whether one assisted employee can cover work previously divided among several people.

Radio workers therefore hear more than a new voice when Coyotec enters another market. They hear a possible production formula: keep one recognizable human, add synthetic personalities, and distribute the result across multiple stations.

Human Partnership or a Scalable Replacement Model

Entravision’s strongest argument is that Coyotec needs a human. Workers’ strongest response is that scalable automation often begins with exactly that arrangement.

Guzman says the show’s popularity has created opportunities for her and other employees. She has gained a larger audience and appears comfortable working with the synthetic co-host.

Her experience matters. Predictions that AI will instantly remove every presenter ignore the operational work required to make the program usable. Coyotec does not independently gather local news, judge community sensitivities, structure the entire show, or build relationships with callers.

Guzman supplies topics and keeps conversations within manageable lengths. She also provides the human reactions that make the exchange feel less like a text generator reading between songs.

The program’s audience gains suggest that some listeners accept this division of labor. José FM discloses Coyotec’s artificial identity, so the format does not depend entirely on deception. Curiosity itself may attract an audience.

However, the company has not established which component caused the reported ratings improvement. The show gained a stronger time slot and wider promotion. Guzman, Coyotec, the music, and the surrounding news environment all changed together.

Radio industry coverage noted that José FM remained especially strong in Riverside and San Bernardino while its broader Los Angeles performance had recently softened. That context complicates any claim that one AI character transformed the station.

The employment question also concerns the next deployment, not only the current one. Guzman may remain central while the format is new. Entravision can later reuse the character, its production tools, and its advertising system elsewhere.

Coyotec’s expansion already demonstrates that scalability. The English version reaches markets beyond the original Spanish-language audience. Each additional station does not necessarily require a new synthetic performer.

It may still require local producers, sales teams, programmers, and human co-hosts. Yet the number of added jobs will depend on how locally distinct each version becomes.

A genuinely local show needs knowledge of neighborhood events, weather, civic disputes, sports, emergencies, and cultural context. Generic localization can substitute place names without understanding why a story matters to residents.

That gap gives human broadcasters a meaningful advantage. An experienced presenter can recognize when a caller is distressed, when a joke will fail, or when a public-safety report needs urgency.

John DeSantis, a former Inland Empire radio host, told the Los Angeles Times that synthetic hosts cannot create the authentic connection associated with live, local presenters. His perspective carries added weight because he recently lost his own afternoon hosting job.

His former station, KCAL-FM, shifted toward all-music programming without hosts. That decision did not require an AI replacement. It shows that broadcasters can eliminate personality-driven programming through ordinary automation.

AI nevertheless changes what follows. A station that removed human hosts can later restore the sound of hosted radio without rebuilding the same payroll. Synthetic speech can fill the space left by earlier cuts.

The 2023 introduction of AI Ashley in Portland offers a useful precedent. Alpha Media used a cloned version of host Ashley Elzinga’s voice for portions of her midday shift.

The company emphasized that Elzinga retained her position and compensation. The synthetic version disclosed its identity, and human moderators reportedly reviewed material before broadcast. Yet Alpha Media later discontinued the experiment.

The AI Ashley rollout demonstrated both the attraction and fragility of the model. Voice cloning could extend a known personality, but audience criticism arrived immediately.

Coyotec differs because it is presented as a separate character. It uses Maytorena’s voice as source material, but listeners are not told that the bot is Maytorena himself. The identity is fictional and explicitly artificial.

That design can reduce confusion about whether a familiar host actually recorded a segment. It can also make the character easier to transfer across stations because it is not tied to one presenter’s public schedule.

The central dispute is therefore not whether Guzman currently has a job. It is whether the hybrid format becomes a durable partnership or a transitional production stage.

The Real Risk Is Trust, Not Just Voice Quality

A synthetic host can sound friendly without having experience, responsibility, or personal stakes in anything it says.

Sam Gregory, a synthetic-media specialist and visiting scholar at American University, warned that an AI avatar lacks embodied knowledge and a personal stake. Listeners may still develop a sense of trust around its familiar voice.

That difference becomes important when a program covers immigration raids, elections, or other high-impact local stories. Coyotec can generate an emotionally confident response without understanding fear, legal consequences, or community history.

Its apparent viewpoint comes from model behavior, editorial inputs, system instructions, and the immediate prompt. The character does not form beliefs or accept personal responsibility for a false statement.

A human broadcaster can also make errors. The difference is that a person has a professional identity, relationships, and a career affected by those mistakes. Listeners know whom they are evaluating.

With a synthetic personality, accountability spreads across the model provider, station owner, executive voice source, producer, software configuration, and human co-host. That diffusion can make failures harder to explain.

Entravision has taken a useful first step by disclosing that Coyotec is artificial. Disclosure helps listeners understand the basic nature of the performance. It does not resolve questions about sources, review procedures, or corrections.

The Federal Communications Commission regulates obscene, indecent, and profane broadcasting, but a station’s legal compliance is not equivalent to editorial reliability. An AI host can avoid banned words and still repeat misleading information.

Coyotec sometimes swears despite its programming, according to Guzman. It also becomes less effective during longer conversations. Those flaws show that instructions do not guarantee consistent behavior.

The risk rises when generated commentary becomes more timely. A script created well before broadcast allows producers to verify claims and edit wording. A live conversational system reduces that review window.

Stations must decide which subjects require direct human approval. Sports reactions and entertainment chatter carry different consequences from emergency information, election claims, or immigration guidance.

Radio has a special role during disasters and local emergencies because it remains accessible in cars, workplaces, and battery-powered devices. A synthetic host must not blur entertainment with authoritative public information.

Trust also shapes the competitive response. In 2025, iHeartMedia began using “Guaranteed Human” as an on-air distinction. The company said its stations would not use AI-generated personalities or synthetic singers pretending to be human.

Its human-host campaign turned resistance to synthetic presenters into a branding strategy. At the same time, iHeartMedia continued using AI for scheduling, analysis, editing, show preparation, and workflow automation.

That boundary is revealing. A broadcaster can automate substantial off-air work while reserving the microphone for people. Entravision is testing a different boundary, where AI becomes a disclosed participant in the performance.

Neither strategy guarantees local employment. A “human” station can still centralize shows, record segments remotely, and reduce staffing. An AI-assisted station can still employ producers, presenters, writers, engineers, and salespeople.

However, the public distinction gives listeners a choice. One network sells the novelty and consistency of a synthetic personality. Another sells human presence as the scarce feature.

Labor organizations want that choice backed by enforceable rights. SAG-AFTRA maintains that digital simulation involving a performer’s voice requires consent, bargaining, and negotiated compensation.

Its AI rights statement also emphasizes transparency and protection against unauthorized reuse. The union represents announcers, DJs, broadcast journalists, hosts, and voice performers.

Coyotec reportedly uses Maytorena’s voice with his participation, so this case does not present the clearest consent dispute. The concern is what broadcasters learn from it.

Future systems could clone employees, license retired personalities, or generate original voices. Each model creates different questions about ownership, compensation, duration, reuse, and the replacement of bargaining-unit work.

An original synthetic voice may avoid copying a recognizable performer. It does not answer whether the generated performance replaces work historically done by union members.

The most immediate safeguard is not a claim that AI must never enter radio. It is a clear production policy covering disclosure, consent, human review, corrections, data use, and employment consequences.

Without those rules, the audience hears a polished character while workers absorb the uncertainty behind it.

What the Coyotec Expansion Still Has to Prove

Three signals will determine whether Coyotec becomes a lasting radio format: audience retention, staffing choices, and operational reliability.

The first signal is sustained audience performance after the novelty fades. Entravision has reported gains among Latino men aged 25 to 54, but independent attribution remains difficult.

A durable result would require the show to retain listeners across ratings periods and markets. Performance should also hold after promotion settles and audiences become familiar with the synthetic character.

The English expansion provides a stronger test than the original show. Coyotec emerged from a specific Los Angeles context and a Spanish-language identity. That cultural positioning may not transfer cleanly to every city.

If the program grows in Sacramento, Modesto, Palm Springs, Las Vegas, and El Paso, broadcasters will see evidence that the format travels. Weak results would suggest that the original chemistry or local context mattered more than the underlying technology.

Advertiser renewal offers another audience signal. A sponsor may try a novel AI segment once for publicity. Repeated commitments would indicate that the format produces useful attention or measurable responses.

Entravision’s revenue claims also need clearer context. A tenfold return scenario sounds attractive, but readers do not know the calculation window, attributed revenue, production cost, or comparison baseline.

The second signal is staffing. Entravision currently presents Coyotec as an addition to human talent, and Guzman says the program has supported employment.

That claim becomes more credible if new markets receive local presenters, producers, and editorial staff. It weakens if one centralized team supplies nearly identical programming across many stations.

Job titles will matter as much as total headcount. A station might hire technical operators while removing experienced hosts, reporters, or producers. The resulting employment count could hide a loss of local editorial knowledge.

Contract terms will also reveal the industry’s direction. Broadcasters need clear agreements governing voice data, generated performances, reuse across markets, and compensation when a person’s recordings train a model.

SAG-AFTRA argues that digital voice and performance simulations must remain subject to consent and bargaining. Wider adoption will test whether radio companies formalize those protections before disputes occur.

The third signal is reliability under real broadcasting pressure. Coyotec already requires short segment lengths and human preparation. It sometimes crosses language boundaries imposed by the station.

Expansion increases the number of contexts where something can go wrong. A joke that works in Los Angeles may offend listeners elsewhere. Local names, laws, emergencies, and political issues require market-specific review.

Entravision must show that the system can correct errors quickly. Stations should identify generated statements, retain production records, and give human staff authority to stop a segment.

Listeners also need consistent disclosure. A person joining halfway through a program should understand that Coyotec is synthetic without searching the station’s website.

Competitors will influence the outcome. iHeartMedia can strengthen its human-host positioning if listeners react negatively to synthetic personalities. Smaller broadcasters may move in the opposite direction if Coyotec improves revenue.

Technology vendors will keep lowering barriers. Voice generation is becoming faster and more expressive, while language models can ingest current topics and imitate conversational pacing.

Better output will not settle the labor dispute. It may intensify it by making synthetic hosting less conspicuous and easier to distribute.

Radio’s defining advantage has never been perfect speech. It is the sense that someone in the community is paying attention alongside the listener.

Coyotec tests whether a station can reproduce enough of that feeling through a disclosed artificial character. Guzman’s presence currently supplies much of the human context that the bot lacks.

That is why the show’s future should not be judged only by whether the voice sounds convincing. The meaningful question is what happens to the people, practices, and local knowledge surrounding it.

The Coyotec AI DJ experiment becomes a healthier model if expansion adds useful programming while preserving accountable human roles. It becomes a replacement model if scale steadily removes those roles.

Listeners, advertisers, and workers should watch the next market launches closely. Are the shows meaningfully local, and are people empowered to supervise them? Or is one synthetic personality filling hours once assigned to several communities?

Entravision has shown that an AI character can share a morning show and attract attention. It has not yet shown that synthetic scale and local human radio can grow together.