The research by Harker Bos Group’s Crowd React Media, conducted in May and June, tested 1,326 weekly radio listeners ages 18-45. Participants listened to station imaging and promotional spots recorded in both human and AI versions without knowing which they were hearing.
Human and AI voices scored nearly identically across key attributes, with listeners showing only slight preference for disclosure of human talent.
In the blind test, just 59% of those who heard the professional human voice (by actor Neil Wilson) correctly identified it as human. Among those who heard the AI version, only 55% correctly identified it as synthetic — meaning nearly half mistook it for a real person.
Ratings for professionalism, authenticity, credibility, energy, and likability were statistically indistinguishable between the two. Overall appeal was essentially tied at 60% for human and 61% for AI on one of the spots. The only notable difference was humor, with 33% calling the human voice funny compared to 26% for AI.
After learning which version they heard, attitudes shifted more on principle than quality:
- 47% said it would make no difference if a station used AI voices.
- 21% said they would feel more favorable.
- 33% said they would feel less favorable.
- Broader sentiment toward AI voices in media and advertising was 44% positive, 26% negative, and 30% neutral.
Crowd React founder Katie Miller said the results underscore the importance of disclosure over audio quality. After the reveal, nearly half of the human-voice group felt more favorable, while the AI group was mixed.
The study highlights a broader industry divide. iHeartMedia has banned AI on-air talent under its “Guaranteed Human” campaign, citing research showing strong consumer preference for real humans.
In contrast, Entravision launched an AI co-host named Coyotec alongside live host Geraldine “GeeGee” Guzman on Jose 97.5 in Los Angeles in 2025, reporting a 75% jump in weekly cume among Hispanic men 25-54.