An investigation that began with a press release
Trade publication Podnews published an investigation on 10 November 2026 into Muse, the personal artificial intelligence assistant Meta launched in September. It all started with an email sent to the newsroom: a US education podcast network, Education Across the States, highlighted more than 1,300 episodes, state-by-state shows and public-records requests filed with 56 government agencies. The attached photo, supposedly showing the team in the studio, actually carried a digital watermark stating that it had been generated by Meta.
By tracing the RSS feed, Podnews found that the audio, artwork and master feed were all served from a muse.ai/podcasts address. In other words, Muse does more than help with writing: the tool can research topics, produce episodes with artificial voices and host them, even though Meta has made no announcement about this feature.
Nearly 23 hours of audio a day
The volumes are striking. The feed examined contained 1,503 episodes published between 19 October and 7 November, with a combined duration of more than 441 hours, or roughly 23 hours of content per day. Asked in writing, the named producer, who signs as Charles Mayfield, said he had no partnership with Meta and simply used Muse as a customer. He acknowledged that the platform provides almost no audience statistics and no monetisation tools: the numbers come from Apple Podcasts and Spotify.
Podnews found no online trace of this person or of the company mentioned, and the exchanges showed several signs of automated operation. Dublin City University (DCU), presented as an internship partner, said that following a review the role would no longer be offered to its students. Podnews did acknowledge that the 57-minute episode it checked was largely accurate, with only one minor error.
Much more than a single case
This network is not alone. Podnews has already counted 22 podcasts hosted on Muse, often voiced by the same synthetic pairs, such as "Alex" and "Jordan": AI briefings, personal finance, street festivals, English or Vietnamese lessons. Yet Muse is only available in some countries, and only since September. Its podcasting capabilities remain little known to the general public.
What it changes for online radio stations
For online radio and podcast publishers, this story raises three very practical issues:
- Competition by volume: a consumer tool can now generate dozens of hours of spoken programming a day at almost no cost. Directories and apps risk being flooded with generic content.
- Trust as an advantage: against these automated feeds, identifiable human voices, a real team and verifiable sources become strong selling points for listeners and advertisers alike.
- Transparency about AI: podcast directories require the use of generated voices to be disclosed. An online station that hands its news bulletins or features to AI should say so clearly on air and in its descriptions.
Artificial intelligence can help a small station fill its schedule, but the Podnews investigation is a reminder that a programme is only as good as the human checking behind it. Being open about your methods remains the best way to protect your credibility.
Source : Podnews