How Radio Mason is made

What's AI

• Seven of our presenters. Folkes, Lacey, Coco, Kit, Miles, Nova and Lane are AI presenters: their voices are licensed synthetic voices, and their links and banter are written by AI in each presenter's character. They are not real people, and no presenter's voice imitates any real person. Every presenter page says which one you are looking at, and so does the presenter list.
• The news bulletins. Our hourly news is assembled by AI from The Paper — our own newsroom engine that reads many independent outlets, cross-checks stories between them, and tells you how well-corroborated each story is ("confirmed by six independent newsrooms" means exactly that). Every story's sources are named and linked on our news page. Bulletins are voiced by AI and say so on air.
• Travel and weather. Read by AI from live official data (National Highways, rail feeds, weather services), written into plain speech moments before you hear it.
• Idents and station sound. AI-voiced and AI-produced.

What's human

• Four of our presenters. Ross Hoare, Wayne J C, Missy M and M-Man are real people presenting real shows, and more are joining as we bring them on. The station is deliberately a mix, and every presenter page carries a badge saying which you are listening to.
• Who covers when. Human shows are pre-recorded. If one isn't delivered in a given week, an AI presenter covers that slot — and as we grow, human presenters take over slots AI holds today. The badges and these lists are kept current, so the honest answer to "who am I listening to?" is always here rather than guessed at.
• The music. Real tracks by real artists — nothing on our playlist is AI-generated music. We're fully licensed by PRS for Music (LC-005092) and PPL (webcast licence C-002544), so the artists get paid when we play them. Backing independent artists is the whole point of this station.
• Responsibility. A human station manager (Wayne Conyers) owns the station, sets the editorial rules, chooses the music, supervises the AI systems and answers for what goes to air. When something's wrong, a person fixes it — tell us.
• The mission. Playing the artists the machine ignores, telling you how sure we are of the news, and never pretending to be something we're not. AI made the station possible; the values are human.

How the station runs itself

A lot of people ask how one small team runs a station that's on air every hour of every day. The honest answer: it's built to look after itself — and to fetch a human the moment it can't.

• The plan is the station. Every hour is assembled ahead of time — the music, the presenter links, the news, the idents — into a running order the playout simply follows. Get the plan right and the output is right, so most of our work is keeping that plan honest rather than sitting at a mixing desk.
• It watches itself, around the clock. A monitor checks the station every few minutes, day and night: is it actually on air, is the news going out on the hour, has a song come round too often, did a track cut short? The safe problems it quietly fixes itself; anything serious — dead air, a missed bulletin — it emails the station manager, even at three in the morning.
• It never goes silent. If a live show isn't delivered that week, or a system hiccups, an AI presenter covers the slot rather than leaving dead air. The music keeps playing, whatever happens.
• Your shout-outs go out fast. Send one at radiomason.com and a presenter reads it on air, usually within minutes. We check every message first — so the daft-but-lovely get through and the ones trying it on don't.
• One human answers for all of it. When the station flags something it can't fix on its own, a person picks it up. The automation does the heavy lifting; the judgement stays human.

How we mark AI content

In line with good practice (and the EU AI Act's transparency rules, which we follow voluntarily as a UK-only station): we disclose on air which of our presenters are AI, and that our bulletins are; this page is the permanent written disclosure; and our AI-generated audio carries machine-readable markers in its file and stream metadata where technically feasible. Our synthetic voices are commercially licensed, and we never clone a real person's voice without their documented consent.

Questions

Curious how any of it works, or think we've got something wrong? Get in touch — straight questions get straight answers here.