At the end of 2025 I started experimenting using AI to help shape and build my shows. Whether we like it or not, AI is growing and is here to stay – and knowing how to use it effectively is becoming more and more important. I think of AI as being like an “artificial intern”. AI is a supporting tool in the process of putting together my show – but it is not the whole process. As presenter/producer, I retain overall control over what goes in, and what doesn’t.
To some extent I’m still figuring out where it adds value. Some AI functions I used early on, were okay for putting together my show – but I realised quite quickly that it didn’t quite work for my particular type of show. So I scaled back my use of those functions.
There are three main features of AI which I now routinely use for my show. Refining content ideas. Producing jingles. Supporting programme material – promotional graphics & visuals for social media/website.
Jingles
This is, by far, the most interesting part of my experiments in to AI. Last year, I found a song on Spotify which I really liked – but when I tried to look up the artist I couldn’t find anything about them. I later learnt the song was made entirely with AI.
While I didn’t want to be playing AI songs on my radio show, I did wonder whether AI could be used to make jingles. I love radio imaging, jingles, and all the production which goes in to it – and since the jingles in my show at that point had been on air since 2021, I thought it time for a refresh.
Initially I found an AI Jingle Maker website, where for a small amount of money, someone (also named Frederick!) would produce the jingle for me according to the brief I provided. The results were surprisingly good, and I wondered whether it would be possible to cut out the middle man and make the jingles myself.
Bit of research, trying different sites, and I settled on using a website called Suno to generate the jingles. But I was unsure exactly how to prompt (brief) the site in order to get the best results. A bad prompt will result in it just making a 3 or 4 minute song – not good for when I want a 20 second radio ident. So, I used ChatGPT to help me – asking ChatGPT to suggest a prompt for the style of music, the genre, the era I wanted to create.
The result is that 3 months down the line, I have well over 100 jingles which have made it in to my own jingle package, on air regularly across my shows. They add quite a lot of texture and flavour to the show, as I quite like mixing and matching the jingles to the songs I play.
The jingles are not perfect, and the normal AI audio defects can be heard in some of them – but to me they’re just a bit of fun. It also means that rather than spending upwards of Β£10000 on a professionally produced package, I spend just Β£20.
So pleased have I been with the response and feedback from my loyal listeners (friends, family, and fellow Cirencester Radio volunteers), that I have put all my jingles on my radio website, so you can click around, read the story of how they were produced, and listen to each individual jingle.
Control vs Automation
Whether it is building jingles, or choosing music for the show, I am clear with myself that while I use AI to assist, I still make the final decision. If I don’t like something AI has generated, I won’t use it. I retain control.
When I started my Saturday morning show in November, I tried using AI to build the entire running order – sometimes giving it a list of songs, sometimes asking it to select songs. While this worked to a degree, it wasn’t the best way of shaping my show.
My music choices on the show are very eclectic – that is one of the main selling points of my show. This means I might have songs by Vera Lynn, Frank Sinatra and Bobby Darin – in the same show as the latest chart releases from Niall Horan, a Country release from Lainey Wilson, and 70s Rock – The Eagles or Elton John.
The issue with AI is that, it doesn’t have the producer’s ear. It might tend to group all the songs from the same decade / genre together, rather than make it a musical journey flowing freely between the genres and decades.
To be able to do this properly, you need to listen to the songs, and decide what works and what doesn’t. AI might explicitly stop me from playing Luke Combs followed by Vera Lynn… but pick the right Combs song, one which is slower and more of a ballad, than a Country rock song – and it absolutely can work. I did so on Monday!
So while I might use AI to ask it to come up with a list of songs about a particular topic / by a certain artist / produced by a certain studio – my expertise is required to curate the songs in to a usable playlist. A show built entirely with AI would not be as eclectic as my shows are – and when shaping my shows I try to take in to account the tone and flow I’m aiming for, and what the audience might expect. AI does not understand this.
I have also previously used AI to collect interesting information about the music, to read out on my show. This also didn’t work that well – because what AI finds interesting, to me isn’t something particularly useful. It often defaulted to a piece of information more akin to an academic commentary on the song, rather than something a music radio presenter might say on air.
This also didn’t fit the style of my show. My show isn’t meant to be a “music expert”. I’m not pitching myself as a muso or anything like that. I’m simply someone who enjoys listening to music, and wants to share that love of music with the listener. The tone of my show is: “I enjoyed this, so I’m sure you will too”, as opposed to knowing everything there is to know about a particular singer, or knowing which recording studio it was made at.
How AI Improves the Show
Where AI really does come in to its own is with all the supporting tasks I ask it to do. As a solo presenter/producer, on a voluntary community radio station, there’s not a marketing budget to produce great big promotional posters. So how do I promote my show? Primarily through social media (Facebook & Instagram) as well as on this website.
But there’s something missing. A photo of me sitting in my bedroom next to the computer isn’t necessarily that interesting every week. So I used AI to produce some promotional graphics for my show. This adds another layer. I often combine my AI graphic with a photo of me either out and about or in the studio. I also try and prompt ChatGPT to make the graphics fit the tone of the show I’m aiming for that week.
This has had a noticeable impact on how people engage with my show. A more interesting graphic makes the online social media presence more noticeable. People tune in to the live show, and they message me when listening – not just hearing the recording on ‘Listen Again’. For me: That’s the difference between a radio station and a podcast website. With live radio, you get that shared experience of tuning in and listening at the same time. I am experiencing Monday evening at the same time as my listeners.
The results are immediately visible when I look at the statistics for my stream. When I broadcast my show, as well as sending the audio ‘down the line’ to Cirencester Radio, the audio is also streamed on my own Fred On the Radio website.
In Q1 2026 – from 1st January to 27th March – I’ve had 53 listeners – 87 listener hours.
During the same period in 2025, I had 5 listeners – 10 listener hours.
(Admittedly I wasn’t on air in January 2025, as I took a 6 month break from broadcasting).
That is a significant growth.
Closing Thoughts
I am still experimenting with AI, and how I use it within my show. I can’t say it is perfect. I can’t say I’ll always get it right either. But that’s the joy of doing community radio. I have that freedom to experiment and do what I want (within reason).
As time goes on I keep what works – and quietly drop what doesn’t. There was a fun music quiz question feature AI suggested for my Saturday morning show – but the trouble is, for it to work properly there needed to be a bit of listener interaction. It didn’t suit the style of my show. So I quietly dropped it.
I continue to use AI, where it helps to improve the show – I’m not chasing technology for the sake of playing with new technology.
Thanks for reading!
FH.
