A few weeks ago I told you I'd be writing about using AI to come up with podcast episode ideas.
Then my diary went absolutely bonkers.
Rather than rush it and send you a half-arsed newsletter, I parked it.
It’s a bit hard to ignore AI nowadays isn’t it? It’s either going to take over the world and nick your job or it’s going to solve every problem you’ve ever had.
None of those are really helpful are they?
So instead of talking about AI...
I thought I'd show you Reader.
First Things First... Don't Panic, It's Just a Fancy Acronym
I can hear you now, “Woah Vic, what you on about now, I thought you were talking about AI?”
Well LLMs are just one type of AI, and it’s the one we’re using today.
Don't let the name put you off.
LLM stands for Large Language Model.
It's the technology powering tools like ChatGPT, Claude and Google's Gemini. Think of it as a very well-read assistant. These tools can do lots of other clever things too, but today we're just using them to help us think up podcast episode ideas.
What they can't do…
...is read your mind.
And that's exactly what today's experiment is all about.
I wanted to answer a question I hear all the time:
"Can AI actually come up with good podcast episode ideas?"
So I decided to put it to the test.
I'm going to ask Claude (you could do exactly the same thing with ChatGPT or Gemini) to come up with podcast episode ideas for me.
But I'm going to ask the same question three different ways.
Each time, I'll give it a little more information.
We'll see what changes. Will the ideas get any better?
Or is AI just full of hot air?
Let's find out…
Experiment #1: I Told AI Almost Nothing
Like loads of people, I started with a very simple prompt.
Give me 10 podcast episode ideas for a social impact podcast.
Claude came back with ten ideas.
And do you know what? They're actually… pretty good.
Things like Failure Files, The Ripple Effect and Small Wins are dead interesting podcast concepts.
But there’s a problem, they're ideas for a social impact podcast.
Not my podcast.
But Claude didn't fail, it did exactly what I asked!
It's me, hi. I'm the problem, it's me.
Experiment #2: Right...Let's Actually Help It
This time I treated Claude more like a new producer joining the team.
I told it:
- what Grab The MIC Live is about
- who my audience are
- my values
- how I like to communicate
- what I wanted it to create
Then the ideas became much more relevant.
Instead of generic social impact ideas, I got things like:
- Should You Even Start a Podcast?
- Finding the Confidence to Hit Record
- How to Get Your First 100 Listeners
Much better.
But...
they still weren't quite right.
Claude assumed Grab The MIC Live had already launched and every episode would be an interview.
Again...that wasn't Claude getting it wrong.
That was me not giving it the full picture.
Now, some people might give up at this point, but the key is to keep giving it context, keep feeding the little AI monster information.
Experiment #3: Turns Out AI Likes Context
AI isn't a vending machine, you don't press a button and get the perfect answer first time, you have a conversation.
A weird, bizarre conversation with a robot, but hey.
So instead of starting again, I simply said, “Great start, but we need a few tweaks…”
I explained that:
- the podcast hasn't launched yet
- I wanted a mixture of solo and interview episodes
- the audience includes people who haven't started podcasting yet
- I'd like the ideas to follow a logical learning journey
The results got even stronger and that's because Claude wasn't guessing anymore.
It was building on everything we'd already discussed.
I didn't stop there, by the way. I kept pushing Claude to critique my ideas, challenge my assumptions and even act like an executive producer... but that's a story for another newsletter.
So... is AI friend or full of rubbish?
Neither.
It's a tool, a very clever one at that.
But like any tool, the results depend on how you use it.
Here's what I'd love you to take away from this little experiment.
🚫 AI doesn't replace your creativity.
Your ideas, experience and judgement are still the most valuable part.
🚫 AI isn't magic.
It only knows what you tell it.
The better the brief, the better the outcome.
✅ Treat it like a collaborator.
Not a search engine or a writer.
Brief and challenge it - and push back!
Ask it what's missing, that's where the really interesting stuff starts happening.
One Last Thing Before You Go…
If you find yourself explaining your podcast idea over and over again, there's an easier way.
ChatGPT, Claude and Gemini let you create your own little AI helper.
ChatGPT calls them GPTs.
Gemini calls them Gems.
Claude uses Projects where you can save instructions and reference files so it understands your podcast every time you come back
Think of them as giving AI a permanent briefing.
Tell it about your podcast idea, who it's for, your tone of voice and what you're trying to achieve.
You can even go a step further and give it things like your podcast description, your show proposition, previous episode descriptions or even a transcript from one of your episodes. The more it understands your podcast, the more useful and relevant its ideas become.
You tell it all of that once...
...and every future conversation starts from there.
It's like working with a producer who's already listened to your show, understands your audience and knows what you're trying to achieve.
Today's experiment wasn't really about AI. It was about briefing. Whether you're working with a producer, a designer, a freelancer or an AI assistant, the same rule applies - the better the brief, the better the outcome.
Podcasting is built on conversations. It turns out AI works best that way too!
Cool Things for You
The Audio Production Awards are now open for entries!
If you're making brilliant audio, this is your little nudge to put yourself forward.
The Audio Production Awards (the APAs), run by AudioUK, celebrate the very best in podcasts, radio and audiobooks across the UK. There are 30 categories this year, including several new ones, recognising everyone from presenters and producers to sound designers and production companies.
Now, I know what some of you are thinking...
"My little podcast isn't award-worthy."
Please don't decide that for the judges.
If you've created something you're proud of over the last year, why not throw your hat in the ring?
There are early bird entry fees until 30th August, with the final deadline on 16th September. And, if cost is a barrier, the Pay What You Can scheme is back too.
Even if you're not entering this year, I'd still recommend having a look through the categories and last year's winners. It's a brilliant reminder of the incredible variety of audio being made and might just spark an idea or two for your own podcast.
Good luck if you decide to enter - I'll be keeping everything crossed for you! 🤞
Find out more here
That's it for this week's newsletter.
I hope this little experiment has shown you that AI isn't about finding the perfect prompt.
It's about having a better conversation.
Next week, we're moving away from AI and onto something that could make more people press play on your podcast... your episode titles. Because even the best episode in the world won't get listened to if nobody clicks on it in the first place.
Oh, and one last thing...
Thank you so much to everyone who's already replied with their favourite podcast for our Grab The MIC Summer Listening List. There have been some absolute belters so far!
If you missed last week's newsletter, it's not too late to join in. Just hit reply and tell me:
- The name of the podcast.
- Why it's the one for you.
- And, if you've got one, the episode you'd tell everyone to start with.
I'm going to pull all your recommendations together into one big summer listening guide for the Grab The MIC community, and I genuinely can't wait to discover some new favourites.
I'll see you next week.
Vx
