Claude Prompts for Podcast Planning

The short version: Claude is better than ChatGPT at long-form podcast planning because it holds more of your show's back catalogue in context without losing the thread, but it's useless if you feed it a vague prompt like "help me plan my podcast". You need structured prompts that give it your audience, your format and your constraints, then you edit hard before anything goes near a guest.

Why Claude specifically, and not another tool

Claude's long context window is the actual reason to use it for podcast work rather than ChatGPT. You can paste in five or six full transcripts, your show notes template and a guest's recent book chapter, and it will still reference details from the start of that material when you ask it something near the end. For podcast planning, where you're juggling a season arc, a guest's back catalogue of interviews, and your own previous episodes so you don't repeat questions, that matters more than cleverness.

It's not better at being witty. It's not better at jokes or hooks. It's better at not forgetting what you told it twelve paragraphs ago. That's the whole case.

The uncomfortable bit nobody says out loud

Most "AI podcast prompt" posts imply the AI will make your planning faster across the board. It won't, not evenly. Claude is excellent at research synthesis and outline structure. It is mediocre at generating good interview questions, because good questions come from genuine curiosity about a specific person, not from a pattern Claude has seen in a thousand podcast transcripts. If you ask Claude for "10 great interview questions for a marketing founder," you'll get ten questions that could apply to almost any marketing founder. That's the tell. Use Claude to do the heavy lifting on structure and research, and spend your own saved time on the three or four questions that only make sense for this specific guest. That's where the actual listener value lives.

A worked example: planning one episode from scratch

Say you run a weekly B2B podcast on AI adoption for small business owners, 30 minutes per episode, and your next guest is the founder of a logistics software company who's been on four other podcasts this year talking about the same two case studies. Here's a realistic prompt sequence.

Step 1: Guest research prompt

"I'm interviewing [name], founder of [company], on [date] for a 30-minute podcast about AI adoption in small business. Here are links to three previous interviews they've done [paste URLs or transcripts]. Summarise the stories and points they've already covered publicly, so I avoid repeating them. Flag anything they said that sounded rehearsed versus anything that sounded like a genuine aside worth following up on."

This is the prompt almost nobody uses, and it's the one that stops you recording an episode that's 80% identical to three others already online.

Step 2: Angle prompt

"Given that summary, suggest three angles for this interview that haven't been covered in those three transcripts. My audience is small business owners with 5 to 50 staff who are AI-curious but not technical. I want at least one angle that's slightly uncomfortable or contrarian for a logistics software founder to answer."

Step 3: Outline prompt

"Build a 30-minute episode outline using angle 2. Structure: 3-minute intro and guest context, 20 minutes on the core topic broken into 3 segments, 5 minutes on practical takeaways for listeners, 2-minute close with a call to action. For each segment give me the question I'd ask and one likely follow-up based on what they might say."

That's a specific, usable document in under ten minutes of your time, not a generic list of "great podcast questions" that could sit on any show.

Step 4: Pre-record gut check

"Here's my draft outline. Read it as if you're a sceptical listener who's heard three podcasts with this guest already this year. Where would you switch off?"

This step is the one people skip and it's the most useful. Claude will quite happily tell you your segment three is thin if you ask it to criticise rather than help.

Prompts for the planning stages beyond one episode

Season or series mapping

"I run a podcast on [topic] for [audience]. Here are the titles and one-line summaries of my last 12 episodes [paste list]. Identify gaps in topics I haven't covered, overlap I should avoid, and suggest 8 episode topics for the next quarter that build a narrative arc rather than feeling random."

Give Claude the actual list, not a description of your show. Vague inputs get vague outputs, every time, no exceptions.

Guest sourcing and vetting

"I need a guest who can speak credibly on [specific niche topic] to an audience of [description]. Based on what you know, name the type of person I should be looking for (role, company size, what makes them credible) rather than guessing specific names, since you may not have current information on who's active right now."

This is the honest way to use Claude for guest sourcing. It cannot reliably tell you who's currently active, available, or good on mic. It can tell you what credibility profile to search for on LinkedIn or in trade press, which saves you time narrowing down.

Show notes and titles

"Here's the full transcript of this episode [paste or summarise key sections]. Write 3 title options under 60 characters, a 100-word episode description for podcast apps, and 5 pull quotes under 20 words each suitable for social captions."

Ask for a number every time: 3 titles, 5 quotes, 100 words. Open-ended prompts get rambling, unusable output.

Where this feeds into your wider content system

None of this matters if the episode then sits as a single audio file nobody finds. Once you've got the transcript, the same Claude conversation can draft the repurposing brief, and that's worth pairing with a real process for how to repurpose a podcast episode into a week of content rather than one tired LinkedIn post. If you're already running an AI content calendar, slot the episode topics in at the planning stage, not after recording, so your social team and your podcast aren't working from two different plans.

A note on prompt quality generally

If you're new to writing prompts that get specific output rather than generic listicle answers, it's worth spending an hour on the fundamentals, Claude and ChatGPT respond to the same structural principles even though the interfaces differ. There's a wider rundown of this in the ChatGPT prompts guide, and most of it transfers directly: give context, give constraints, give a number, ask it to critique its own output before you accept it.

What to not bother with

Don't ask Claude to write your intro script word for word unless your show has a very fixed, formal format. Scripted intros read as scripted, and podcast listeners can hear the difference between a human talking and a human reading an AI paragraph. Use Claude for the bones (outline, questions, structure, summaries) and keep your own voice for anything that gets spoken aloud. The same logic applies if you're turning clips into other formats, whether that's audiograms or something more visual like using AI to turn a still image into video for a quote card.

If this is part of a bigger push to get a year's worth of content out of a handful of podcast episodes, it's worth reading how to plan that rather than doing it episode by episode in a panic each week, which is covered in the piece on writing a year of social content in one afternoon.

Frequently asked questions

Is Claude better than ChatGPT for podcast planning?

For long documents like transcripts, back catalogues and research notes, yes, because Claude's context window holds more detail without losing track of earlier points. For snappy hooks and titles the two tools are roughly comparable, so test both if you have access.

Can Claude write my actual interview questions?

It can draft a solid starting list, but generic prompts produce generic questions that could suit any guest. The best use is to have Claude cover the structural questions and spend your own time on the two or three that only make sense for this specific person.

Can Claude find me podcast guests?

Not reliably by name, since it may not have current information on who's active or available. Use it to describe the credibility profile you need, then do the actual searching on LinkedIn or in trade publications.

How much time does this save?

A full research, angle and outline prompt sequence like the one above typically takes under 15 minutes to run, against what can be an hour or more doing the same research by hand, particularly the step of checking what a guest has already said on previous podcasts.