AI Prompts for Podcast Script Writing and Episode Outline Generation in 2026
I produce three podcasts and have been using AI for episode prep since early 2025. The workflow has cut prep time from 4-6 hours per episode to about 90 minutes without reducing output quality. The key is using AI for the parts of podcast production that are time-consuming and template-able — research, question development, show notes — while keeping the creative and conversation elements human. These are the specific prompts that work in production.
Guest Research and Interview Preparation Prompts
Before any interview, I run a two-part research prompt. Part 1: 'Research [guest name] as if you were a journalist preparing for a 45-minute interview. Based on publicly available information (books published, podcast appearances, articles written, LinkedIn background), compile: (1) their primary expertise and the specific topic they're most associated with, (2) the 3 most interesting or counterintuitive things they've said publicly — include the source, (3) any major position changes or evolutions in their thinking over time, (4) what are they known for NOT saying or actively disagreeing with in their field — what is their controversy or counter-position, (5) topics they've been asked about many times and would likely give a rehearsed answer to (avoid these or approach from a different angle).' Part 2 of the research: 'Given this background, what questions would most likely produce an answer this guest has NOT given in a previous interview? List 5 such questions, each targeting a gap in their public commentary.' The 'previous interview' gap question is the single most valuable research prompt for podcast prep. It systematically surfaces angles that aren't in the guest's standard talking points.
Combine AI guest research with a manual step: listen to the guest's 2 most recent podcast appearances for 20 minutes each. Note the questions that got the most interesting responses and the questions that produced rehearsed nonanswers. Feed these observations back into the AI research prompt for sharpened question design.
Part 1: expertise, counterintuitive positions, evolution, controversies, rehearsed topics
Part 2: 'what would they not have answered in previous interviews?'
Combine AI research with manual review of 2 recent appearances
Target questions that produce unrehearsed answers, not performance of known positions
For technical experts: 'what is the most common misconception about [their field] they haven't addressed publicly?'
Controversy framing: 'what do you disagree with that most people in your field believe?'
Episode Outline and Run-of-Show Prompts for Structured Conversations
Unstructured conversations wander. A run-of-show (a timed sequence of topics and transitions) keeps episodes focused without making them feel scripted. My outline prompt: 'Create a run-of-show for a 45-minute podcast episode with guest [name] on the topic [topic]. Structure: (1) cold open — one provocative question or statement to start before any introduction (30 seconds), (2) guest intro — 2 minutes, keep it tight, what does the audience need to know right now to understand the rest of the conversation, (3) core topic block 1 — [primary angle, specific guest expertise] — 10-12 minutes with 3 key questions and one expected tangent to follow, (4) core topic block 2 — [secondary angle, different facet] — 10-12 minutes, (5) controversy or challenge moment — one question designed to create productive disagreement or tension, 5 minutes, (6) listener-applied segment — 'what should listeners actually do with this information?' — 5 minutes, (7) close — one last insight or prediction, 2 minutes. For each block include: opening question, 2 follow-up questions if the opener lands, and 1 pivot question if it doesn't.' The 'pivot question if it doesn't land' instruction is something most podcasters skip and regret — every experienced host has been stranded by an opener that got a one-sentence answer.
The controversy block (point 5) is the highest-yield section for listener engagement. Podcast listeners report that moments of respectful disagreement are the most memorable and shareable. The question should challenge something the guest has said, not attack them personally — 'I've heard the counter-argument that X — how do you respond to that?' is a healthy frame.
Cold open: one question before intro — creates immediate energy without setup
Guest intro: 2 minutes max — what they need to know for NOW, not complete bio
Always include pivot questions for openers that might not land
Controversy block: respectful disagreement, not attack — 'counter-argument that X' framing
Listener-applied segment: 'what should someone actually do with this?' — critically missing from most shows
End with prediction or forward-looking question: invites return listeners to check back
Show Notes and Transcript Summary Prompts for SEO and Audience Retention
Show notes turned me from a skeptic to a believer when I started ranking episode titles on Google. A well-written show note is a 300-600 word blog post that ranks for the topics discussed in the episode. My show note prompt: 'Write show notes for a podcast episode where [host] interviewed [guest] about [topic]. Based on this transcript/outline [paste], write: (1) a 100-word summary optimized for search around the keyword [target keyword] — use the keyword in the first sentence, (2) 5 key takeaways from the episode as bullet points — each should work as a standalone insight, not just a topic label, (3) time-stamped highlights for 5 key moments — include the approximate time, the topic label (H3), and 2 sentences describing the specific insight at that moment, (4) 3-5 resources mentioned in the episode with brief context, (5) a 1-paragraph bio of the guest optimized for their professional keywords.' The timestamped highlights are the most clicked element of show notes in my analytics — listeners use them to navigate or rescan episodes. They're also naturally SEO-optimized because they describe specific content at depth.
For YouTube podcast uploads, the show notes format plus a transcript embed creates a text-heavy page that ranks well for long-tail queries. Google indexes podcast transcripts, and the combination of transcript + structured show notes regularly puts episode pages on page 1 for moderately competitive niche topics.
Show notes keyword target: match episode topic to a searchable phrase, not just a subject label
Key takeaways as standalone insights: not 'we discussed X' but 'X is better than Y because Z'
Timestamps are the most clicked show notes element — always include 5+ key moments
Guest bio in show notes: optimize for their professional keywords (helps them find the episode too)
Upload transcript to YouTube description: Google indexes it and ranks for long-tail queries
Repurpose 3 takeaways as social posts the same day the episode drops
Solo Episode and Monologue Script Prompts for Confident Delivery
Solo episodes require different scripting than interviews — without a guest to prompt the next thought, pacing and structure carry more weight. The solo episode prompt: 'I want to record a 20-minute solo episode on [topic]. My audience is [description]. The episode's core argument is [one sentence]. Help me structure this as a spoken monologue — not a written essay. Requirements: (1) open with a story or specific scenario that illustrates the central point, not a statement of the central point, (2) use 3-4 main points with clear verbal signposting ('the first thing I want you to understand is...', 'and this connects to something most people get wrong...'), (3) include at least one moment of vulnerability or self-challenge — something I got wrong or had to change my mind about, (4) use short sentences for emphasis — break up complex ideas into 2-3 word punchlines after longer explanations, (5) write the closing as a 'what to do next' instruction, not a summary. Word count: ~2,400 words for 20 minutes at conversational speaking pace.' The distinction between written essay structure and spoken monologue structure is the most common failure in solo podcast scripts. Written language frontloads complexity; spoken language builds to it. AI defaults to essay structure unless you explicitly specify oral delivery conventions.
For regular solo episodes: build a running bank of 'story starters' — specific scenarios from your own experience that illustrate common points in your niche. Prompt AI monthly: 'Given my last 10 episode topics [list], what are 5 new angles I haven't explored? For each angle, suggest a personal story framing I could use to open the episode.'
Open with a specific story/scenario, not a statement of the main point
Use verbal signposting throughout — spoken structure is more explicit than written
Include a genuine self-challenge moment — audiences reward intellectual honesty
Short sentences after complex ideas: punchlines that land orally
2,400 words = ~20 minutes at conversational speaking pace
Word count calibration: record yourself speaking 500 words, time it to calibrate per-minute word rate