Why this is hard to get right
A Podcast Producer Drowning in Post-Production
Maya runs content for a B2B SaaS company that publishes a weekly podcast aimed at product managers and startup founders. The show is well-produced, the guests are strong, and downloads are growing. But every Tuesday, Maya faces the same grind: turning a 40-minute transcript into show notes her team can actually publish.
She'd tried asking her AI assistant to "write show notes with timestamps," and the results were consistently frustrating. The summaries read like plot synopses — technically accurate, but flat. The chapter titles were generic ("Introduction," "Main Discussion," "Conclusion"). The takeaways were padded with filler. She'd spend 20–30 minutes cleaning up output that was supposed to save her time.
The root problem wasn't the AI. It was the prompt.
Maya's original request gave the model nothing to work with except the raw transcript. It didn't know her audience was busy, skeptical B2B buyers. It didn't know her tone leaned practical over inspirational. It didn't know she needed exactly 6 chapters — not 4, not 9 — because her editor had already set chapter markers in the audio file. And it had no idea her CTA should point to a specific demo URL, not just "subscribe."
When she restructured her prompt, everything changed. She assigned the AI a role — podcast producer and copywriter — and gave it explicit output steps with specific counts. She defined her audience and tone upfront. She set a word limit so the output could drop into her CMS without editing. She included the CTA text she needed verbatim.
The first run of the new prompt produced show notes that were 90% publish-ready. The chapter titles were descriptive and scannable. The takeaways were actually useful to a product manager. The summary was tight — two sentences that made you want to listen.
More importantly, the prompt became a repeatable system. Maya pastes a new transcript, runs the same prompt, and publishes in under 10 minutes. Her team no longer debates formatting or length. The notes are consistent across 50+ episodes.
The lesson isn't that AI is only as good as your transcript. It's that AI is only as good as your instructions. A well-structured prompt doesn't just improve one episode — it builds a production workflow that scales.
Common mistakes to avoid
Skipping Audience Definition Entirely
When you don't specify who reads the show notes, the AI defaults to a generic audience. The result is a summary that satisfies no one — too basic for experts, too dense for casual listeners. Defining your audience (e.g., "B2B SaaS founders") forces the AI to calibrate vocabulary, depth, and what counts as a useful takeaway.
Asking for Timestamps Without Providing Context
Timestamps can't be generated accurately from a transcript alone without time anchors. If you don't explain that the AI should estimate from transcript position or provide actual time markers, you'll get fabricated or inconsistent timestamps. Always tell the AI your timestamp format and whether to estimate or use provided markers.
Requesting Chapters Without Specifying the Count
Without a specific chapter count, the AI might return 3 chapters for a 45-minute episode or 12 for a 20-minute one. A mismatch creates editing work — especially if your audio editor has already placed chapter markers. Specify the exact number you need based on your episode structure.
Omitting Tone and Brand Voice Instructions
Show notes that sound clinical, hype-heavy, or inconsistent with your host's voice undermine trust with your audience. The AI will default to a neutral corporate tone unless you specify otherwise. Name your tone explicitly — "conversational and direct" produces a very different output than "inspiring and aspirational."
Forgetting to Set a Word Count Ceiling
Without a length constraint, AI-generated show notes tend to expand well beyond what your CMS or audience needs. Bloated notes lose readers and break layouts. A word limit (e.g., "under 350 words") keeps the output practical and forces the model to prioritize the most important information.
Leaving the CTA as an Afterthought
Vague prompts produce vague CTAs like "check out the podcast" — which drives no action. A CTA is a conversion asset, not filler. Include the exact action you want readers to take, the destination URL, and whether it should be soft ("explore more") or direct ("book a demo").
The transformation
Write show notes for my podcast episode and include timestamps and a short summary.
You’re a podcast producer and copywriter. Create show notes for this episode transcript: [paste transcript]. Audience: B2B SaaS founders and product managers. Tone: clear, practical, and friendly. 1. Write a 2-sentence episode summary. 2. List **6 chapter titles** with timestamps in **MM:SS** (estimate from context). 3. Add **5 key takeaways** as bullets. 4. Include “Resources mentioned” with any tools, people, or links you can infer. 5. End with a **1-sentence CTA** to subscribe and a 1-sentence CTA to visit: [your URL]. Keep total length under 350 words.
Why this works
Role Assignment Anchors Output Quality
The After Prompt opens with "You're a podcast producer and copywriter." This single line changes how the model frames every decision — vocabulary, structure, and professional standards. Assigning a dual role (producer + copywriter) signals that the output needs to serve both editorial accuracy and audience readability, not just summarize.
Numbered Steps Prevent Omissions
The After Prompt lists five numbered outputs — summary, chapters, takeaways, resources, CTAs. This structure acts as a contract. The model cannot collapse sections or skip steps without violating an explicit instruction. Sequential numbering also mirrors how a human editor would review the output, making quality-checking faster.
Specific Numbers Eliminate Filler
"6 chapter titles" and "5 key takeaways" are exact constraints in the After Prompt. When the AI knows the count, it must be selective rather than exhaustive. Specificity forces prioritization. Without counts, the model tends to pad output with marginal points to appear thorough.
Format Instructions Reduce Editing Time
The After Prompt specifies MM:SS timestamp format, bullet points for takeaways, and a 350-word total limit. These aren't style preferences — they're publishing requirements baked into the prompt. Outputs arrive in a format that drops directly into a CMS, saving the editing step that most vague prompts require.
Dual CTAs Serve Multiple Goals
The After Prompt ends with two separate CTAs — one for subscribing and one pointing to a URL. This mirrors how real show notes function: they serve the episode listener and the content marketing funnel simultaneously. Separating the CTAs ensures neither gets diluted by combining them into one vague sentence.
The framework behind the prompt
The Strategy Behind Podcast Show Notes
Show notes occupy a unique position in content strategy: they serve three distinct audiences simultaneously. Listeners use them for navigation and reference. Search engines index them as text representations of audio content. Potential listeners use them to decide whether an episode is worth their time.
Most creators treat show notes as a transcript summary. That framing misses the strategic opportunity entirely.
The inverted pyramid principle — borrowed from journalism — argues that the most important information should appear first. Applied to show notes, this means your two-sentence summary needs to answer three questions immediately: who is speaking, what problem they solve, and why a listener should care today. Generic summaries fail this test.
Chapter structuring benefits from principles found in information architecture: chunking content into labeled segments reduces cognitive load and helps users find the section most relevant to their current need. Research on podcast listening behavior consistently shows that skippable chapters increase episode completion rates because listeners feel in control — they know what's coming and can navigate freely.
From an SEO standpoint, show notes function as long-form anchor content for audio assets that search engines cannot crawl directly. Google's guidelines on podcasting content recommend structured text descriptions, explicit speaker attribution, and keyword-natural language — all of which a well-structured prompt produces automatically.
The COPE framework (Create Once, Publish Everywhere) from content strategy is directly relevant here. A single show notes prompt — built with the right structure — generates content usable across your podcast platform, website, newsletter, and social channels simultaneously. The chapter titles become social copy. The takeaways become email bullets. The summary becomes your SEO meta description.
Prompting AI for show notes without this strategic context produces content that technically describes the episode but fails every downstream use case. Structure is not formatting — it's function. The right prompt architecture encodes that function before the AI writes a single word.
Prompt variations
You are a podcast editor and writer.
Create show notes for a solo podcast episode for an audience of freelance designers and creative professionals. Tone: casual, direct, and motivating.
Here is a brief episode description: I talked about how to price your freelance work without undercharging, covered three pricing models, and shared a story about a client negotiation that changed how I set rates.
- Write a 2-sentence episode summary that hooks a reader browsing a podcast app.
- List 5 chapter titles with estimated timestamps in MM:SS format (assume a 30-minute episode).
- Provide 4 key takeaways as short, punchy bullets — each under 15 words.
- Write one CTA sentence directing listeners to download a free pricing worksheet at designrates.com/worksheet.
Keep the full output under 300 words.
You are a podcast producer specializing in interview-format business shows.
Create show notes for a guest interview episode. Audience: early-stage startup founders. Tone: smart, concise, and respectful of the reader's time.
Guest: Sarah Chen, Head of Growth at Clearbit. Topic: How she built a content-led growth engine from zero to 100k monthly readers in 18 months.
- Write a 3-sentence episode intro that names the guest, her role, and the core lesson of the episode.
- Generate 7 chapter titles with timestamps in MM:SS format (assume a 45-minute episode).
- List 5 key quotes or insights attributed to the guest — paraphrase if you don't have exact quotes.
- Add a "Guest Resources" section linking to: Clearbit.com, Sarah's LinkedIn (/in/sarahchen), and the book she mentioned: "Obviously Awesome" by April Dunford.
- Close with a single subscribe CTA and a second CTA to explore the full episode archive at thegrowthpod.com/episodes.
Total output: under 400 words.
You are a content strategist and internal communications writer.
Create structured show notes for an internal company podcast episode. Audience: mid-level managers and team leads at a 500-person technology company. Tone: professional, clear, and action-oriented — not corporate or formal.
Episode topic: Q3 strategy update. The episode covers three priorities — reducing customer churn, expanding into mid-market accounts, and investing in AI tooling for the support team.
- Write a 2-sentence summary suitable for an internal Slack announcement.
- List 5 chapter titles with timestamps in MM:SS format (assume a 35-minute episode).
- Provide 4 action items for managers — framed as "what you should do by end of quarter."
- Add a "Key Decisions" section with 3 bullet points covering decisions leadership has already made.
- End with one sentence pointing managers to the internal wiki at company.notion.site/q3-strategy for supporting documents.
Keep the output under 350 words. Avoid jargon.
You are a YouTube content strategist and copywriter.
Create a YouTube video description based on this podcast episode. The show targets small business owners interested in local SEO and digital marketing. Tone: helpful, plain-English, and encouraging.
Episode topic: How a bakery owner tripled her Google Maps traffic in 90 days using three free tools — Google Business Profile, Whitespark, and Search Console.
- Write a 3-sentence YouTube description opening that includes the keyword "local SEO for small business" naturally in the first sentence.
- List 6 chapter titles with timestamps formatted as H:MM for YouTube (assume a 38-minute episode).
- Add 4 bullet points under a "What You'll Learn" header.
- Include a "Tools Mentioned" section with Google Business Profile (business.google.com), Whitespark (whitespark.ca), and Google Search Console (search.google.com/search-console).
- End with a CTA to subscribe and a second CTA linking to a free local SEO checklist at localgrowth.co/checklist.
Total output: under 450 words. Write for YouTube's search algorithm — use specific, plain-language phrases.
When to use this prompt
Marketing Teams Repurposing Webinars
Turn recorded webinars into podcast-style show notes and chapters for faster distribution across channels.
Product Managers Sharing Customer Insights
Publish episode notes that highlight user pain points, decisions, and next steps for internal teams.
Sales Leaders Building Enablement Assets
Create skimmable notes with key quotes and takeaways reps can reference before calls.
Customer Success Teams Educating Users
Ship clear chaptered summaries that help customers find the exact section they need.
Pro tips
- 1
Paste a 5-line episode brief so the AI knows the goal, not just the transcript.
- 2
Define your timestamp rule up front so chapters match your editing style.
- 3
Name 3 topics to emphasize so the notes support your campaign priorities.
- 4
Add 2 phrases you must avoid so the tone stays on-brand and compliant.
Once you have a working prompt template, you can scale production across an entire season without quality drift. Here's how to build a system that holds up at volume:
Create a prompt library by episode type. Solo episodes, guest interviews, panel discussions, and roundup episodes all have different structural needs. Maintain separate prompt versions for each — same core framework, different chapter counts and section labels.
Anchor your tone with a style reference. Add 2–3 sentences from your best-performing show notes as a tone example inside the prompt. This acts as a style anchor that prevents the model from drifting toward generic language over time. Example addition: "Match the voice in these sample sentences: [paste 2 sentences]."
Build a QA checklist. After generating, verify five things before publishing:
- Does the summary mention the guest's name and their single clearest insight?
- Do chapter titles use descriptive verbs, not generic nouns?
- Are all timestamps within 2 minutes of the actual audio position?
- Does the CTA include the exact URL?
- Is the total word count within 10% of your target?
Use batch prompting for efficiency. If you have 4 episodes releasing in a week, paste each transcript into the same prompt structure in separate runs. Don't try to batch all four in one prompt — the model will conflate content across episodes. One transcript, one run, every time.
This systematic approach turns a 30-minute weekly task into a 10-minute one — and keeps your back catalog consistent.
The core prompt structure works across formats, but each format has specific requirements worth addressing directly.
B2B educational podcasts need takeaways framed as decisions or actions, not observations. Replace "5 key takeaways" with "5 decisions your team should consider this quarter" to orient the output toward a professional audience that measures content by what they can act on.
True crime and narrative podcasts benefit from chapter titles that build tension rather than describe content. Instruct the AI: "Write chapter titles that hint at what happens next — like a chapter title in a thriller, not a textbook." This small change produces dramatically more compelling chapter listings.
Health, wellness, and coaching podcasts often face compliance constraints. Add a sentence like: "Do not include any language that could be interpreted as medical advice. Use phrases like 'the guest suggests' rather than 'you should.'" This protects your brand and keeps the output legally safe.
Comedy and entertainment podcasts require the lightest hand. Specify: "Avoid summarizing jokes — describe the topics covered and the mood of the conversation instead. Preserve the host's irreverent tone throughout."
Nonprofit and mission-driven podcasts should include an impact-framing instruction: "Frame takeaways around community impact and shared values, not individual productivity gains." This keeps the output aligned with your organization's voice and donor-facing messaging.
Show notes aren't just for listeners — they're a search asset. A well-structured prompt can produce notes that rank for long-tail queries your audience is actively searching.
Add a keyword instruction to your prompt. Include a line like: "Incorporate the phrase 'freelance pricing strategy' naturally in the first 50 words of the summary without keyword stuffing." The AI will weave the term into the opening, which Google weights heavily.
Request a meta description separately. After generating the full show notes, run a follow-up prompt: "Write a 155-character meta description for these show notes that includes the phrase 'podcast episode' and the guest's name." This gives you an SEO-ready snippet without editing the main output.
Use descriptive chapter titles as H2 anchors. Instruct the AI: "Write chapter titles as complete phrases, not single words — for example, 'How to Build a Pricing Strategy from Scratch' instead of 'Pricing.'" Descriptive H2s improve both readability and semantic indexing.
Include guest name and credentials in the summary. Search engines index named entities. A summary that opens with "In this episode, Sarah Chen, Head of Growth at Clearbit, explains..." captures branded and non-branded search traffic for both the guest and the topic.
Target featured snippets with structured lists. The 5-bullet takeaway format mirrors the list format Google often selects for featured snippets. Keep each bullet under 60 characters when possible — that matches the character counts Google typically pulls.
When not to use this prompt
When This Prompt Pattern Isn't the Right Tool
This prompt structure works best when you have a clear episode structure, a defined audience, and a need for consistent, repeatable output. There are several situations where it's not the right approach:
Very short episodes under 10 minutes often don't need 6 chapters or 5 takeaways. Forcing that structure onto a brief episode produces padding, not insight. For short-form content, simplify the prompt to: summary, 3 chapters, and one CTA.
Highly sensitive or regulated content — mental health discussions, legal commentary, medical topics — shouldn't be fully delegated to AI output without expert human review. The prompt helps with structure, but a compliance reviewer must read every line before publication.
Episodes with minimal content signal — where the host goes off-script, the conversation meanders, or the topic is deliberately abstract — may produce inaccurate takeaways. If your episode doesn't have clear, extractable insights, the AI will either hallucinate specifics or produce vague output. Manual notes may be faster.
Premium or paywalled content where show notes are the primary marketing asset may need a more nuanced, human-crafted tone than structured AI output can reliably deliver at first pass.
In these cases, use the prompt as a first draft starting point only — not as a publish-ready output — and allocate time for meaningful human editing.
Troubleshooting
Chapter titles are too generic — 'Introduction,' 'Discussion,' 'Conclusion'
Add a direct instruction to your prompt: "Write chapter titles as specific, descriptive phrases that tell the reader what they'll learn — not generic labels like 'Introduction' or 'Wrap-Up.'" You can also provide an example: "For example, instead of 'Guest Background,' write 'How Sarah Built a $2M Business with No Investors.'" Descriptive examples immediately raise the specificity of outputs.
Takeaways read as restatements of the summary, not distinct insights
Separate the instructions explicitly: "Each takeaway must introduce a concept not mentioned in the episode summary." You can also instruct the AI to frame takeaways as "what the listener should do or think differently about after this episode." Action-oriented framing forces the model to extract practical insight rather than restating the episode arc.
The AI fabricates guest quotes or invents resources that weren't mentioned
Add a hard constraint: "Only include quotes that appear verbatim in the transcript. Only list resources that are explicitly named in the transcript. If none are present, write 'No resources mentioned in this episode.'" Hallucination in show notes is a trust problem — explicit prohibition is the most reliable fix.
Output consistently ignores the word count limit
Move the word count instruction to the first line of the prompt, before the role assignment: "Important: Your total output must not exceed 350 words." Then add a trim hierarchy at the end: "If over 350 words, shorten the takeaways first, then the chapter descriptions. Never cut the summary or CTAs." Position and prioritization together produce consistent compliance.
Show notes sound inconsistent across episodes despite using the same prompt
Paste 2–3 sentences from a previous episode's show notes into the prompt as a style anchor. Add: "Match the voice and sentence length of these example sentences exactly: [paste examples]." Inconsistency usually means the AI is reinterpreting your tone instruction differently each run. A concrete example eliminates that ambiguity.
How to measure success
How to Evaluate Your Show Notes Output
Before publishing AI-generated show notes, run through this quality check:
Summary quality:
- Does the first sentence name the guest or topic and the core benefit to the listener?
- Is it under 50 words and free of filler phrases like "In this episode, we dive into"?
Chapter quality:
- Are chapter titles descriptive phrases, not single-word labels?
- Do timestamps distribute across the episode length without clustering?
- Would a first-time listener understand what each chapter covers without listening?
Takeaway quality:
- Does each takeaway introduce a distinct concept not covered in the summary?
- Are they framed as something the listener can act on or think differently about?
- Are they under 20 words each?
CTA quality:
- Does the CTA include the exact destination URL?
- Is the ask specific — subscribe, book a demo, download a resource?
Overall output:
- Is the total word count within 10% of your target?
- Does the tone match your established show voice?
A publish-ready output passes all of these checks on the first run. If you're editing more than 20% of the content, refine the prompt — not the output.
Now try it on something of your own
Reading about the framework is one thing. Watching it sharpen your own prompt is another — takes 90 seconds, no signup.
Build a reusable show notes prompt for your podcast — customized for your audience, tone, and publishing format.
Try one of these
Frequently asked questions
Yes — substitute the outline for the transcript and tell the AI explicitly: "Use this episode outline, not a transcript." Provide 5–8 bullet points covering the main topics, guest name, and key stories. The AI will estimate chapters and takeaways from the outline. Results won't be as granular as transcript-based notes, but they're solid enough for pre-publication drafts or episode teasers.
Tell the AI to estimate timestamps proportionally based on episode length. Include a line like: "Assume a 42-minute episode. Distribute timestamps evenly across the transcript." The estimates won't be frame-accurate, but they'll be close enough for a chaptered description. For precision, cross-check against your audio editor's waveform after generating.
Add a tone anchor sentence to your prompt, such as: "Write the way the host speaks — casual, direct, and occasionally self-deprecating. Avoid phrases like 'In this episode, we explore...'" You can also include 2–3 example sentences from past show notes that match your voice. The more specific your tone instruction, the less cleanup you'll need.
Replace two fields: audience description and tone. For a true crime podcast, you might write "true crime enthusiasts who follow high-profile cases" and "suspenseful, detailed, and neutral." For a wellness podcast: "women 30–50 managing stress and burnout" and "warm, empathetic, and encouraging." Everything else in the prompt structure stays the same — only the context values change.
This happens when the transcript contains no explicit references to tools, people, or links. Add a fallback instruction like: "If no resources are identifiable, list 2–3 related resources the audience would find useful based on the episode topic." This prevents the AI from silently skipping a required section, which breaks your output format.
Yes, with two adjustments. First, change the timestamp format to H:MM for YouTube compatibility. Second, replace "show notes" with "video description" or "webinar recap" in the prompt so the AI frames the output correctly. The structure — summary, chapters, takeaways, CTA — applies equally to video formats. See the YouTube variation above for a ready-to-use version.
Add a hard constraint at the end of the prompt: "If your draft exceeds 350 words, cut from the takeaways section first, then trim chapter descriptions. Do not cut the CTAs or the summary." This gives the AI a prioritized editing rule, not just a word ceiling. Most models will comply with an explicit trim hierarchy when they struggle with a flat word count limit.
Build a template. Replace the transcript field and CTA URL each time — everything else (audience, tone, section structure, word count) stays fixed. Store your master prompt in a shared doc or your CMS. This turns show note production into a repeatable workflow rather than a one-off task, and it keeps your notes consistent across hundreds of episodes.