The operator-grade guide to AI-generated podcast show notes — the 6-section template that ranks for episode-page queries, the exact AI tells to ban, the tools that generate them (Castmagic, Descript, Kompozy), and the Persona Brief override that fixes generic tone for every future episode.
AI-generated show notes default to a "summary plus chapter list" format that reads like AI and ranks for almost nothing. The structure that works is a 6-section template: a 2-3 sentence hook, a 2-sentence guest intro with a link, 5-7 timestamped key topics, 3-5 verbatim quotable moments, every link and resource mentioned, and a brief CTA. Total 400-600 words. Tools like Castmagic ($23/mo), Descript ($24/mo), and Kompozy (Creator $49/mo) can all generate this — but only after a Persona Brief override that bans the standard AI tells ("In this episode, we discuss…", "Tune in to learn…") and forces the 6-section shape. Show notes are the SEO surface of every episode; the template is the difference between an episode page that ranks for the guest name and a bullet list nobody reads.
Show notes are the only durable, indexable text most podcast episodes ever produce. The audio sits inside Apple and Spotify where Google barely reaches; the show notes page on your own site is the surface a search engine can actually crawl, rank, and send traffic to. Done well, a single episode page ranks for the guest\x27s name, the episode topic, and dozens of long-tail queries that compound month after month. Done with default AI output, it is a bulleted summary that nobody reads and nothing links to.
The gap between those two outcomes is almost entirely structural. Every consumer show-notes tool in 2026 — Castmagic, Descript, Capsho, the show-notes module inside an orchestration platform — ships the same default: a paragraph of summary followed by a chapter list. That format reads as machine-generated to a human and ranks badly because it does not match what a person searching for an episode actually wants. They want to know who the guest is, what got said, where to jump, and what was referenced. The summary-and-chapters format answers none of those directly.
This guide is the operator-grade reference: the 6-section template that ranks, the specific AI tells to strip out of podcast show notes (they differ from the tells in blog or social copy), the honest tool comparison with verified 2026 pricing, and the one-time Persona Brief override that makes every future episode\x27s show notes match your voice automatically. Pairs with the broader [ai-podcast-tools-2026](/ai-podcasting/ai-podcast-tools-2026) stack reference and the [transcription-quality](/ai-podcasting/transcription-quality) deep-dive, since show-notes quality is capped by transcript quality upstream.
Audio is a closed surface. When a listener finds your episode inside Apple Podcasts or Spotify, that discovery happened inside an app whose ranking signals you barely influence and whose catalog Google does not meaningfully index. The episode page on your own website is the opposite — it is open, crawlable text that a search engine reads, ranks, and routes traffic through. For a podcast that wants to grow beyond the listeners the podcast apps happen to surface, the show-notes page is the highest-leverage owned asset per episode.
The mistake most podcasters make is treating show notes as a chore to dispatch in the publishing flow rather than a ranking asset to engineer. They paste whatever Castmagic or Descript generated, hit publish, and move on. The output is a competent summary that satisfies the "did I write show notes" checkbox and ranks for nothing, because a summary-plus-chapters block does not match the shape of the queries real people type when they are looking for a specific episode. Engineering the structure to match query intent is the entire game, and it is a one-time investment that compounds across every episode you ever publish.
Two practical consequences follow. First, the transcript quality upstream sets the ceiling: misheard guest names and mis-transcribed numbers poison the highest-value parts of the show notes (the quotable moments and the guest intro). Get the transcript right first — see [transcription-quality](/ai-podcasting/transcription-quality) for the engine comparison. Second, the structure has to be enforced at the prompt level, not fixed by hand after generation, or you are back to spending the manual time the AI was supposed to remove.
The template that ranks is six sections in a fixed order. Each section earns its place by answering a question a searcher or a scanning reader actually has, and the order matters because it front-loads the two things that drive a listen decision (the hook and the guest) before the reference material.
Total length lands at 400-600 words — long enough to cover the topic for a search engine, short enough that a human actually reads it. The discipline is in the order and the verbatim quotes, not the word count. A 450-word document in this shape outranks a 1,200-word AI summary of the same episode because it matches intent and the long one does not.
| Section | Reader question it answers | Primary SEO value | Most common AI failure |
|---|---|---|---|
| Hook | Is this worth my time? | Topic-query relevance | Opens with "In this episode we discuss" |
| Guest intro | Who is this person? | Ranks for "[guest name]" queries | Generic "our amazing guest" with no link |
| Key topics | Where do I jump? | Long-tail sub-topic queries | Chapters named "Introduction / Conclusion" |
| Quotable moments | What actually got said? | Shareable, screenshot-ready | Paraphrases instead of verbatim quotes |
| Links mentioned | What did they reference? | Inbound links + adjacent queries | Section omitted entirely |
| CTA | What do I do next? | Conversion, not ranking | Performative subscribe-and-review pile-on |
The generic "signs of AI writing" lists matter, but podcast show notes have their own dialect of tells — phrases every show-notes tool reaches for by default because they were trained on a decade of podcast-template boilerplate. These are worse than ordinary AI vocabulary because experienced podcast listeners have seen them ten thousand times and read them as "nobody wrote this." Ban every one of them at the Persona Brief level so they never appear in the first place.
The pattern across all of these: they are framing phrases that describe the episode instead of delivering it. A human writing show notes leads with what was actually said; an AI on default settings leads with meta-description of the conversation. Banning the phrases forces the model into the human pattern.
Episode pages compete for two distinct query types, and the 6-section template is built to win both where the default format wins only one.
The default summary-and-chapters format only addresses the first query type, and weakly: it mentions the guest, lists generic chapters, and stops. It has no verbatim quotes to be shared (which is how episode pages earn natural links), no named topics (which is how they rank for sub-topic queries), and no links section (which is how they earn inbound authority). The 6-section template is not stylistically nicer than the default — it is structurally aimed at the queries that actually exist.
Three tools own the show-notes generation slot in 2026, and they are not interchangeable. Castmagic is a dedicated transcript-to-content engine; Descript is an editor that produces show notes as a side effect; an orchestration layer treats show notes as one of many fan-out outputs from a single Persona Brief. All three require the override below to escape the generic tone — the override is the work, the tool is the delivery.
| Tool | Entry price (2026) | What it is | Best for | Watch-out |
|---|---|---|---|---|
| Castmagic | $23/mo | Dedicated transcript-to-content engine | Show notes as the primary output, fastest path | Default tone is "SaaS landing page"; needs the override |
| Descript | $24/mo | Text-based audio/video editor; notes are a byproduct | Podcasters who already edit in Descript | Show notes are a secondary feature, thinner structure control |
| Kompozy | Creator $49/mo | Orchestration layer; notes are one fan-out of many | One Persona Brief driving notes + clips + blog + newsletter | Worth it once you need 3+ output types per episode, not for notes alone |
The decision rule is straightforward. If show notes are the only AI output you need and you want the cleanest dedicated tool, Castmagic at $23/mo is the specialist. If you already edit episodes in Descript ($24/mo), the show notes fall out of a tool you are already paying for. If you are fanning each episode into clips, a blog draft, a newsletter, and social posts on one consistent voice, an orchestration layer is the pick because the show notes inherit the same Persona Brief as every other output — there is no separate voice calibration for the notes. See [pricing](/pricing) for the orchestration tiers and [content-repurposing](/repurpose) for the full episode fan-out pattern.
Generic show-notes AI does not know your podcast\x27s voice, so it averages to the "Inc.com explainer" register that every other show also gets. The fix is a one-time Persona Brief addition that you write once and that applies to every episode forever. Four components:
This is the highest-ROI 30 minutes in the entire show-notes workflow. Written once, it converts every future episode\x27s output from generic to on-voice with no per-episode effort. The override is what separates podcasters who ship AI slop from podcasters who ship publishable show notes end-to-end — same tool, same transcript, completely different output, driven entirely by the brief.
With the override in place, the per-episode process collapses to a short review pass. The structure is enforced by the brief, so the human is checking accuracy, not rewriting prose.
Episode one of this workflow takes 15-20 minutes of review. By episode ten, with a stable custom vocabulary feeding the transcription engine and a calibrated brief, it drops to 5-8 minutes. The compounding is the point: the structure and voice are solved once at the brief level, and every subsequent episode rides that investment.
The honest limits matter because trusting the AI past them is how a credible show ships a wrong attribution to its whole audience. AI owns the operator layer of show notes — structure, drafting, formatting, link extraction. It does not own the accuracy layer.
Three things stay human. First, quote accuracy: the model will occasionally attribute a line to the wrong speaker or smooth a verbatim quote into a paraphrase, and both are fatal for a pull-quote whose entire value is that a real person said it. Second, proper-noun and number fidelity: a guest who said "$4.2 million in 11 months" must not become "around four million in under a year" — the specificity is the credibility, and the transcription layer is where it gets lost. Third, the editorial judgment of which moment is actually the hook: a clip-detection model and a show-notes model both score surface signals, and the moment your audience has been waiting for is often the one with no audio-energy spike. Reclaim the time the structure and drafting would have cost, and spend it on those three checks.
If you remember one thing: the structure is the ranking, not the word count. The 6-section template (hook / guest / topics / quotes / links / CTA) ranks because it matches the queries real searchers type; the default AI summary-and-chapters format does not. Pick a tool to deliver it — Castmagic at $23/mo as the specialist, Descript at $24/mo if you already edit there, or an orchestration layer if show notes are one of several outputs you fan from one [content-repurposing](/repurpose) brief. Then write the Persona Brief override once: ban the tells, force the 6 sections, require verbatim quotes. Everything after that is a 5-minute accuracy check per episode. Start with [pricing](/pricing) to size the orchestration tier, or read the full stack in [ai-podcast-tools-2026](/ai-podcasting/ai-podcast-tools-2026).
400-600 words. Long enough to rank for episode-specific and topic queries, short enough that a human actually reads it. Shorter under-covers the topic for search; longer and the reader scrolls past. The discipline is in the 6-section structure and verbatim quotes, not in hitting a word count.
Hook (2-3 sentences), guest intro (2 sentences with a link), key topics covered (5-7 timestamped bullets), 3-5 verbatim quotable moments attributed to the speaker, links and resources mentioned, and a brief CTA. It ranks because each section answers a question a real searcher or scanning reader has — unlike the default AI summary-plus-chapters format, which only weakly addresses guest-name queries and ignores topic queries.
Castmagic ($23/mo) is the dedicated specialist and the fastest path if show notes are your only output. Descript ($24/mo) is right if you already edit episodes there. Kompozy (Creator $49/mo) is the pick when show notes are one of several fan-out outputs (clips, blog, newsletter) on one Persona Brief. All three need the Persona Brief override to escape generic tone. Prices verified 2026-06-18.
Because every tool defaults to a "summary plus chapter list" trained on a decade of podcast-template boilerplate, full of tells like "In this episode we discuss" and "Tune in to learn". The fix is a one-time Persona Brief override that bans those phrases, forces the 6-section structure, and calibrates tone against a reference show. Written once, it applies to every future episode automatically.
Yes. The entire value of a pull quote is that a real person said it, word for word — that is what makes it shareable and screenshot-worthy. AI paraphrases of quotes are the single most damaging tell in show notes because they read as machine-smoothed. Pull quotes directly from the transcript with only filler-word cleanup, and verify them against the transcript before publishing.
On your podcast\x27s own website, with a unique URL per episode. Your site is the only crawlable, rankable surface you control — the podcast apps barely surface to Google. Most hosts (Buzzsprout, Captivate, Transistor) auto-generate an episode page where you can override the show-notes content. The owned page is the SEO asset; the in-app description is not.
Optional, and usually not in the same block. The full transcript helps search-engine indexing of episode-specific queries but hurts readability if embedded inline. The common pattern is publishing the full transcript on a separate linked URL ("Read the transcript") while the show-notes page stays in the 400-600 word 6-section format that humans actually read.
No. Google\x27s guidance is explicit that AI-assisted content is not penalized when output quality is high. Show notes built on a tight Persona Brief, with the 6-section structure, accurate verbatim quotes, and real links, rank the same as hand-written notes. What gets penalized is thin, generic, low-value output — which is exactly what the default summary-and-chapters format produces, regardless of whether a human or an AI wrote it.