goulahai.com

Articles · Published · By Goulah

How to Write Podcast Show Notes, Chapters, and a YouTube Description With AI

I'm a content creator and AI strategist, and I've been creating content for my own brands for about 9 years. Recording the episode is the fun part. Writing the show notes, the description, and the chapter list afterward is the part that gets skipped, even though it's what helps people find the episode and jump to the part they care about.

An AI assistant can draft all of it from one transcript in a few minutes, as long as you set it up so it can't make things up. Here's the workflow, step by step, using YouTube's official help pages for chapters and descriptions and the prompting guidance Anthropic publishes for Claude. The prompts work in any chat assistant that accepts long pasted text.

1. Start with a clean, timestamped transcript

Everything below depends on the transcript. If it has the guest's name wrong, the AI will put the wrong name in your description. I covered how to get a transcript out of YouTube Studio and clean it up in How to Find the Best Short Clips in a Long Video With AI (steps 2 and 3). The short version:

2. Set up the prompt the way long documents work best

A full episode transcript is long, so structure matters. Anthropic's prompting docs say that for large inputs (20k+ tokens), you should put the long document near the top of the prompt, above your instructions, and that queries at the end can improve response quality by up to 30 percent in their tests. They also recommend wrapping your material in XML-style tags so the model can tell it apart from your instructions. Source: Claude prompting best practices: long context prompting (Anthropic).

Give it context too. The same docs say to be clear and direct and to treat the model like a brilliant new employee who doesn't know your norms. For show notes, that means telling it the show name, the guest, and who the audience is.

3. Ask for quotes first, then everything else

This is the step that keeps the summary honest. Anthropic recommends asking the model to pull word-for-word quotes from a long document before doing the task, which grounds the answer in the actual text. Its guide on reducing hallucinations also says to explicitly give the model permission to say it doesn't know, which can drastically reduce false information. Source: Reduce hallucinations (Anthropic).

Here's a first prompt you can adapt. It gets the quotes, a short summary, the chapters, and a list of everything mentioned by name, all in one pass:

<transcript>
[paste your cleaned-up transcript here, with timestamps and speaker names]
</transcript>

<episode_info>
Show: [show name]
Guest: [guest name and what they do, spelled correctly]
Audience: [who watches or listens]
Main topic: [one line]
</episode_info>

Using only the transcript above:

1. Quote the 5 to 8 most important lines word for word, each inside <quote> tags,
   with the timestamp where it starts.
2. Write a 2-sentence summary of the episode based only on those quotes.
3. List 6 to 10 chapters for the full episode. The first chapter must start at 00:00.
   Timestamps in ascending order, each chapter at least 10 seconds long,
   titles under 50 characters that say what the section is about.
4. List every book, tool, company, or person mentioned by name, with the timestamp.

If something isn't in the transcript, don't guess. Write "not in transcript" instead.

4. Turn the chapter list into YouTube chapters

YouTube builds chapters from timestamps in your description, and it has specific rules. Per YouTube's help page, as of this post's date:

Source: Video Chapters (YouTube Help). That's why the prompt above spells out the 00:00 start, the order, and the 10-second minimum. Even so, check the list yourself: jump to each timestamp and make sure the chapter actually starts there. Transcript timestamps mark when a caption line appears, so you may need to nudge a few by a second or two.

5. Write the description

YouTube's tips for descriptions give you a clear brief to hand the AI:

Source: Tips for video descriptions (YouTube Help). In the same chat, follow up with:

Now write a YouTube description for this episode.
- The first 2 lines should say what the viewer gets from the episode and use
  the phrase "[your main keyword]" naturally. These lines show before "more".
- Then the chapter list from above, exactly as written.
- Then the "mentioned in this episode" list.
- Keep it under 5,000 characters and don't add links I haven't given you.

Add your own links (guest's site, your newsletter, sponsors) yourself. Telling the AI not to add links stops it from guessing at URLs.

6. Write show notes that sound like you

Show notes for your website, podcast app, or newsletter can be longer than the YouTube description. The fastest way to get them in your voice is to show the AI what yours look like. Anthropic's docs call examples one of the most reliable ways to steer format, tone, and structure, recommend 3 to 5 of them, and suggest wrapping them in <example> tags so the model can tell them apart from your instructions (source). The same docs say examples should be relevant and varied enough that the model doesn't copy unintended patterns.

Here are 3 sets of show notes I wrote for past episodes. Match their
length, tone, and structure. Don't copy their content.

<examples>
<example>[paste past show notes 1]</example>
<example>[paste past show notes 2]</example>
<example>[paste past show notes 3]</example>
</examples>

Write show notes for this episode using only the transcript and quotes above.

If you don't have past show notes yet, write one set by hand for this episode, then use it as the example next time.

7. Make it check its own work, then check it yourself

Anthropic's hallucination guide suggests asking the model to find a supporting quote for each claim after it writes, and to retract any claim it can't support. Run this last:

Review every claim in the description and show notes you just wrote.
For each one, find a direct quote from the transcript that supports it.
If you can't find a supporting quote, remove the claim and mark where it was
removed with empty [] brackets. Then list what you removed.

Then do a human pass before you publish:

8. Reuse the quotes for clips

The quotes from step 3 are a head start on short clips: they're already the strongest lines in the episode, with timestamps. The full clipping workflow, including how to verify each moment and cut it, is in How to Find the Best Short Clips in a Long Video With AI.

Mistakes to avoid

Official sources

Platform rules and limits in this guide are from the official pages above, accurate as of the published date. They can change, so check the linked pages before you rely on a specific rule.