What AI search engines are
They're tools that, instead of giving you a list of links, give you a written answer and cite where it came from: ChatGPT with search, Perplexity, Gemini, Copilot or Google's AI answers. To answer, they crawl the web, pick the pages that answer best and link to them. If your podcast is among those sources, the AI recommends it by name. This is sometimes called GEO, though at heart it's good old SEO with a few twists.
AI doesn't listen: it reads
This is the key point. An hour of brilliant conversation uploaded only to Spotify is, to a search engine, a title and two lines of description. What it can read from your podcast:
- The transcript of the episode, checked and with names spelled correctly.
- A summary that says what it's about and what listeners will learn.
- Chapters with timestamps and clear titles.
- A guest profile: name, role and what they do, linked to their website.
- Subtitles on YouTube and in social clips.
- Resources mentioned: books, studies and links from the conversation.
Automatic transcripts are a good starting point, but check them: a misspelled name or a nonsensical sentence undermines credibility.
A page for every episode
Audio platforms are shop windows, but they're not yours. Your website is. Ideally, every episode has its own page with its own address and all its content as text:
- A title that explains the topic.
- A summary right at the top: what it's about and who it's for.
- An embedded audio or video player.
- Chapters and key points with question-style subheadings.
- The full transcript, visible on the page and not hidden in a PDF.
- Links to every platform.
Titles and descriptions that answer questions
People talk to AI the way they'd talk to a person: in full questions. The more your titles and descriptions resemble those questions, the easier it is to be chosen. “How to speak in public without freezing, with…” works far better than “Episode 23: we talk to…”. A good episode structure helps, because each segment already answers a question.
YouTube, your best ally
It's one of the sources that appears most in AI answers, especially Google's, and it has one huge advantage: subtitles can be read.
- Upload the full episode, even with a still image; video is better.
- Check the automatic subtitles or upload your own.
- Add chapters with timestamps in the description.
- Write a full description, with a summary and links.
- Post short clips that each answer a single question.
Let the crawlers in
It happens more often than you'd think: websites accidentally block the bots AI tools use to search. Check with whoever runs your site:
If you're unsure about the feed, it's explained in how to publish your podcast.
How AI decides what to cite
For anyone with some SEO experience, this is the shift in mindset. When someone asks a question, the AI search engine doesn't search for that exact phrase: it breaks it into several smaller queries, retrieves passages from different pages and builds the answer from those that best address each one. What gets cited isn't your whole page, but a specific paragraph.
Two practical consequences. First: ranking well on Google isn't enough; AI answers increasingly cite pages that aren't in the classic top 10, because they answer one of those sub-questions better. Second: the winners are the ones with passages that make sense on their own, without needing to read what comes before or after.
Write passages that can be cited
- Self-contained answers. Under each question-style subheading, a 40 to 60-word paragraph that answers it fully. That's what AI copies and links to.
- Your own data. A figure from your experience (“across 7 episodes we recorded…”), a study you cite with its source or a fact your guest provides. Original content gets cited; generic content doesn't.
- Clear definitions of the concepts in your field, in one sentence.
- Direct quotes from your guest, with their name and role. They add authority and are easy to attribute.
- FAQs for each episode: three or four real questions with short answers at the bottom of the page.
An example: from episode 23 to a page that gets cited
Here's how the same episode listing changes when you design it to be found:
Before
Description: “Today we have a very special guest. Don't miss it!”
Chapters: none
Transcript: no
After
Description: two sentences covering the problem, who it's for and which techniques are explained.
Chapters: “04:10 Why your mouth goes dry”, “12:30 The 4-2-6 breathing exercise”…
Transcript: full and checked, on the page
The second isn't longer for the sake of it: each chapter answers a question someone might ask an AI, and each one is another chance to appear.
Crawlers: search isn't the same as training
The big AI companies have split their bots into two groups. Some crawl to train models (OpenAI's GPTBot or Anthropic's ClaudeBot). Others crawl to search and cite in answers (OAI-SearchBot, Claude-SearchBot, PerplexityBot), and there's a third type that visits your page when a user asks for it to be read right then (ChatGPT-User, Claude-User, Perplexity-User).
You're free to decide whether your content is used for training, but if you block the search bots, you disappear from their answers. A starting point for your robots.txt:
# AI search engines: let them in (they decide whether to cite you) User-agent: OAI-SearchBot Allow: / User-agent: ChatGPT-User Allow: / User-agent: Claude-SearchBot Allow: / User-agent: Claude-User Allow: / User-agent: PerplexityBot Allow: / User-agent: Perplexity-User Allow: / # Model training: your call (blocked here) User-agent: GPTBot Disallow: / User-agent: ClaudeBot Disallow: / Sitemap: https://yourdomain.com/sitemap.xml
Structured data for every episode
Structured data tells search engines unambiguously that the page is an episode, which show it belongs to, who takes part and where the audio is. It won't work miracles, but it removes doubt. Here's a PodcastEpisode example to paste into the page:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "PodcastEpisode",
"name": "How to speak in public without freezing",
"url": "https://yourdomain.com/podcast/public-speaking/",
"datePublished": "2026-11-10",
"description": "Voice and breathing techniques for speaking confidently in public.",
"timeRequired": "PT38M",
"partOfSeries": {
"@type": "PodcastSeries",
"name": "Your podcast name",
"url": "https://yourdomain.com/podcast/"
},
"actor": [{ "@type": "Person", "name": "Guest name" }],
"associatedMedia": {
"@type": "MediaObject",
"contentUrl": "https://yourhost.com/episode-24.mp3"
}
}
</script>
You can validate it with Google's Rich Results Test or the schema.org validator.
Authority and mentions
AI prefers reliable sources, and trust is built through consistency and what others say about you. Use the same podcast name and description on your website, platforms and social media; explain who's behind it with a genuine background page; and get yourself mentioned: guests linking to their episode, the media, other podcasts, industry directories. Every mention is another signal that you exist and what you're about.
How to tell if it's working
- Ask several AIs about your topic yourself and see if you appear.
- Check your website analytics for visits coming from chatgpt.com, perplexity.ai or other AIs.
- Ask people who contact you how they found you.
Going further: advanced measurement
- AI traffic in Google Analytics. Create a segment or referral source report with this regular expression:
chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai. You'll see how many people arrive from each assistant and which pages they land on. - A question panel. Write down 15 or 20 questions your ideal listener would ask and put them to ChatGPT, Perplexity, Gemini and Claude once a month. Note whether you appear, in what position and which page is cited. It's the most direct metric available today.
- Server logs. Check that OAI-SearchBot, Claude-SearchBot or PerplexityBot are visiting your episode pages. If they aren't, something is stopping them.
- What your clients say. Add “an AI assistant (ChatGPT, Perplexity…)” as an option in the “how did you hear about us?” field of your form.
What you don't need (for now)
There's a lot of noise around this topic. Two examples. The llms.txt file, an “index for AI” proposed a while ago, isn't an official standard, and Google has said explicitly that it doesn't use it for search or its AI answers; having one does no harm, but don't expect anything from it. And rewriting your copy “for AI”, with artificial phrasing or keyword lists, doesn't help either: what works is clear, original, well-organised content, the same thing that already worked for people.
A plan for the next 30 days
- 01Week 1Review robots.txt, your firewall and the sitemap. Set up the AI traffic report and your question panel.
- 02Weeks 2 and 3Build the full page for your 5 most useful episodes: question-style title, summary, chapters, checked transcript, FAQs and structured data.
- 03Week 4Upload those episodes to YouTube with subtitles and chapters, ask guests to link to their page and repeat the question panel.
To sum up
For an AI to recommend your podcast, it has to be able to read it: transcripts, a page for every episode, titles that answer questions, YouTube with subtitles and a website that doesn't block crawlers. It isn't magic: it's organising what you already have.