Podcast SEO. How episodes get found in Google and AI answers.
Podcast SEO means your episodes show up when someone searches Google or asks an AI assistant, and the episode page is the weakest part of it. On Future of Hospitality, a show we produce, about 80% of the Google impressions in October went to the articles on the site, and 4% to the episode pages. Below is what we measured, and the setup we now use for every show.
What we measured on our own show
We read Google Search Console for futureofhospitality.org over the 28 days up to 31 October 2026: 204 clicks and about 17,000 impressions in total. The show has 11 series and 130 episodes, published in English, Dutch and German, which is why there are more episode pages than episodes.
| Page type | Pages seen in Google | Clicks | Impressions |
|---|---|---|---|
| Articles | 348 | 120 | 13,711 |
| Series pages | 33 | 15 | 950 |
| Episode pages | 176 | 7 | 735 |
| Homepage (three languages) | 3 | 40 | 695 |
The episode pages are not broken. Each one has the full video, a transcript with chapters and a summary, and 277 of the show's videos are in Google's video index. They answer the question “what was said in this episode”, and almost nobody types that. The articles answer a question people do type, such as how to handle guest complaints in a hotel, and the episode is the evidence inside them. Since October, every episode of the show has its own article built around one question from the conversation.
Why an episode page alone rarely ranks
A search engine ranks pages for questions. An episode title is usually written for people who already listen: a guest's name, a phrase from the talk, an episode number. That is fine for your feed and weak for search.
AI answers do not change this. Google says its AI features rely on the same Search index and core ranking systems as normal results, so a page that does not rank in Google rarely ends up in an AI answer there. Other assistants, like ChatGPT and Claude, read the HTML your server sends. A transcript that only appears after a script runs may never be read.
The setup we use for every episode
- One article per episode, built around one question a buyer would type. The question goes in the title, and the first two sentences answer it. The episode is the proof: a quote, a number, a story from the guest.
- One episode page, with the video as the main thing on the page and a player that loads without a click. Google calls that a watch page, and only a watch page can show up as a video result.
- The full transcript on the episode page only, in the HTML, split into chapters. The article quotes it and links to it, so the two pages never compete for the same question.
- A page per guest, with their role and every episode they were on.
- Links in both directions: the article links to the episode and the series, the episode links back to the article.
- Structured data that matches what is on the page: PodcastEpisode, VideoObject with the chapters as key moments, and a breadcrumb. It does not lift rankings on its own, but it tells Google which page is the episode and where each chapter starts.
The show notes that go with every episode follow the same idea; we wrote them up in our show notes guide.
The mistake that kept 108 videos out of Google
In October, Search Console listed 108 videos on Future of Hospitality as not indexed, with one reason: “Video isn't on a watch page”. On those episode pages the video sat behind a play button and only loaded after a click. Google's video guidelines say it plainly: don't rely on user actions such as clicking to load the video.
We changed the player so the video loads with the page, on 30 October, and asked Google to check the pages again. If you run a podcast site, open the Video indexing report in Search Console. It names this problem in one line.
What AI assistants need from a podcast site
- The text in the page itself: title, summary, chapters and transcript in the HTML the server sends.
- One clear answer per page, in the first sentences, in the words a buyer would use.
- Names with roles: who the guest is, where they work, who hosts. An assistant can only quote someone it can name.
- No second version of a page written for AI. Google says separate pages for variations of a question, made mainly to steer rankings or AI answers, break its scaled content abuse policy.
We also publish an llms.txt file, but we don't count on it: Google ignores it, and in an Ahrefs study 97% of llms.txt files got no requests at all in May 2026.
How we do this for a client
Our website agent, Page, builds the episode page, the guest page and the article for every episode, on your own site, and a producer checks them before you approve. If you want to see it on your own show first, send us your next episode and we produce it for free. What the whole setup costs next to software or an agency is on the podcast cost page.
Sources
- Google Search Central: Video SEO best practices
- Google Search Central: Optimizing your website for generative AI features
- Anthropic: Web fetch tool (no JavaScript)
- Ahrefs: llms.txt study
- Own data: Google Search Console of futureofhospitality.org, read on 3 November 2026.