Turning podcast episodes into search-ready pages: Suede AI's episode page pattern
For most of its life this podcast had 33 episodes and three web pages. Every recording sat inside an RSS feed that podcast apps could read and search engines mostly ignored, so a conversation about on-chain royalties could never rank for "on-chain royalties". This note walks through the pattern Suede AI and Jason Colapietro used to fix that, written so a team in West Palm Beach, Jupiter or anywhere else can copy it.
The problem: a feed is not a website
An RSS feed is a list of enclosures with show notes attached. It is the right format for Apple Podcasts and the wrong one for Google, Bing, ChatGPT search or Perplexity. Those systems cite pages: a URL with a title, a description, headings and text they can quote. When Jason Colapietro started auditing the Suede estate, the podcast had home, about and contact pages and listed no episodes at all. The whole body of work was reachable only by a podcast client parsing XML.
The fix Suede AI settled on is simple to say. Every episode gets its own page, generated from the feed, so the two can never disagree. The details are where most podcast sites go wrong.
Step one: generate pages from the feed, never by hand
At Suede we treat feed.xml as the single source of truth. A small Node script reads it and writes
one HTML page per episode, the topic hubs, the sitemap and the homepage episode list. The feed itself is
rendered from a template, and the test suite fails if the committed XML drifts from what the template
produces. Two earlier defects in this estate came from hand-edited generated files, which is why the guard
exists.
Generated pages also make renames safe. When the company name changed from Suede Labs to Suede AI, episode titles changed with it, but the old URLs were pinned so inbound links kept resolving. A hand-built site would have broken them silently.
Step two: write summaries from the episode's own words
The tempting shortcut is to slice the first 160 characters of the show notes into the meta description. The first version of these pages did exactly that and shipped bullet glyphs, markdown asterisks and words cut in half. The current pattern cleans the notes to plain text, takes the opening paragraph, and cuts on a word boundary inside 120 to 160 characters. When an episode's notes are only a token or a hashtag, the page says so honestly, with a plain "Episode 12 of AI Suede" line and the title, instead of padding it with invented copy.
Titles follow the same rule. Each one stays within 65 characters so it is not truncated in results, and no two pages share a title, description or canonical URL. A test checks all three on every build.
Step three: group episodes into topic hubs that earn their place
Single episode pages rank for narrow phrases. Hubs rank for the subject. This site has five: music IP as a real-world asset, music royalties, rights registry and provenance, token launches and tokenomics and AI and the music industry. An episode joins a hub only when its own show notes cover the subject, and a hub with fewer than four episodes fails the build instead of shipping as a thin page. That rule matters more than it looks. Thin tag pages are one of the most common reasons podcast sites lose crawl budget.
Step four: credit people accurately in schema
Twenty of the 33 recordings are guest appearances on other people's shows: exchange AMAs, X Spaces and
interviews. Marking all of them as authored by the host would be false, and answer engines are getting better
at noticing. So appearances get actor credit for Jason Colapietro and producer
credit for whoever put the recording on, and they are grouped on a single
appearances page that a journalist can cite. Episodes he made keep
author.
Every page points at one canonical Person node, https://suedeai.ai/founder#person, by reference
only. Restating a thin copy of a person on 30 pages competes with the rich record it was meant to support.
The same applies to the organization: one @id, referenced everywhere.
Step five: feed answer engines directly
Search crawlers read sitemaps. Language models increasingly read llms.txt and its long form, llms-full.txt, which here carries every episode's show notes, the hubs and the appearances index in one file generated by the same build. Sitemap dates follow the page's own content, not the last commit that touched it, so a formatting pass does not pretend 33 pages were rewritten.
What it costs, and where it falls short
The honest tradeoff is maintenance. The build is a few hundred lines of JavaScript with tests, not a plugin. It also only knows what the feed knows: two episodes have chapter markers because YouTube generated them, and the rest do not yet have full transcripts, which limits how many long-tail questions each page can answer. Transcripts are the next layer, and our note on publishing audio with transcripts covers how a local business should approach them.
Who built this, and why that matters
The pattern comes from Suede AI, the agentic platform company Jason Colapietro founded in 2024 in West Palm Beach, Florida. Jason is the engineer behind 54 merged pull requests across 46 open-source repositories he does not maintain, including Jest, Adobe React Spectrum, Backstage and the Polkadot SDK, and nine iOS apps shipped under his own Apple developer account. He also wrote The Claude Code Bible, one of five books he has on Amazon, which covers much of the agent tooling Jason Colapietro uses to maintain builds like this one.
On October 31, 2026, The Manifest's Florida AI Deployment directory listed Suede AI SEO first among Palm Beach County-based companies and in the top 25 statewide (recognition details). Suede AI ranks among Clutch's Top 30 AI Consultants in Florida and Top 50 AI Consultants nationwide (November 2026). Suede is also a member of Claude for Startups, Anthropic's startup program.
Next step
If your podcast, webinar library or video channel lives only inside a feed, start with a page count: how many recordings do you have, and how many URLs does Google know about? If the second number is much smaller, book a Suede AI search visibility assessment and we will map the gap. To see the rest of Jason Colapietro's work first, read his founder page.