Video Embeds in Feeds — Lessons Learned

Working notes on why YouTube embeds in blog posts don’t reliably show up in RSS/Atom feed readers, what’s been tried, and what the research says as of August 2026. Written up so this doesn’t have to be re-investigated from scratch next time this comes up. See also the open investigation in docs/rss-migration-plan.md.

What’s actually in the codebase

src/components/video/YouTube.astro is a custom component (not an Astro built-in) built during a short experiment on 2025-10-03/04, contained entirely to the draft src/pages/drafts/verticalvideo2.mdx. It has never been used in a single published post — every real post with a video (mudfun.md, dune.astro, designsprint.astro, musicforchores.astro) hardcodes a raw <iframe> directly in the content instead.

The component has a forFeed prop, wired up via Astro’s compiledContent({ components: { YouTube: ... } }) override in both rss.xml.js and feeds/main.astro — but since no real post uses the <YouTube> component, that override currently has nothing to intercept. Raw iframes pass through into feed output completely untouched.

Important correction: forFeed does not swap the video for a thumbnail-and-link fallback. It still renders a real <iframe> — the only thing it changes is the embed domain: youtube-nocookie.com (used on the website) becomes plain youtube.com for the feed. See below for why that distinction turned out to matter, and why it’s not a full fix either way.

Why nocookie on the website specifically: at the time, the youtube-nocookie.com embed had a more minimal player UI than the regular youtube.com embed. Worth re-checking whether that’s still true — YouTube changes embed UI periodically.

What the research found (2026-08-14, medium confidence overall)

Full agent report is not reproduced verbatim here — this is the distilled version.

Do feed readers render or strip iframes? It’s genuinely per-reader, not standardized:

Does youtube.com vs youtube-nocookie.com actually matter? Yes, this is real, not folklore — but it cuts both directions. Some sanitizers (Miniflux) explicitly prefer the nocookie domain; others (a common DOMPurify pattern seen in the wild, FreshRSS’s manual-allow requirement) are written against the plain youtube.com domain specifically and would silently drop a nocookie embed unless someone extended the allowlist. There is no universal answer — it depends entirely on each reader’s own hand-written allowlist.

What’s the standard fallback? A thumbnail image linking out to the video, not an iframe at all. This is what feed-generation tools (e.g. WP RSS Aggregator) commonly do by default, and the W3C Feed Validator flags <iframe> inside content:encoded as a warning specifically because of inconsistent reader support. A YouTube thumbnail needs no API key: https://i.ytimg.com/vi/{id}/hqdefault.jpg, wrapped in a link to https://www.youtube.com/watch?v={id}.

Correction (2026-08-14): the fallback pattern already exists — in main-static.html

Ran /feeds/main and /feeds/main-static through Granary’s live Atom conversion to check the mudfun videos directly. As expected, /feeds/main (sourced from the live mudfun.md) carries the raw <iframe> straight through into the Atom <content> untouched — confirms the diagnosis above empirically, not just theoretically.

But /feeds/main-static — the old hand-maintained static file, restored as main-static.html — does not have the raw iframe for either mudfun video. It already has exactly the thumbnail+link fallback the research above recommends as the standard practice:

<a href="https://www.youtube.com/watch?v=4jV1f4xQNO4" style="display: inline-block; margin: 1rem 0;">
  <img src="https://img.youtube.com/vi/4jV1f4xQNO4/maxresdefault.jpg" alt="Cleaning mud off the bicycle - Watch on YouTube" style="width: auto; height: 200px; display: block; aspect-ratio: 9 / 16; object-fit: cover; border-radius: 4px;" loading="lazy">
</a>
<p style="margin-top: 0.5rem; font-size: 0.85em; color: rgba(255,255,255,0.6); font-style: italic;">(<a href="https://www.youtube.com/watch?v=4jV1f4xQNO4">Watch on YouTube</a>)</p>

So this was already solved by hand at some point while building the static prototype — it just never got backported into mudfun.md itself (still raw iframes there) or generalized into YouTube.astro/rss.xml.js. This markup is a ready-made template for whatever the eventual fix looks like — no need to design the fallback from scratch, just decide how to wire it in (hand-edit the two posts directly vs. building it as a real fallback path in the YouTube component/feed generators).

Takeaways for whenever this gets picked back up

  1. Swapping forFeed to use youtube.com instead of youtube-nocookie.com is a narrow, unreliable workaround — it helps only the subset of readers whose allowlist happens to prefer the plain domain, and does nothing for readers that strip all iframes outright (which appears to be most of them, absent specific enrichment like Feedbin’s).
  2. A real fix means a thumbnail+link fallback in feed output specifically (website keeps the live iframe player), not a domain swap — and the exact markup for it already exists in main-static.html, see above.
  3. This has been true since the videos in mudfun.md first published (2025-10-23) — it’s a pre-existing, already-live gap, not something any recent branch work introduced. Not a blocker for anything else; a standalone task whenever there’s appetite for it.