The undocumented Atom endpoint that outlasted every official syndication feature YouTube ever built.

YouTube launched in 2005. RSS was already the dominant protocol for following websites without visiting them. The obvious move — surfacing a feed per channel — was made, quietly, and then never spoken of again.

Every YouTube channel exposes an Atom feed at a predictable URL, constructed from the channel's unique identifier: https://www.youtube.com/feeds/videos.xml?channel_id=CHANNEL_ID. The channel ID appears in the channel's URL and in the page source. No authentication is required. The feed is valid XML, returns recent uploads in reverse-chronological order, and includes the video title, publication timestamp, and a link to the watch page. YouTube has never documented this endpoint in any public API reference, and the company has not announced it, promoted it, or, notably, removed it.

Adult leaning over a laptop showing a Feedly unread list several hundred items long, desk lamp casting warm light on the screen
A laptop, a newspaper and a pair of headphones on one desk. Three delivery routes for the same reporting, and only one of them is a feed.Photo: Beyzanur K. / Pexels
Feed reader export dialog on a monitor screen showing OPML file download prompt, adult hand on mouse
Moving between readers is a file transfer, not a migration. OPML carries the URLs; read state, starred items and folders mostly stay behind.Photo: SHVETS production / Pexels

What it does not include is also worth noting. There is no enclosure element carrying a media file — the feed points to YouTube's walled player, not to a downloadable stream. Chapters, descriptions, and thumbnail URLs sometimes appear as extensions but inconsistently. The feed is a pointer, not a delivery mechanism.

Feed-reader users have treated the endpoint as a first-class subscription method regardless. In Feedly, Inoreader, NewsBlur, and self-hosted readers such as FreshRSS and Miniflux, pasting the constructed URL produces a working subscription. The practical effect is a channel view stripped of recommendations, suggested videos, comment sections, and autoplay — a chronological list of uploads and nothing else. For anyone who has decided that the algorithmic feed reader is a different kind of problem than the one it solves, a plain Atom subscription is the cleaner tool.

From the record

How to find it

  • Channel IDthe UC… string in a channel's URL or page source; substitute into https://www.youtube.com/feeds/videos.xml?channel_id= to get a working Atom feed
  • No login requiredthe endpoint is publicly accessible without a Google account
  • No autodiscoveryYouTube's HTML does not include a <link rel="alternate"> tag pointing to the feed; the URL must be constructed manually

What the feed does and doesn't carry

  • Carriesvideo title, upload date, link to watch page, sometimes thumbnail and description extensions
  • Does not carryan enclosure element with a media file; the feed is a pointer to YouTube's player, not a download

The persistence of this endpoint is itself a data point. YouTube has deprecated and removed official RSS features before — the activity feeds, the friends system, and various v2 Data API endpoints were all retired during the 2010s as Google progressively tightened control over what left the platform in structured form. The per-channel Atom feed survived each of those contractions. Whether that reflects deliberate policy, institutional inertia, or simply the low cost of leaving a read-only XML endpoint running is not something YouTube has explained.

The robots.txt convention does not govern this — the feed is served openly by YouTube's own infrastructure. What constrains it is the absence of a media file and the dependence on a channel ID that casual users have no obvious way to find. The feed works; it is just not advertised, not linked from channel pages, and not included in any autodiscovery tag on youtube.com. It exists in the manner of many useful things on the open web: reliably, and in silence.