**Dev.

I noticed something weird while pulling articles from Dev.to's API this morning — the new AI disclosure labels are getting stripped out entirely in RSS feeds and third-party readers. The embed tag shows fine in the browser, but the moment it hits a feed parser, the whole disclosure section just... disappears.

Here's what I'm seeing in the raw RSS output:

<description>
  Excited to see how this feature rolls out and evolves over time
  {% embed https://dev.to/devteam/introducing-ai-disclosure-on-dev-tools-for-nuance-clarity-and-better-feeds-34mk minimal %}
</description>

The {% embed %} tag isn't being processed at all — it's being passed through as literal text. Most feed readers either ignore it or display it as ugly raw markup. I tried a few different parsers (rss-parser, feedparser, even a manual curl to the RSS endpoint) and they all behave the same way.

My suspicion is that Dev.to is rendering these embeds server-side for the web version but not applying the same processing pipeline to RSS output. That's a pretty common oversight when rolling out new content features, but it's worth flagging because it defeats the whole purpose of the disclosure — if readers never actually see the label, it doesn't help with transparency.

I did a quick test with a few other embed types (Twitter oEmbed, YouTube, etc.) and they seem to render fine in feeds, so it's specifically the {% embed %} liquid-style tag that's the problem. Either they need to resolve those embeds before generating the RSS XML, or provide a plain-text fallback that includes the disclosure info directly.

Has anyone else run into this? Or better yet, found a workaround? I'd rather not have to manually scrape the article content and re-parse it just to get disclosure status. Also curious if this is intentional — maybe they're treating RSS as a degraded experience on purpose?

Update — Found a partial workaround

For anyone else hitting this, the article's HTML endpoint does include the disclosure in a <div class="disclosure"> wrapper, so you can at least detect whether a given post has been labeled as AI-assisted. Not ideal, but it's something if you're building tooling around this.