In Settings, Reading, there’s a line that says “For each article in a feed, include”, with two choices: Summary or Full text. It looks like a decision from another decade, and it is one, but it decides whether any WordPress content automation built on top of a site gets real content or a teaser. Everything downstream (the model writing the social post, the video descriptions, the Showcase copy) inherits whatever that dropdown puts in the feed.
Nicholas Garofalo has a walkthrough on Make WordPress Marketing about the AI and Zapier workflows now running the project’s social accounts, the WordPress.tv to YouTube pipeline, and the Showcase. The detail I keep coming back to sits in the comments, in his reply to a reader. The AI turns out to be the smallest part of it. The triggers are almost all RSS.
Aside from the AI bits though, almost everything is triggered by RSS, which is an over 20-year-old WordPress feature.
WordPress content automation starts at the feed
Each social amplification trigger polls an RSS feed from a WordPress.org blog, a Make team, or a WordCamp site, hands the item to one shared writing sub-workflow, and files the result as a GitHub issue labeled as an amplification request. The writing step takes its instructions from a brand style guide kept in the marketing repo, which means the prompt context is versioned like code and reviewed in the open. It’s the same move as going from pasted prompts to repeatable workflows, done at project scale. By the post’s count, reach across the eleven official accounts is 2.4 million followers, and release posts are now auto-posted immediately, a step the team only took once confidence in the drafts grew.
The design decision worth stealing is the queue. Every draft lands in GitHub, where anyone can see what was generated, what was changed, and what shipped. That visibility is what made auto-posting safe enough to switch on.
The automation turned all the way up
The WordPress.tv automation launched in 2022 and closed a Trac ticket that had been open for seven years. Before that, the WordPress YouTube channel had roughly 14,000 subscribers; it now has over 117,000, and annual watch time grew from about 15,000 hours to over 75,000. The pipeline got there through error handling: retries and branches for YouTube API rate limits, failed uploads, and duplicate entries, so videos auto-publish with thumbnails and metadata instead of waiting on a person.
The Showcase pipeline is the more aggressive one. Since 2025, submissions land in a shared Google Sheet, get verified as running WordPress, are enriched with BuiltWith data, and are rated on a five-star scale by AI against content, authority, and design, with external signals like page rank mixed in. Every rating comes with a written rationale, so a reviewer can see why a site scored the way it did. Qualifying sites become GitHub issues with AI-written copy; humans still review everything through the sheet, and published entries then trigger the social workflow. What used to take weeks of coordination now takes minutes, per the post, and last year the Showcase drew over 600,000 users and more than 1.3 million page views, with organic search as the most common entry point.
Pointing this at client sites
Most sites are one setting away from being ready. I’d treat the feed as an asset and check three things before building any workflow on top of it.
- Feed depth. Summary or Full text decides whether the model writing your drafts reads the piece or guesses from a 55-word excerpt. Full text for anything feeding automation.
- Stable identity. Feed readers and Zapier’s RSS trigger key off the item GUID, and WordPress keeps it stable across edits. Migrations and permalink changes are where it breaks, and the failure mode is every workflow firing twice for the same post.
- A dedicated feed when the main one has editorial duties. If machines should read a curated stream, register one with add_feed() instead of bending the main feed:
add_action( 'init', 'bbioon_register_release_feed' );
function bbioon_register_release_feed() {
add_feed( 'releases', 'bbioon_render_release_feed' );
}
function bbioon_render_release_feed() {
header( 'Content-Type: application/rss+xml; charset=' . get_option( 'blog_charset' ), true );
load_template( get_template_directory() . '/feeds/releases.php' );
}
After adding the feed, visit Settings, Permalinks once to flush rewrite rules, or the new URL 404s until you do. I haven’t tested how the RSS triggers behave with WooCommerce product feeds. Those usually come from a plugin, and each one shapes GUIDs differently, so treat that as unverified until you watch a few items cycle through.
One caveat on the Showcase half: the rating leans on authority and page rank, so the gate now bakes in whatever those signals already reward. A genuinely good site on a fresh domain starts at a disadvantage a human reviewer might not apply. The written rationale is the counterweight: you can argue with it, but only if someone is looking.
If you want a pipeline like this behind a client store or a content site, that’s work I do: audit the feed, register a machine-facing one, wire the workflow, and keep a human approval step where it earns its place. The hard part is deciding what deserves to go in the feed.
The question I’m still sitting on is who audits the model’s taste over time. The Showcase worked because a reviewer could be argued with, and a five-star score with a rationale looks arguable too, but drift is quiet. Nobody files a ticket when the front page slowly gets more conventional, and by the time anyone notices, the examples the model learned from are mostly things it already liked.