Your call-to-action can be present and still dead-end. Audit for clickable, not just for the URL.

I run an automated content pipeline. It checks, among other things, that every published video description contains a call-to-action — a link to the offer. For weeks that check was green. Conversions were still zero.

The check was lying to me, and the reason is worth thirty seconds of your attention if you automate any content.

The bug

My audit did roughly this:

// ❌ looks reasonable, is wrong
const hasCTA = description.includes("penloomstudio.com");

Every description that mentioned the store passed. But a few of them said things like:

Get the pack → [penloomstudio.com/fix](https://penloomstudio.com/fix)

That is a bare domain in prose. YouTube — like most platforms — does not linkify a bare URL in a description. No http://, no auto-link. So the viewer sees grey, unclickable text. The CTA was present and useless at the same time. My audit only knew how to check the first thing.

The fix

Audit for a clickable link, not for the domain string:

// ✅ requires a real scheme so the platform will linkify it
const hasClickableCTA = /https?://S*penloomstudio.com/i.test(description);

One character of intent — the https?:// — is the entire difference between “we mentioned the offer” and “a human can reach the offer.”

The part that actually generalizes

The same failure showed up in a second channel weeks later: cross-posting articles where I’d written [penloomstudio.com/x](https://penloomstudio.com/x) in the body. Same dead-end, same cause. So the durable fix wasn’t a smarter audit — it was a single chokepoint every outbound post flows through, that upgrades bare domains to real links before anything ships:

// idempotent: leaves existing markdown links and full URLs untouched,
// only promotes a bare domain to a clickable one
// (swap example.com for your own domain)
function linkify(body) {
  return body.replace(
    /[[^]]*]([^)]*)|https?://S+|(?<![w@/.])(example.com/[^s)*`]]+)/g,
    (m, bare) => (bare ? `[${bare}](https://${bare})` : m)
  );
}

The alternation matters: it consumes any already-linked URL whole first, so the bare-domain branch can never partial-match text that’s already inside a link. (That was the second bug — the naive version nested and truncated links that were already fine.)

The lesson, one line

An automated check that tests for the presence of a thing, when what you needed was the usability of that thing, is worse than no check — because it’s green.

“Is the link there?” and “can a human click it?” are different questions. If your funnel is quietly converting nobody, go look at whether your own tooling is answering the easier one.

I write these up as I hit them, building a small studio’s content system in public. Notes and the tools live at penloomstudio.com.