The short version
- Make the standard a script that exits with an error. A checklist gets skipped in exactly the week you needed it.
- Give the script only what it can be certain about. A validator that fails good pages for a bad reason teaches everybody where the skip flag is.
- One command re-checks the whole library, so a new post costs nothing to validate and a rule change applies backwards across everything already published.
A publish gate is an automated check that runs before content goes live and refuses to ship anything below the standard. We put 44 content pages for a construction procurement SaaS behind one, after most of those pages had already been written.
What is a publish gate?
A script that runs before a release and exits with an error if a page misses the bar. Not a checklist, not a review meeting. A command run by the same person who was about to publish, on their own machine, before anything has a URL.
What our own gate checks, and who does the checking:
| Check | Bar | Enforced by | What it prevents |
|---|---|---|---|
| Em-dashes | zero | Script, hard fail | The loudest tell in our house style |
| Rendered word count | 900 minimum | Script, hard fail | Thin pages that rank for nothing |
| FAQ entries | 5 minimum | Script, hard fail | A page with no FAQ schema |
| Data table | at least one | Script, hard fail | Nothing an answer engine can lift |
| Internal links | at least one | Script, hard fail | Orphan pages nothing points at |
| Cover image | unused by any other post | Script, hard fail | Two essays wearing the same picture |
| En-dashes | zero | Human, final read | A rule the script does not cover yet |
| Every number sourced | one row in the fact bank | Human, fact-check pass | A claim nobody can defend on a call |
| Client-identifying detail | zero | Human, final read | A broken anonymity promise |
Our house style bans both characters and the script tests for one of them.
Why did we add one after the fact?
Because writing a rule down does not enforce it. We had a documented house style and it was restated in every brief. An audit of our own two hundred essays then turned up the banned character in post after post, against a house rule of zero. Nobody had decided to ignore it. The rule had nothing behind it.
One writer holds a standard in their head and mostly applies it. Add a second writer and a generation step and it becomes several standards, each defensible on its own terms. The script collapses them back into one.
Ours only gates posts carrying the current frontmatter kit and leaves the older essays alone. A single opinion piece is exempt from the depth checks because that one is measured on links rather than rankings. The exemption list sits in the script itself.
What belongs in the gate, and what does not?
Anything mechanical. Characters, counts, thresholds, the presence of a required element. If a person can be wrong about it and a script cannot, the script owns it.
Judgment stays with a person: whether a claim is true, and whether a story identifies the client who trusted us not to write it. No validator reads a paragraph and tells you it is boring.
Getting that split wrong is how gates die. Put a judgment call in the script and it starts failing pages that are fine. Somebody adds a flag to skip it for the release that has to go out today, and the flag stays in the command everybody copies afterwards.
What does a gate cost to run?
One file and one command. Ours re-checks every post in the library on each run, so validating a new post costs nothing, and changing a rule is one edit that applies backwards across everything already published.
Six of the nine checks are a script. The other three are a person reading the page.
Why is a retraction so expensive?
The edit itself is trivial. The email explaining the edit is not. Somebody re-reads everything else that shipped that month, because if one page got through, the client is right to ask what else did.
The two findings that put us here are enforcing style rules with a script and why volume without structure stays invisible.
Frequently asked questions
What is a publish gate in content operations?
A script that runs before content goes live and blocks the release if a page misses the standard. It checks mechanical rules such as banned characters, minimum rendered word count, required elements and internal links. Unlike a checklist, it cannot be quietly skipped by the person under the most pressure to skip it.
What should an automated content validator check?
Anything a script can be certain about: banned characters, minimum rendered word count, a minimum number of FAQ entries, the presence of a data table, at least one internal link, and a cover image no other post is already using. Ours hard-fails on every one of those.
Why do written style rules get broken anyway?
Because stating a rule does not enforce it. An audit of our own two hundred essays against a house rule banning one character found that character in post after post, with the rule sitting in every brief the whole time. A standard degrades with every writer and every generation step you add to it.
Should a content gate check quality as well as mechanics?
No. Encode only what a script can be objectively right about. Judgment calls, such as whether a claim is true or whether a story identifies a client, belong to a human review pass. A gate that fails good pages gets a skip flag added for one urgent release, and the flag stays in the command everybody copies.
How much does a publish gate cost to build and run?
One file and one command. Ours re-validates the entire library on every run, so adding a post costs nothing and changing a rule is a single edit that applies retroactively to everything already published.
What is the real cost of publishing a mistake?
Far more than the correction. The client is right to ask what else shipped that month, so everything published in it gets re-read.
Sources
- Chua Network delivery data across 8 client accounts (internal fact bank)
- Chua Network engagement records, anonymized (internal experience bank)
