An article goes out under a client's name with an invented statistic and an internal link to a page that doesn't exist. The client's marketing manager finds both in ten minutes. The damage costs more than the article was ever worth, because every future deliverable now gets read with suspicion.
Between draft and client, you need a gate: a fixed set of checks, run every time, each with a pass or a fail.
A gate is a check anyone can run
An editor's taste doesn't scale to the volume a content retainer demands. It varies by editor, by article, and by how tired the editor is. A gate doesn't vary. It's binary, it's written down, and it gives the same answer whoever runs it. "Does every number have a source the reader could open" is a gate. "Is this good" is an opinion.
Gates don't replace editorial judgement. They protect it. Once a checklist handles the mechanical checks, the editor's attention goes to the argument, the angle and the client's voice. Only a human can judge those.
Three gates cover what actually goes wrong: facts, intent and structure, and links and metadata.
Gate one: every claim survives a source check
AI drafts invent statistics, studies and quotes. Tired human writers do the same by half-remembering. So the rule is strict: a number appears in the article only with a named source the reader could click. A claim with no source either becomes qualitative or comes out.
If the draft mentions a study by name, find it before you publish. Citing research that doesn't exist is the fastest way to embarrass a client. Generators that attach citations make the check quicker, but they never make it unnecessary.
Client facts get the same treatment. Check prices, service areas, guarantees, product names and anything phrased as a promise against the client's actual business. The client knows their offer better than you do and will catch what you miss. To them, an invented claim under their name reads as carelessness with their reputation.
Gate two: the page matches the search it targets
Open the target query and look at what ranks. If the query wants a how-to and your draft is a sales page, no polish fixes it. This is the most common way a well-written article fails. Answer the searcher's question within the first screen, then earn the rest of their time. Keyword strategy per client covers how the target got chosen. This gate checks that the draft actually serves it.
Structure is the measurable half. You want one H1, headings that follow a hierarchy, paragraphs short enough to scan, and enough substance for the page type. A deterministic content scorer is the right tool for this. It measures form the same way every time, so the same page always gets the same score and two scores are comparable.
What a scorer can't do is tell you whether the claims are true or the advice is any good. It measures form, not correctness. That's why gate one stays human.
Gate three: links, images and metadata resolve
Internal links must point at pages that exist on the client's site. Generators asked to add internal links will invent plausible-looking paths unless you give them real targets. A human skimming the draft won't spot the difference. Click every one.
External links go to primary sources and resolve without errors. They never point at the client's competitors.
Images need alt text and a proper licence. The title and meta description should read correctly at SERP length. They should also be unique against the client's other pages, because two pages with the same title compete with each other before they compete with anyone else. The free SERP snippet preview shows the title and description as Google would display them, which settles the length question in seconds.
Fail closed, and keep the record
When a gate fails, the article goes back a stage. It never moves forward with a note attached, because a note is just a way of publishing a known defect. The first month, this feels slow. Then a gate catches a fabricated statistic on a page the client planned to run ads to, and it stops feeling slow.
Also keep a simple record of which gate failed on each article. The pattern is the payoff. If most articles fail the intent gate, your briefs are weak. If most fail the link gate, your drafting setup isn't getting real link targets. A gate that fails often is showing you where the pipeline leaks, and a checklist is a cheaper way to learn that than a client.
Frequently asked questions
What should a content SEO checklist include?
Three gates cover it. The first checks facts: every number has a source, and client details match their real business. The second checks intent and structure. The third checks that links, images, titles and meta descriptions all work.
Can a tool run the whole checklist for me?
Only part of it. A content scorer measures structure the same way every time, and a SERP snippet preview settles title and description length. Whether the claims are true and the advice is good stays a human check.
What should I do when an article fails a gate?
Send it back a stage. Don't ship it with a note attached, since that publishes a known defect. Log which gate failed so you can see whether the briefs, the drafting setup or the review step needs fixing.
How do I check that an article matches search intent?
Open the target query and look at what ranks. If those pages are how-tos, your draft shouldn't be a sales page. It should also answer the searcher's question within the first screen.
What to take away
- A gate is a binary, written-down check that gives the same answer whoever runs it, which is what lets quality survive volume.
- Every number needs a source the reader could open, and client facts get verified against the client's real business before anything ships.
- A scorer measures form and does it consistently, so use it for structure, and keep the truth of claims as a human check.
- Fail closed: an article that fails a gate goes back a stage, and the record of which gates fail tells you where the pipeline leaks.
Next
Once an article clears the gates, it still has to reach the client's site intact. Publishing to a client's CMS covers access, clean publishing and proving the page is live.