Most SEO content gets written twice. You draft it, run it through Yoast, then spend an hour chasing orange and red traffic lights. That is backwards. So we built an AI SEO content writer that applies every rule before a word hits the page. Here is the full story: every snag we hit, what caused each one, and what fixed it.
What is an AI SEO content writer?
An AI SEO content writer is a structured skill, not just a prompt. A prompt tells an AI what to write. A skill tells it how to write, with keyphrase distribution, readability rules, heading structure, and traffic light checks baked in from the start.
We built ours to hit every Yoast and Rank Math content check in one pass. It opens with automatic keyword research, runs a context-aware questionnaire, produces a full SEO brief and outline for your approval, and then writes the complete post on request. A compliance checklist appears at the end so you know exactly where the content stands before it goes near your CMS.
How we built the automated SEO content workflow
We started by listing every check Yoast and Rank Math run: keyphrase in the H1, within the first 100 words, in 30 to 75% of subheadings, and in the conclusion. Sentences under 20 words. Passive voice below 10%. Transition words in more than 30% of sentences. Each rule became an instruction.
We then split those rules into four reference files: keyphrase rules, meta field specs, readability rules, and the compliance checklist. Each file loads only when it is needed. If Yoast changes a threshold, you update one file and nothing else changes.
Before the questionnaire runs, the skill searches the topic automatically and returns long-tail variants and LSI terms for your approval. Throughout, the goal is copy that reads like a knowledgeable person wrote it.
Every snag we hit and how we fixed it
Getting the automated SEO content workflow to install correctly and produce clean output took considerably longer than building it. Here is what went wrong, grouped by type.
Packaging, installation and questionnaire snags
The references folder would not install. After packaging the skill, only the main file appeared on install. The cause was that our zip file never declared the references folder as an explicit directory entry. As a result, the installer had no instruction to create it. We repackaged with explicit entries in the right order and it installed cleanly.
SKILL.md would not update on reinstall. Every repackage installed the reference files correctly, but the main skill file consistently came through as an older cached version. Consequently, we resolved this by writing the zip using in-memory content and by using a new skill name each time to force a clean install.
The questionnaire confused users. A button labelled ‘Something else, I will type it’ gave users no way to type anything after clicking it. The tool already renders a free-text field below the buttons automatically, so the button was redundant. Therefore, we removed it.
The questionnaire was not context-aware. Generic audience options appeared regardless of topic. So we updated the audience, tone, and CTA questions to read the topic prompt and generate contextually relevant options, with sensible defaults as fallback.
Content structure, formatting and output snags
Subheadings came out in title case. The AI SEO content writer had no capitalisation rule, so every heading rendered in title case. As a result, we added a clear rule: all headings use sentence case, first word and proper nouns only, no exceptions including H1.
Partition lines broke CMS paste. The original output used decorative divider lines between sections. These caused formatting noise and broken blocks when pasted into WordPress. So we removed them entirely. The skill now outputs clean markdown headings.
Em dashes kept appearing. The readability rules banned em dashes from body copy, but the Word document build script applied its own rules and em dashes crept back in. Consequently, we upgraded to zero tolerance: no em dashes anywhere in chat copy, Word output, build script strings, or any message to the user while the skill is running.
H3s were fragmenting the page. Early versions used one H3 per snag entry, which created a page of thin labels and short paragraphs. Instead, we now use thematic H3 groups, each covering 3 to 4 related entries as bold-lead paragraphs beneath a single descriptive heading. This solves Yoast’s subheading distribution check without fragmenting the reading experience.
SEO compliance, scoring and distribution snags
The keyphrase was not at the start of the title tag. Yoast flags this as an improvement when the keyphrase appears but not in the first three words. Accordingly, we tightened the rule: the AI SEO content writer must place the keyphrase as close to the beginning of the title as possible.
The meta description range was wrong. We had targeted 120 to 155 characters. In practice, Yoast gives orange traffic lights outside 120 to 140 characters. We updated the target and added a character count check to the compliance checklist.
The compliance checklist graded things green that Yoast failed. Internal links, external links, and images were marked as passes despite not being present in the copy. Therefore, we rebuilt the checklist to mark them as manual items with a clear publisher action block.
The skill used invented terminology. We had been calling Yoast’s scoring system ‘the green zone’ throughout the skill files. Yoast calls it a traffic light system: green, orange, and red. So we removed the invented phrase from every file and replaced it with accurate language.
Our SEO services apply the same standards we built into this skill across every client brief we produce at Oriel.
What the AI SEO content writer produces today
The skill opens with a topic prompt. It searches the topic automatically and returns long-tail variants and LSI terms for your approval. A short questionnaire follows with context-aware options. From those answers it produces a full SEO brief: title tag, meta description, slug, OG title, primary keyphrase, synonyms for the Yoast synonym field, long-tail variants, and LSI terms. You approve the outline before any copy is written.
On confirmation, the skill asks whether you want links included. If yes, it searches your site for internal candidates and the web for authoritative sources for each named reference. It writes the complete post, inserts real markdown links, and appends a compliance checklist covering every check.
Why this matters for your content workflow
Good SEO copy is rule-based. Remembering every rule on every brief is slow and inconsistent. This AI writing skill removes that load entirely. Because rules live in separate reference files, updating one does not affect the others. For marketing managers and small business owners who publish regularly, that consistency compounds fast.
This is not a one-off prompt. Install it once and it runs on every brief. See how we apply these standards in practice on our marketing page. Download the skill below and run it on your next post.
Download the AI SEO content writer skill by completing the form below.
At Oriel, we connect business tools and cut the manual work that slows teams down. If you would like to explore what this could look like for your team, get in touch.