Case study · SEO Engine

The SEO Engine: an SEO content system built in Claude Code

A full SEO blog post used to take a full day by hand. Now it takes 15 to 20 minutes, from research to paste-ready CMS metadata.

See it in action Book a call with Christine

Christine Mendoza, FounderBuilt for a US marketing company's in-house blog and SEO team, who create content for the company's clients · Design and build

The problem

A good SEO blog post isn't one job. It's several: picking a topic that hasn't been covered, researching it, outlining, writing, checking keywords, then preparing everything the CMS needs. Done by hand, one post took the team a full day.

Research comes first

Current stats and sources have to be found for each audience before a word gets written.

Topics are easy to repeat

Without a record of what's been published for each audience, ideas start to overlap.

SEO checks get rushed

Keyword placement in the title, the opening paragraph and the headings is easy to miss at the end of a long day.

One-shot AI drafts need rework

A single prompt gives a wall of text that someone still has to restructure and check.

What I built

The SEO Engine runs inside VS Code with Claude Code. The team types "start" and answers two questions: who the post is for, and which topic.

Topic memoryBefore suggesting ideas, it checks what's already been published for that audience, so suggestions stay fresh.
A person approves the outlineAfter research, it stops and presents H1 title options, a section plan and highlights. Nothing gets drafted until someone says yes.
Code checks the SEOPython scripts grade the draft instead of the AI grading itself, and flagged issues get a targeted fix and a re-check.
Publish-ready outputA post summary plus meta title, meta description, focus keyword, slug and tags, ready to paste into the CMS.

Built with: Claude Code in VS Code, with Python scripts for topic tracking, saving files, SEO analysis and reporting.

How it works

Step by step

Research. Live web searches for current stats and best practices, saved as research notes.
Outline. H1 title options, a section-by-section plan and highlights, presented for approval.
Draft. The full post, written to its own file.
Review. A self-check against the style rules: active voice, no filler, short paragraphs, stats cited or flagged.
SEO and formatting. The scripts run, issues get fixed, and the check repeats until the keyword is in the title, the opening paragraph and an H2.
Final delivery. Paste-ready CMS metadata, and the topic is logged so it isn't suggested again.

Results

What changed

1 day → 15–20 minper full SEO post, from research to paste-ready CMS metadata.

6 phaseseach saved to its own file, so the team can check any step.

Human-approvedNothing is drafted until a person approves the outline.

Honest about gapsWhen something still needs a person, like a placeholder meta description, it says so instead of passing it off as finished.

See it in action

Inside the build

▶ Watch the 1-minute demo

Start screen: pick an audience, then a topic

Phase 2: outline and H1 options waiting for approval

Phase 5: the SEO check flags the keyword and the engine fixes it

Phase 6: post summary and paste-ready CMS metadata

The speed comes from the structure, not the model. Which parts of your team's content process really need a person, and which are steps waiting to become a system?Christine Mendoza, Founder

Book a call with Christine

Tell me what your team keeps doing by hand.

A 30-minute call with me, not a sales team. I'll ask why your business exists, what's getting in the way and what you want, and tell you honestly whether a system would help.

Book a call with Christine Message me on LinkedIn