Claude Code SEO Skill — AI skill for Claude Code

Claude Code SEO skill

Write content that ranks in Google, gets cited by ChatGPT and Perplexity, and reads like a person wrote it.

The problem

AI-written content has tells, and the obvious ones (delve, landscape, em dashes) are the least of them. The tells that survive a vocabulary cleanup are structural: "It's not X. It's Y." A one-line closer after every section. "Here's the thing." "What nobody tells you." Ideas in threes. A bold label on every bullet. A recap paragraph at the end.

Readers spot them. Google's helpful content system is built to spot template content. And AI answer engines cite specific, sourced, plainly written pages over pages that sound like they were generated.

There's a second problem that only shows up in SEO content. Guidance that says "add statistics, quotes, and citations" produces fabricated statistics, quotes, and citations when a model follows it without a constraint. This guide has the constraint.

What's in the repo

`seo-geo-content-guide.md` is the guide. It's the source of truth and the file to read.

`SKILL.md` makes the guide installable as a Claude Code skill and says how to apply it.

`CHANGELOG.md` tracks versions.

What the guide covers

Four rules sit above everything else: never invent a fact, the author's voice samples win, every sentence adds something the reader didn't have, and the portability test (if a sentence could move unchanged to another company, it's filler).

The SEO section covers content types and lengths, the last-longest-click standard from Google's NavBoost system, keyword placement that doesn't turn every H2 into a template, E-E-A-T signals, topical clusters, and the technical layer with JSON-LD schema examples.

The GEO section covers the nine Princeton GEO methods with their measured visibility boosts, what sentence structures get quoted, how each answer engine differs, and a research process for finding gaps in current AI answers.

The sourcing rule ties the two together. Statistics, quo