Search is splitting in two. There’s the Google you’ve been optimizing for over the last fifteen years, and there’s the layer that’s quietly taking control: ChatGPT, Claude, Perplexity, Gemini, and Google’s own AI Overviews answering questions before anyone clicks a blue link. geo-seo-claude is a Claude Code skill built for that second layer — and it’s sitting at 8.2K stars.
The premise is in the name: GEO-first, SEO-supported. GEO stands for Generative Engine Optimization, and the bet behind this skill is straightforward — optimize for where the traffic is going, not where it was. The fundamentals of traditional SEO still matter, but now they’re just the opening move. The central question the tool tries to answer is whether an AI engine will cite your site when composing an answer.
Internally it’s not a single script but an orchestrated system: 11 specialized sub-skills coordinated by 5 subagents running in parallel. The breakdown covers the entire surface of AI discoverability:
- geo-citability scores how ready your content is to be cited by an AI engine.
- geo-crawlers analyzes whether AI crawlers (GPTBot, ClaudeBot, PerplexityBot and company) can actually reach your pages.
- geo-llmstxt analyzes and generates the emerging
llms.txtstandard — the equivalent of robots.txt for language models. - geo-brand-mentions tracks your brand’s presence on the platforms that AI engines pull information from.
- geo-schema handles structured data so machines can parse your pages correctly.
- geo-platform-optimizer fine-tunes for specific AI-powered search platforms.
- geo-technical and geo-content cover the basics of traditional technical SEO and E-E-A-T.
- geo-audit runs the full orchestration and scoring.
- geo-report and geo-report-pdf convert results into a markdown report ready for the client and a professional PDF with charts.
That last pair is the clue about who this is for. A PDF report with charts isn’t a feature you build to tweak your personal blog — it’s made to hand off to a client. It’s a skill aimed at agencies and consultants who need to show GEO work as a deliverable, which also explains the 1.3K forks: people are adapting it for their own offerings.
Installation is a single command from your terminal (Mac/Linux, or Git Bash on Windows). Python dependencies go into a dedicated virtual environment at ~/.claude/skills/geo/.venv/, so your system Python stays intact and a clean uninstall takes everything with it. Once installed, you point it at a URL and let the audit run.
A couple of honest caveats. GEO is a young discipline — the llms.txt standard it relies on is still emerging and not all crawlers respect it, and the “citability scoring” is an interpretation of how AI engines choose their sources, not a published ranking formula that anyone’s confirmed. Take the scores as a structured heuristic, not gospel. And the repo author points toward a paid community for monetization advice; the skill itself is open-source and free, which is the part that deserves your attention.
The bigger picture: GEO is moving from buzzword to quick workflow, and the tooling is arriving to match. Whether or not this particular skill becomes your go-to tool, the shift it’s built on — optimizing for the engine that answers instead of the one that lists — is the direction the whole field is heading.