Source: rampstackco/claude-skills · Maintainer: community: RampStack · License: MIT · Checked: 24 October 2026
What is the seo-aeo-geo skill?
The seo-aeo-geo skill gives Claude Code, Codex and other agents a framework for getting a site cited in AI answers: AI overviews, ChatGPT search, Perplexity, Gemini and other assistants. It is for marketers, SEO consultants and developers who want an agent to audit a site for AI search and write a concrete fix plan, including an llms.txt file.
It lives in RampStack's claude-skills catalog, and an identical copy ships in rampstackco/claude-skills-seo, a 12-skill SEO subset of that catalog. The skill's description opens: "Optimize content and site structure for AI-driven search experiences including AI overviews, large language model citations, generative answer engines, and AI assistants."
The SKILL.md sets its own expectations in one line: "This skill encodes principles. AI search products evolve fast. The principles age slower than the products." It does not promise a ranking hack. It explicitly says strong AEO and GEO need strong fundamental SEO underneath.
What does it do?
The skill is model-invoked, so the agent can load it when your request fits, or you can type /seo-aeo-geo. Its description lists trigger words including AEO, GEO, AI search, AI overview, llms.txt, AI citation, ChatGPT search, Perplexity, Gemini and answer engine. It also triggers when you worry about AI eating your organic traffic.
The core is a 5-layer framework. The skill says each layer compounds:
| Layer | What the agent checks |
|---|---|
| 1. Extractable content structure | Direct 1 to 3 sentence answers, question-form headings, tables, numbered steps, inline definitions |
| 2. Citation worthiness | Original data, specific numbers, named authors, visible dates, methodology, outbound citations |
| 3. Structured data depth | Schema.org types with required and recommended properties, Person and Organization schema, FAQPage only for real FAQs |
| 4. AI-readable accessibility | /llms.txt, an explicit robots.txt decision on AI crawlers, semantic HTML, no client-only rendering for key content |
| 5. Real-world entity signals | Wikidata, consistent name and address, brand mentions, sameAs profile links, reviews |
The workflow has six steps: audit the site against the five layers, pick 10 to 20 priority queries, test current AI visibility for those queries, find the gaps, plan fixes layer by layer, then re-test quarterly. The default output is a Markdown plan saved as aeo-geo-strategy.md, with a visibility audit, a 5-layer scorecard, the priority queries, a remediation plan, a roadmap and a re-test schedule.
What's in the folder:
SKILL.md: the framework, workflow, failure patterns and output format.references/llms-txt-guide.md: what llms.txt is for, the format, two worked examples (a content site and a SaaS site), what to include,llms-full.txt, common mistakes and how to verify the file.references/extraction-friendly-patterns.md: 12 before-and-after patterns, such as question-led headings, answer-first paragraphs, tables for comparisons, self-contained sections and explicit units.
One rule stands out for anyone who has seen an agent invent numbers. Under "If required data is unavailable", the skill says: "A stated gap is a complete answer." The agent is told never to estimate or interpolate a number it could not measure.
How do you install it?
The skills.sh page for the full catalog gives this command for the skills CLI:
npx skills add https://github.com/rampstackco/claude-skills --skill seo-aeo-geo
The CLI's documented options let you add -a claude-code to target Claude Code, -g to install for every project and -y to skip prompts:
npx skills add rampstackco/claude-skills --skill seo-aeo-geo -a claude-code -g -y
The claude-skills README also documents a Claude Code plugin marketplace. The rampstack-seo plugin installs the 12-skill SEO subset, which includes this skill:
/plugin marketplace add rampstackco/claude-skills
/plugin install rampstack-seo@rampstack
The same README describes copying skill folders from a clone into ~/.claude/skills/ (every project) or .claude/skills/ (one project). Copy only skills/seo-aeo-geo if you do not want the whole catalog.
The standalone claude-skills-seo repo holds an identical copy of the skill. Its README says skill content changes are made in the full catalog first and re-synced. Its own skills.sh command is npx skills add https://github.com/rampstackco/claude-skills-seo --skill seo-aeo-geo.
In Claude Code, a personal skill lives in ~/.claude/skills/seo-aeo-geo/ and a project skill in .claude/skills/seo-aeo-geo/, per the Claude Code skills docs. Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills. To use it in the Claude apps, Anthropic's help article on using skills in Claude covers uploading a skill folder as a ZIP; the MIT license allows that.
How do you use it?
Give it a site or page, the topics you want to be cited for, and a way to see the rendered HTML. Those are the skill's three required inputs.
Example prompts (suggestions, not quotes from the repo):
Audit example.com for AI search using the 5-layer framework and save the plan to aeo-geo-strategy.md.Write an llms.txt for this Next.js docs site. Pick the key pages by hand and explain each choice.Rewrite this pricing FAQ page so AI overviews can extract clean answers. Keep the facts unchanged.Check our robots.txt and tell me which AI crawlers we block today, then recommend a policy.Our organic traffic dropped after AI overviews launched for our main queries. What should we change first?
Requirements are light. There are no packages, API keys or scripts. For step 3 of the workflow (testing visibility in AI products), the agent needs some way to run those queries, such as web access or results you paste in. Without that, expect the plan to state the gap rather than guess which sources get cited.
Tips and limits
These come from reading the skill and its repo on 24 October 2026, not from testing:
- Install counts are modest. skills.sh listed 313 installs for the copy in the full claude-skills catalog and 5 for the claude-skills-seo copy (snapshot, 24 October 2026). The repos had about 990 and 6 GitHub stars. skills.sh first saw the catalog copy on 1 May 2026, so read the skill before relying on it.
- Security audits: all three scanners on skills.sh (Gen Agent Trust Hub, Socket, Snyk) showed Pass for both copies on 24 October 2026.
- Siblings depend on where you install from. The "When NOT to use" section points to
seo-onpage,seo-technical,seo-keywordandseo-offpage. All four are in the full catalog;seo-offpageis not in the 12-skill claude-skills-seo subset. - FAQ schema is not a trick. The skill warns that "Stuffing FAQ schema on pages that have no genuine FAQs" gets penalized by search engines and AI alike.
- It is principles, not product rules. The failure patterns warn against optimizing for one AI product's quirks and against static optimization; plan to re-audit at least quarterly.
- Entity work is slow. Layer 5 includes Wikipedia and Wikidata, and the skill says a Wikipedia entry requires genuine notability, so do not force it.
The skill writes one Markdown file and runs nothing. Under the MIT License you can copy and change it as long as you keep the copyright notice.
Save it in your Kipt library
In Kipt, open Skills and paste the skill's GitHub folder link into the import box. A link to a repo with many skills, such as claude-skills-seo, shows a list so you can also pick seo-onpage or seo-technical. Your skills stay in your own private GitHub repo, and you can install all of them into Claude Code or Codex with npx skills add https://github.com/<you>/<repo>.git --all -g. Kipt is not affiliated with RampStack.
Similar skills
- Copywriting skill: writes the page copy that this skill then structures for extraction.
- Web performance skill: speed and Core Web Vitals work that sits under any search strategy.
- Web design guidelines skill: Vercel's UI review rules, including accessibility checks.
- Best Claude Code skills: a wider list to pair with it.