geo-llm-optimization
Skill7 starsChanged 4 months agoMITView on GitHub
Generative Engine Optimization (GEO) - make a website citable and recommended INSIDE AI-assistant answers (ChatGPT, Claude, Perplexity, Gemini), not just ranked on Google. Use when the user wants to be found now that people ask AI directly, wants to be a source LLMs pull from, or asks about llms.txt, AI crawler access, JSON-LD/structured data for AI, or "SEO for AI". Covers letting AI crawlers in, extractable content, structured data, authority/E-E-A-T signals, and llms.txt - with concrete implementation.
In this repository9 other files it gives its agents.
- filament-plugin-developmentSkill
- lahjaSkill
- laravel-architecture-reviewSkill
- laravel-filamentSkill
- laravel-mixpostSkill
- laravel-mysql-to-postgresSkill
- laravel-upgradeSkill
- livewireSkill
- skill-authoringSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 31
Verify you are not accidentally blocking them at the CDN/WAF layer (Cloudflare "block AI bots" toggles, etc.).
- Very specific: line 56
Add a clean sitemap.xml, canonical URLs, and OG/Twitter meta.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
GEO - Generative Engine Optimization
More and more people skip Google and ask ChatGPT/Claude/Perplexity/Gemini directly. When they do, there is no results page - one synthesized answer. If your content isn't inside that answer, you don't exist for that person: no impression, no click, nothing. GEO is how you get into the answer.
SEO vs GEO: SEO optimizes so a search engine ranks your page and you win a click. GEO optimizes so an AI assistant cites or recommends you inside its answer - you win a mention. They share most fundamentals; the target differs. SEO is not dead - do the shared fundamentals well and you play both games with one codebase.
The five things GEO actually asks (in priority order)
1. Let the AI crawlers in (the #1 thing people get backwards)
AI assistants read the web through their own bots. If robots.txt blocks them, you've opted out of every AI answer on purpose. Allow them:
# robots.txt - allow AI crawlers explicitly (or just Allow: / for User-agent: *)
User-agent: GPTBot
Allow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Google-Extended # Gemini / Vertex training + grounding
Allow: /
Sitemap: https://example.com/sitemap.xmlVerify you are not accidentally blocking them at the CDN/WAF layer (Cloudflare "block AI bots" toggles, etc.).
2. Be extractable - LLMs lift claims
Models pull discrete claims out of your text. Make them easy to lift verbatim:
- Clear, descriptive headings; a direct definition near the top of each section.
- Short declarative sentences with real numbers and sources ("cut p95 from 800ms to 120ms").
- Answer the actual question plainly instead of burying it under vague prose - the page that says the thing directly gets quoted; the wall of fluff gets skipped.
3. Give machines structure (JSON-LD + semantic HTML)
Structured data tells both Google and the models exactly what the page is, who wrote it, and when. Use real Article/BlogPosting, WebSite, and Person/Organization types - and a @graph linking author to site:
{
"@context": "https://schema.org",
"@type": "BlogPosting",
"headline": "…",
"datePublished": "2026-07-28",
"author": {
"@type": "Person",
"name": "Your Name",
"jobTitle": "…",
"sameAs": ["https://www.linkedin.com/in/you", "https://github.com/you"]
},
"publisher": { "@type": "Organization", "name": "…" }
}Add a clean sitemap.xml, canonical URLs, and OG/Twitter meta. Do not fake schema (e.g. a FAQPage for Q&As that aren't really on the page) - fabricated structured data is a liability, not a GEO win.
4. Build authority / E-E-A-T signals
Models weight sources they can trust: a named author with a real bio and sameAs links to real profiles (LinkedIn, GitHub), and being talked about elsewhere. A large share of GEO is won off your own site - in what the rest of the web says about you. Get cited, guest-post, answer where your audience already asks.
5. Ship an llms.txt
An emerging convention (llmstxt.org): a plain-text/markdown file at your root that points AI crawlers at your best, cleanest content - like a sitemap.xml written for models. Generate it from your content:
# Example Site
> One-line description of what this site is and who it's for.
## Articles
- [Article title](https://example.com/blog/slug): one-line summary.
- [Another](https://example.com/blog/slug2): one-line summary.
## About
- [Author on LinkedIn](https://www.linkedin.com/in/you)
- [Author on GitHub](https://github.com/you)On Next.js, serve it from a route (e.g. app/llms.txt/route.ts) generated from your posts, with ISR so it stays current.
Implementation checklist
robots.txtallows GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended (and CDN/WAF isn't blocking them)- Every page: canonical URL, semantic headings, a plain-language answer up top
- Per-page JSON-LD (
Article/BlogPosting) + site-levelWebSite+Person/Organization@graph - Author has a real bio and
sameAsprofile links sitemap.xml+ OG/Twitter meta presentllms.txtat root, generated from real content, kept current- No fabricated structured data (no fake FAQPage)
- A plan for off-site mentions/citations (the part you can't do on your own domain)
Note on non-English
AI answers in under-served languages have far less competition. A deep, well-structured article in a language with little quality web content becomes a default source models reach for - a durable GEO moat.
Sources
llms.txt spec · OpenAI GPTBot docs · Google-Extended · schema.org BlogPosting
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

