Two distinct optimization targets:
- SEO — ranking in traditional search engines (Google, Bing)
- GEO — being cited by AI search engines (ChatGPT, Perplexity, Gemini, Copilot)
Key insight: AI search engines do not rank pages — they cite sources. Being cited is the new ranking #1.
Documentation
The SEO and GEO principles in this skill are stable, but the llms.txt standard is actively evolving.
Before generating or advising on llms.txt files, use WebFetch to read the current spec:
WebFetch("https://llmstxt.org")
Other reference sources (use WebFetch when you need specifics):
- GEO research: https://arxiv.org/abs/2311.09735 (Princeton study)
- Schema reference: https://schema.org (use
references/schema-patterns.mdfor common templates) - Google guidelines: https://developers.google.com/search/docs
Core Principles
Apply these whenever writing or reviewing public-facing content or markup, regardless of framework.
GEO: The 9 Princeton Methods
From Princeton research on what makes content get cited by AI engines:
| Method | Visibility boost | How to apply |
|---|---|---|
| Cite authoritative sources | +40% | Link to studies, official docs, standards |
| Include specific statistics | +37% | Use real numbers, not vague claims |
| Add expert quotes | +30% | Attribute statements to named sources |
| Authoritative tone | +25% | Write with confidence and precision |
| Readable and clear | +20% | Short paragraphs, plain language |
| Use technical terminology | +18% | Domain-specific terms signal expertise |
| Vocabulary diversity | +15% | Avoid repetition of the same words |
| Fluency optimization | +15-30% | Natural flow, no awkward phrasing |
| Keyword stuffing | -10% | Never — hurts both SEO and GEO |
Best combination: fluency + statistics = maximum boost.
Content Structure for GEO
- Answer-first: put the direct answer at the top, then elaborate
- Clear heading hierarchy (H1 > H2 > H3)
- Use bullet points, numbered lists, and comparison tables
- FAQ sections significantly increase AI citation rate
- Short paragraphs (2-3 sentences)
Traditional SEO Fundamentals
- One H1 per page, containing the primary keyword
- Meta title: 50-60 characters, keyword near the front
- Meta description: 150-160 characters, include keyword and a call to action
- Every image needs descriptive alt text
- Canonical URL on every page
- Sitemap.xml and robots.txt present and correct
Platform-Specific Behaviour
Each AI engine has different citation preferences. See references/geo-principles.md for full detail.
ChatGPT: favours branded domains; content updated within 30 days cited 3.2x more Perplexity: allow PerplexityBot in robots.txt; prioritises PDF documents and FAQ schema Google AI Overview: E-E-A-T signals, structured data, topical authority through content clusters Microsoft Copilot: requires Bing indexing; page speed under 2 seconds; LinkedIn/GitHub mentions help Anthropic AI: uses Brave Search indexing (not Google); prioritises factual density and structural clarity
llms.txt
llms.txt is to AI engines what robots.txt is to crawlers — a file at your site root that tells AI systems how to understand your content. Always add it to any public-facing site.
Place at https://yourdomain.com/llms.txt. Format is Markdown:
# Site or Product Name
> One-paragraph summary of what the site is and who it is for.
Optional additional context (key facts, caveats, important notes).
## Docs
- [Page title](https://yourdomain.com/page.md): What this page covers
## API
- [API reference](https://yourdomain.com/api.md): Endpoint overview
## Optional
- [Changelog](https://yourdomain.com/changelog.md): Recent changes
Key rules:
- H1 = site/product name (required)
- Blockquote = brief summary of what the site is (required for AI comprehension)
- H2 sections = curated links to your most important pages with descriptions
- Link to
.mdversions of pages where possible — AI engines prefer clean Markdown over HTML - The
## Optionalsection signals lower-priority content that can be skipped in short-context situations - Keep it concise — this is a navigation aid, not a full content dump
Structured Data
Always add JSON-LD schema. See references/schema-patterns.md for templates.
Priority schema types:
- FAQPage — +40% AI visibility; add to any page with questions and answers
- Article — for blog posts and editorial content
- Organization — on the homepage and about page
- Product — for product pages
- BreadcrumbList — for navigation context
Checklist
Apply when reviewing any public-facing page:
-
llms.txtpresent at site root with H1, summary blockquote, and key page links - Unique, keyword-containing H1
- Meta title and description present and correct length
- Open Graph tags (og:title, og:description, og:image, og:url)
- JSON-LD schema appropriate to page type
- FAQPage schema if page contains Q&A content
- robots.txt allows relevant bots (including PerplexityBot, GPTBot, ClaudeBot, anthropic-ai)
- Sitemap includes this page
- Images have descriptive alt text
- Answer-first content structure
- At least one statistic or data point per key claim
- Mobile-friendly layout
- Page loads in under 3 seconds