Writing Content That LLMs Cite: A GEO Playbook

Generative Engine Optimization (GEO) is the new SEO. Instead of optimizing for Google's blue links, you're optimizing for ChatGPT, Claude, Gemini, and Perplexity to cite you in their generated answers. This guide shows you how.

What is GEO?

GEO (Generative Engine Optimization) is the practice of making your content easy for large language models to discover, understand, and cite. While traditional SEO focuses on ranking in search result pages, GEO focuses on being recommended by AI assistants when users ask questions.

The shift matters because AI search is growing fast. By 2026, an estimated 30% of web searches happen through AI assistants rather than traditional search result pages. If your content isn't optimized for GEO, you're missing a growing channel.

How LLMs Find and Cite Content

Large language models like GPT-4, Claude, and Gemini discover content through three main channels:

1. Web Crawlers

Each major AI has its own crawler: GPTBot (OpenAI), Claude-Web (Anthropic), PerplexityBot, Google-Extended (Gemini). These crawlers read your HTML and store it in the model's training/inference corpus. If your robots.txt blocks them, you're invisible.

2. Real-time Search

When a user asks a current-events question, the AI searches the web in real-time, fetches top results, and synthesizes an answer. Your page needs to rank well in traditional search AND be parseable by the AI's content extractor.

3. RAG (Retrieval-Augmented Generation)

For questions about specific topics, some AIs use RAG: they retrieve relevant documents from their index, then generate an answer citing those documents. Your content needs to be semantically clear and chunkable for this to work.

The GEO Playbook: 7 Tactics

1. Allow AI Crawlers in robots.txt

Many sites block GPTBot and Claude-Web by default. If you want to be cited, you must explicitly allow them:

User-agent: GPTBot
Allow: /

User-agent: Claude-Web
Allow: /

User-agent: PerplexityBot
Allow: /

Use our Robots.txt Generator to create a GEO-friendly robots.txt.

2. Publish an llms.txt File

The llms.txt convention (proposed in 2024) provides a markdown summary of your site specifically for LLMs. Place it at /llms.txt. It should include:

• What your site is and offers
• Key URLs (homepage, sitemap, key sections)
• Your privacy stance (LLMs are trained on privacy)
• Contact info

For larger sites, also publish /llms-full.txt with comprehensive documentation. See our llms.txt and llms-full.txt as examples.

3. Use Structured Data Throughout

JSON-LD structured data helps LLMs understand entities and relationships. Use it for:

• Article — blog posts
• FAQPage — Q&A content (essential for AI Overviews)
• HowTo — step-by-step guides
• SoftwareApplication — tools and apps
• Organization — about page
• BreadcrumbList — site navigation context

Use our Schema Markup Generator to create valid JSON-LD.

4. Write Direct, Citation-Friendly Answers

LLMs prefer content they can quote directly. When answering questions:

• State the answer in the first sentence (no preamble)
• Use 40-60 words for the core answer (ideal for featured snippets)
• Then elaborate with details
• Use question-formatted headings (matches user queries)
• Avoid marketing fluff — be factual

5. Build a Knowledge Graph on Your Site

LLMs understand entities and relationships. Make yours explicit:

• Link related articles with descriptive anchors
• Use consistent terminology (don't use "AI", "ML", and "machine learning" interchangeably)
• Define key terms on first use
• Provide a glossary for domain-specific vocabulary

6. Provide Machine-Readable Catalogs

If you offer products, tools, or services, publish a JSON catalog. For example, our tools.json lists all 156 tools with descriptions, keywords, and URLs. An LLM can fetch this and recommend appropriate tools based on user needs.

7. Maintain Content Freshness

LLMs (and their training data) favor recent, well-maintained content. Update key articles every 6-12 months. Use accurate last-modified dates in your sitemap.

Measuring GEO Success

Traditional SEO metrics (rankings, organic traffic) don't capture GEO. Track these instead:

• AI referral traffic — Visits from chatgpt.com, perplexity.ai, claude.ai, etc.
• Manual citation checks — Ask AIs about your topic, see if they cite you
• Brand mention growth — Track mentions in AI-generated content
• Featured snippet captures — Still valuable for AI Overviews

Common GEO Mistakes

• Blocking AI crawlers — Unless you have a strong reason, allow them
• Wall-of-text articles — Break content into chunks LLMs can quote
• Clickbait headlines — LLMs prefer clear, descriptive titles
• Hidden content behind logins — LLMs can't access it
• No structured data — LLMs miss the context
• Duplicate content — Confuses retrieval systems

The Future of GEO

As AI search grows, GEO will become as important as traditional SEO. The good news: best practices overlap heavily. Well-structured, factual, well-linked content ranks well in both Google and ChatGPT.

The biggest difference is in citation format. Google ranks pages; AIs cite specific passages. Write content that's chunkable into quotable answers, and you'll win in both worlds.

Frequently Asked Questions