geo-optimisation

GEO: Writing for AI Citation

Generative Engine Optimisation (GEO) ensures documentation surfaces correctly in AI-generated answers — ChatGPT, Perplexity, Google AI Overviews, and Claude. These principles apply to all public-facing docs, not just READMEs.

Crisp Definitions First

Put a one-sentence definition of the project at the very top of the README, before badges or navigation. LLMs preferentially quote top-of-page definitions when answering "what is X?" queries. The definition must be standalone — it should make sense if extracted with no surrounding context.

Atomic Sections

Each H2 section should have one clear intent, answerable as a standalone snippet. AI retrieval systems (RAG) chunk documents by heading, so a section that mixes installation with architecture reduces citation accuracy.

Rules:

• One topic per H2 — don't combine "Features" and "Configuration"
• Strict heading hierarchy: H1 > H2 > H3 without skipping levels
• Descriptive headings with topic keywords — "## TypeScript Configuration" not "## Config"
• Each section should be comprehensible without reading prior sections

Concrete Statistics

Content with concrete statistics can boost visibility in AI responses by up to 28% (Aggarwal et al., "GEO: Generative Engine Optimization", 2023). Include benchmarks, performance numbers, and measurable outcomes wherever evidence exists.

Rules:

• Every statistic must trace to actual code, a benchmark file, or a verifiable measurement
• Prefer relative comparisons ("40% faster than X") over absolute numbers when the alternative is well-known

Comparison Tables

LLMs frequently surface comparison tables when answering "X vs Y" queries. Use a descriptive H2 heading ("How It Compares"), be factually accurate about competitors, and include at least one quantitative row alongside qualitative ones.

TL;DR and Key Concepts Blocks

For long guides (200+ lines), add a TL;DR block immediately after the title. RAG systems often extract the first paragraph under a heading — make it count.