The GEO glossary.
Twelve terms, defined the way we'd want a model to quote them: one clear sentence first, nuance after.
- Generative engine
- An AI system that answers queries by generating a synthesized response from retrieved sources rather than returning a list of links. Claude, ChatGPT, Perplexity and Gemini in search mode are generative engines.
- Generative Engine Optimization (GEO)
- The practice of optimizing content to be retrieved, quoted and cited by generative engines. Coined in Aggarwal et al., KDD 2024.
- Answer Engine Optimization (AEO)
- A near-synonym of GEO, emphasizing direct-answer formats. In practice the terms are used interchangeably.
- Retrievability
- Whether a page can appear in the search results an AI assistant reads: indexed, crawlable, and ranked high enough to be in the retrieved set.
- Citability
- Whether a retrieved passage is worth quoting: self-contained, direct, evidenced. High retrievability with low citability means being read but never cited.
- Citation rate
- The share of AI answers on a defined query set in which a source is cited. The primary GEO success metric.
- Share of voice (AI)
- A brand's proportion of mentions across AI answers for a query set, relative to competitors.
- Grounding
- An AI model anchoring its answer in retrieved documents rather than its training data alone. GEO targets the grounding step.
- llms.txt
- A proposed convention: a markdown file at a site's root (/llms.txt) giving language models a concise, curated summary of the site and its key pages.
- AI crawler
- A bot that fetches web content for AI systems — ClaudeBot (Anthropic), GPTBot (OpenAI), PerplexityBot, Google-Extended. Controlled via robots.txt.
- JSON-LD
- A machine-readable format for structured data (schema.org vocabulary) embedded in pages, declaring entities: organization, services, FAQs, articles.
- Corroboration
- Third-party confirmation of a brand's claims. Generative engines weight sources that other retrieved sources agree with.