Cross-cutting
AI optimisation glossary
The domain vocabulary, carefully defined. Essential for understanding the ChatGPT, Perplexity, Gemini, Claude and AI Overviews ecosystems — and for holding a precise conversation with your teams.
-
AEO
Answer Engine Optimization
- Optimisation for answer engines — a term that predates GEO, originally tied to Google featured snippets and voice assistants. Now largely used as a synonym for GEO.
-
AI Overviews
AIO
- AI-generated summaries shown above the regular results in Google Search for some queries. Formerly SGE (Search Generative Experience).
-
Chunking
- Splitting a document into segments (chunks) for retrieval. The size, semantic coherence and boundaries (headings, paragraphs) of those chunks drive the quality of LLM extraction.
-
Citation-friendly content
- Content designed to be quoted: standalone sentences, sourced claims, clear semantic structure, explicit and dated figures.
-
ClaudeBot
- Anthropic’s user-agent for crawling content used in Claude training and search.
-
Core Web Vitals
- Performance metrics (LCP, CLS, INP) treated as quality signals by Google. Still relevant for classic SEO and contribute to crawlability by AI bots.
-
Corpus
- The set of documents an AI engine queries or was trained on. Varies significantly between ChatGPT, Perplexity, Gemini and Claude.
-
Entity
- A named object in the world (person, brand, product, place, concept). LLMs reason in entities more than in keywords: disambiguating an entity (co-occurrence, schema, topical authority) is a core lever.
-
GEO
Generative Engine Optimization
- The set of practices aimed at improving a site’s visibility in responses generated by AI answer engines (ChatGPT, Perplexity, Gemini, Claude, AI Overviews). Formalised in the academic literature in 2024 (Aggarwal et al., Princeton).
-
Google-Extended
- A robots.txt directive specific to Google’s AI training (Gemini, Vertex AI). Distinct from regular Google Search indexing.
-
GPTBot
- OpenAI’s official user-agent for crawling content used for training and (together with OAI-SearchBot) for ChatGPT Search.
-
Knowledge graph
- A structured representation of knowledge as entities and relationships. Google, Microsoft and others use KGs upstream of LLMs.
-
LLM Optimization
- Umbrella term for optimising a website to be understood, cited and reused correctly by large language models and the search systems that rely on them.
-
llms.txt
- A Markdown file served at the root of a website, proposed as a convention to expose a curated table of contents to LLMs. Adoption is early. See the dedicated resource for the full spec.
-
LOOP
- llmoptimisation.fr’s proprietary framework: Legibility, Ontology, Operations, Performance. Four dimensions for steering AI optimisation.
-
Passage retrieval
- A technique that retrieves not a whole document but a passage (typically a few sentences) judged relevant to a query. Content structured as standalone sections performs better here.
-
PerplexityBot
- Perplexity’s user-agent, used to build the index the engine queries via RAG.
-
RAG
Retrieval-Augmented Generation
- An architecture that pairs a language model with a retrieval step. Engines like Perplexity and ChatGPT Search rely on RAG: corpus relevance and content structure matter as much as the model itself.
-
Share of AI Voice
- The share of a brand’s citations in answers generated by AI engines across a given set of queries. A proxy metric for AI visibility.
-
Standalone passage
- A paragraph or section that can be extracted and read out of context without losing its meaning or precision. Fundamental for LLM reuse.