18
- Discipline
AEO
Answer Engine Optimization
-
A set of practices intended to make an answer easy for an answer engine to extract and present. The term has no standard definition and its scope often overlaps with SEO and GEO.
- Product
AI Overviews
-
A Google Search feature that may show an AI-generated summary with links to sources. A page needs no special markup beyond ordinary Google Search requirements to be eligible.
Primary source or specification: Google Search documentation - Architecture
Chunking
-
Splitting a document into segments used during indexing or retrieval. The appropriate size depends on the content, embedding model, retrieval system and user need.
- Web measurement
Core Web Vitals
-
A set of three field user-experience metrics: LCP for loading, INP for responsiveness and CLS for visual stability. They do not measure editorial quality.
Primary source or specification: web.dev documentation - Semantics
Entity
-
An identifiable object to which properties or relationships can be assigned, such as an organisation, person, product or concept. Explicit context and identifiers reduce ambiguity.
- Quality
Evaluation
Eval
-
A reproducible procedure that measures an expected AI-system behaviour on a set of cases. A useful evaluation describes the data, criterion, model, settings and limitations.
Primary source or specification: OpenAI Evals guide - Discipline
GEO
Generative Engine Optimization
-
A name used for optimising visibility in generative answers. A KDD 2024 research paper studied several methods in an experimental environment, without establishing a citation guarantee for commercial engines.
Primary source or specification: GEO paper, arXiv 2311.09735 - Structured data
JSON-LD
-
A JSON syntax for expressing linked data. On the web, it is notably used to embed Schema.org vocabulary in an HTML page.
Primary source or specification: W3C JSON-LD 1.1 specification - Umbrella term
LLM Optimization
LLMO
-
An ambiguous expression that can refer to optimising an LLM application, a model, costs, evaluations, or content visibility in AI answers. The context should identify the discipline.
Primary source or specification: Five-disciplines framework - Proposal
llms.txt
-
A proposed Markdown format for presenting a selected set of site resources to models. It is neither a web standard nor a guaranteed search-engine directive.
Primary source or specification: llms.txt proposal - Protocol
MCP
Model Context Protocol
-
An open protocol that standardises how an AI application can connect to data sources, tools and workflows. It does not by itself determine an integration's permissions or security.
Primary source or specification: MCP documentation - Crawler
OAI-SearchBot
-
An OpenAI crawler associated with search in ChatGPT. OpenAI distinguishes it from GPTBot, related to training, and ChatGPT-User, used for certain user-initiated actions.
Primary source or specification: OpenAI bots documentation - Performance
Prompt caching
-
A mechanism that reuses computation associated with an identical prompt prefix. Its effects on cost and latency depend on the provider, model and cache-retention rules.
Primary source or specification: OpenAI Prompt Caching guide - Model
Quantization
-
A technique that represents weights or activations at lower numerical precision to reduce memory or compute cost. Gains and quality loss must be measured for each model and task.
Primary source or specification: Hugging Face documentation - Architecture
RAG
Retrieval-Augmented Generation
-
An architecture that retrieves documents or passages before generation to provide context to the model. Its quality depends on the corpus, indexing, retrieval and end-to-end evaluation.
Primary source or specification: Original RAG paper - Web protocol
robots.txt
-
A file that tells compliant crawlers which resources they may crawl. The protocol is not an access-control mechanism and does not guarantee that a URL will not be indexed.
Primary source or specification: RFC 9309 - Vocabulary
Schema.org
-
A shared vocabulary for describing entities and their properties in structured data. It helps machines interpret a page but guarantees neither a rich result nor an AI citation.
Primary source or specification: Schema.org documentation - Measurement
Share of AI Voice
-
A constructed metric for tracking a brand or source across a defined panel of AI answers. It is comparable over time only when queries, engines, settings, dates and counting rules are retained.