Glossary. AI Search
terms defined.

Agentic AI refers to AI systems that can plan, decide, and take actions on behalf of a user rather than only returning information. In search, agentic AI changes discovery because assistants may resea...

Agentic Capabilities are the functions that allow AI systems to plan, browse, compare, call tools, remember context, and take multi-step actions. In AI search, these capabilities move assistants from ...

AI Attribution Rate measures how often an AI answer attributes information to your brand, website, or content when responding to relevant prompts. It is a visibility metric for the citation layer of A...

AI Citation Count is the number of times AI search engines cite your domain, pages, or assets in generated answers. It helps teams understand whether their content is being used as source material by ...

AI Search Optimization is the practice of making your brand, products, and content easier for AI search engines and assistants to find, understand, cite, and recommend. It extends traditional SEO beyo...

Article schema is structured data that helps search engines and AI systems understand article metadata such as headline, description, author, publisher, date published, date modified, image, and main ...

Bi-encoders and cross-encoders are model architectures used to evaluate relevance between a query and a document. A bi-encoder creates separate embeddings for the query and content, making it fast for...

BreadcrumbList schema is structured data that describes a page's position in a site's hierarchy, such as Home > Glossary > AI Search Optimization. It helps search engines understand navigation paths a...

Chunk Overlap is the repeated text shared between adjacent content chunks when documents are split for retrieval. It helps preserve context so a passage does not lose important setup, definitions, or ...

Chunkability is the ability of a page to be broken into useful, self-contained sections that AI systems can retrieve and reuse. AI search often works at the passage or chunk level, so a strong page is...

Citation Drift occurs when AI systems start citing different sources, pages, or domains for the same prompt over time. Drift can happen because content changes, competitors publish stronger pages, cra...

Citation-First Search describes AI answer experiences that foreground sources, references, or footnotes as part of the response. This format is common in tools that want users to verify claims or cont...

ColBERT is a neural retrieval approach that represents queries and documents with multiple token-level embeddings rather than a single dense vector. This allows more precise matching because the model...

Content Update Cadence is the rhythm at which important pages are reviewed, refreshed, and improved. In AI search, freshness matters because AI assistants and search engines need current facts, pricin...

A context window is the amount of text, data, or tokens an LLM can consider at one time when generating an answer. Larger context windows allow AI assistants to process more information, but they do n...

Core Web Vitals still matter for AI Search because performance, accessibility, and usability affect how pages are crawled, rendered, and experienced by users after an AI citation or recommendation. AI...

Embeddings are numerical representations of text, images, pages, or other content that capture meaning and relationships. In AI search, embeddings help systems match a user's prompt to semantically re...

Hallucination in AI is when a model produces information that sounds plausible but is inaccurate, unsupported, outdated, or fabricated. In AI search, hallucinations can affect brand descriptions, prod...

Hybrid Retrieval combines multiple retrieval methods, usually lexical search and vector search, to find the best content for a user's prompt. Lexical retrieval is strong when exact words matter, while...

Internal Linking for entities means connecting pages that explain related brands, products, features, concepts, personas, and use cases. For AI search, internal links help crawlers and retrieval syste...

A Knowledge Graph is a structured representation of entities and the relationships between them, such as companies, products, people, topics, locations, and concepts. In AI search, knowledge graphs he...

Large Language Models, or LLMs, are AI systems trained to understand, generate, summarize, and reason over language. In AI search, LLMs power many answer experiences by interpreting a user's prompt, r...

LLM Answer Coverage measures how many relevant questions or prompts a page can help an AI system answer accurately. High answer coverage does not mean writing one unfocused page for every possible top...

Programmatic GEO is the structured creation and optimization of many AI-search-friendly pages from reliable data, templates, and repeatable content patterns. It is similar to programmatic SEO, but the...

Query Decomposition is the process of breaking a complex prompt into smaller sub-questions that can be answered, retrieved, or reasoned through separately. AI search engines use decomposition when use...

Reciprocal Rank Fusion, or RRF, is a ranking method that combines results from multiple search systems by rewarding documents that rank well across more than one list. In AI search, RRF is useful beca...

Recommendability is the likelihood that an AI assistant will present your brand, product, or content as a good option when a user asks for advice. It goes beyond visibility because being mentioned is ...

Robots.txt for AI is the practice of using robots directives to control how AI crawlers, search bots, and training-related user agents access your site. It matters because different AI systems use dif...

RRF Rank Contribution describes how much a document contributes to final ranking when Reciprocal Rank Fusion combines multiple retrieval results. Many AI search systems blend lexical search, vector se...

Semantic Density Score describes how much useful meaning, entity context, and relationship information is present in a piece of content. In AI search, semantically dense content gives retrieval system...

Source Diversity refers to the range of distinct domains, formats, and perspectives an AI system uses when generating an answer. It matters because AI assistants often look for corroboration rather th...

Static Site Generation, or SSG, prebuilds pages as HTML so content is available quickly without requiring heavy runtime rendering. For AI search and SEO, SSG is valuable because crawlers can access th...

Tokens are the units of text that LLMs process, often representing words, word parts, punctuation, or spaces. Token limits shape how much content an AI assistant can read, retrieve, summarize, and inc...

User-Generated Content, or UGC, includes discussions, reviews, comments, forum posts, videos, and community answers created by users rather than brands. AI search engines often draw on UGC because it ...

Vector Index Presence Rate estimates how consistently your content appears in the retrieval layer of AI systems that use vector search. In practical terms, it asks whether your pages are semantically ...

Vector search algorithms find content by comparing embeddings, which represent meaning in a numerical space. Common approaches include approximate nearest neighbor methods such as HNSW, IVF, and produ...

A Vectorization Pipeline is the process that turns content into embeddings and stores them for semantic retrieval. It usually includes crawling or ingesting content, cleaning text, splitting it into c...

Versioned Docs are documentation pages that preserve information for specific product, API, or platform versions. They are important for AI search because assistants may answer technical prompts using...

VideoObject schema is structured data that helps search engines and AI systems understand a video's title, description, duration, thumbnail, upload date, transcript, and source URL. For AI search, vid...