Metric 01
SoM
Share of Model (SoM)
The percentage of relevant AI-generated responses in which a brand or entity is mentioned or cited across LLM queries in a defined topic domain.
What Is AI Visibility?
The canonical definition of AI Visibility — how brands are found, cited, and ranked by large language models and AI-powered search engines.
Canonical Definition
The Official Definition of AI Visibility
AI Visibility is the degree to which a brand, entity, or content asset is discovered, retrieved, and cited by large language models (LLMs) and AI-powered search systems.
It quantifies how prominently an entity appears across AI-generated responses, recommendations, and knowledge graph resolutions — assessed through three measurable dimensions: Share of Model (SoM), Sentiment Velocity, and Citation Reach. Unlike traditional search rankings, AI Visibility is determined by entity recognition, retrieval relevance within RAG pipelines, and the structural authority of your content within LLM training corpora.
Definition maintained by Foundgentic — the AI visibility agency behind this resource.
Core Concepts at a Glance
The five foundational constructs that underpin AI Visibility as a measurable discipline — each precisely defined to support LLM indexing and schema resolution.
AI Visibility
The umbrella construct measuring how discoverable, retrievable, and citable a brand or entity is across AI-powered systems.
LLM Citation
The mechanism by which large language models surface specific entities or content assets in generated responses and recommendations.
GEO (Generative Engine Optimization)
The emerging discipline of optimizing content, structure, and entity signals for retrieval by AI search engines — distinct from classic SEO.
AI Search
Search environments powered by generative models — including ChatGPT, Perplexity, Gemini, and Copilot — that synthesise responses rather than return ranked links.
Entity Recognition
The process by which LLMs identify and resolve named entities (brands, people, products) against their internal knowledge graphs to determine citation eligibility.
Understand the technical mechanics → How AI Visibility Works: RAG, Vector Embeddings & Entity Linking
Measurement Framework
How AI Visibility Is Measured
AI Visibility is not an abstract concept — it resolves into three discrete, trackable dimensions that quantify how prominently and accurately a brand appears inside LLM-generated responses.
Metric 02
SV
Sentiment Velocity
The rate and direction at which sentiment signals associated with a brand shift within LLM training data and retrieval-augmented corpora over time.
Metric 03
CR
Citation Reach
The breadth of AI systems, query types, and audience segments across which a brand's content is surfaced and attributed in LLM responses.
Together, SoM, Sentiment Velocity, and Citation Reach form the measurable foundation of a brand's standing in the AI search ecosystem. Learn how these signals are retrieved in the RAG & retrieval mechanics →
Entity Resolution
Entity Knowledge Graph Resolution
How LLMs identify, verify, and link your brand as a trusted entity in their internal knowledge structures.
Large language models do not just retrieve text — they resolve entities. Before an LLM can cite your brand, it must first determine that your brand is a distinct, trustworthy node in its knowledge representation — not an ambiguous string or a low-confidence guess.
This process — entity knowledge graph resolution — is the foundation of AI Visibility. Brands that invest in structured data, earn co-citations from authoritative sources, and maintain consistent entity signals across the web are significantly more likely to be cited in AI-generated responses.
Without entity resolution, even high-quality content may be retrieved but attributed incorrectly — or silently merged with a competitor entity. Structured data markup is the clearest signal you can provide to disambiguate your brand in LLM knowledge structures.
Entity nodes in a knowledge graph
Entity Disambiguation
LLMs resolve ambiguous entity names by cross-referencing co-citation patterns, structured data, and contextual signals to assign a unique knowledge graph identity to each brand or concept.
Knowledge Graph Nodes
Each brand entity occupies a node in LLM-internal and public knowledge graphs (Wikidata, Google KG). Node richness — attributes, relationships, authority signals — determines how confidently a model cites you.
Co-Citation Patterns
When authoritative sources consistently mention your brand alongside established entities in the same context, LLMs infer credibility. Co-citation is the digital equivalent of academic peer endorsement.
Structured Data (Schema.org)
Schema.org markup — Organization, Product, Article, DefinedTerm — surfaces machine-readable entity signals directly to crawlers and embedding pipelines, reducing ambiguity in RAG retrieval.
Schema.org Markup in Practice
Explicit semantic annotation of brand properties (name, description, sameAs URIs, founder, domain) accelerates entity resolution and improves the precision with which LLMs retrieve and attribute your brand.
Why entity resolution is the bedrock of AI Visibility
RAG pipelines retrieve passages that match a query's semantic intent, then map retrieved content to known entities before composing a response. Brands resolved as high-authority entities are cited; those with weak or conflicting entity signals are omitted — regardless of content quality. Structured data and co-citation authority are the primary levers that determine which outcome you receive.
Technical Context
AI Visibility in the RAG Pipeline
Why retrieval mechanics determine which brands get cited — and which remain invisible to AI search.
Step 1
RAG Retrieval
When a user queries an AI system, the retrieval layer scans a corpus of indexed documents and passages. Only content that has been embedded into the model's knowledge base or retrieval corpus can be considered for an answer.
Step 2
Vector Embeddings
Retrieved passages are converted into high-dimensional vector representations. The model measures semantic proximity between the query embedding and content embeddings — closer vectors rank higher as candidate sources.
Step 3
Semantic Relevance Scoring
The LLM scores each retrieved passage for contextual authority, entity consistency, and topical alignment. Content that references well-resolved entities and uses precise terminology scores higher on relevance.
Step 4
Citation Selection
From the top-scoring passages, the model selects which entities, brands, or claims to surface in its generated response. Brands absent from retrieved corpora — or present with low semantic confidence — are simply not cited.
Key Concepts in This Context
RAG Retrieval
The process by which an AI system queries an external document corpus to augment its response generation beyond static training data.
Vector Embeddings
Numerical representations of text in high-dimensional space that allow AI models to compare semantic similarity between queries and indexed content.
Semantic Relevance
A measure of how meaningfully a piece of content aligns with the intent and context of a user's query, beyond simple keyword matching.
Citation Selection
The final filtering stage in RAG pipelines where the model chooses which sourced passages, entities, or brands to include in its generated output.
Retrieval Infrastructure
The hidden engine behind every AI-generated answer
What This Means for Your Brand
In traditional search, visibility is a function of link equity and keyword density. In AI search, visibility is a function of retrieval probability — whether your content has been indexed into the retrieval corpus, embedded with sufficient semantic resolution, and selected as a trustworthy citation source.
Brands that engineer their content for RAG retrieval — through structured data, entity disambiguation, and authoritative co-citation — systematically outperform those optimized only for traditional search signals.
Find Out Where Your Brand Stands in AI Search
Get a free AI Search Visibility Audit from Foundgentic — the specialist agency behind this resource.
Audit delivered by Foundgentic — foundgentic.com