SEO #
Improving a site’s accessibility, usefulness and presentation in organic search. SEO includes technical, content and reputation work and remains relevant to Google’s AI search features.
Executive AI-search reference
The language behind AI visibility
Decision-ready definitions for CMOs, marketing leaders, and conference audiences navigating the shift from rankings to citations and agentic discovery.
Search all 43 definitions.
Start with the decision
The terminology matters only when it changes what your team does next.
Visibility strategyWhat are we measuring?
Separate mentions, citations and competitive share.
Open the concepts →Category leadershipWhich questions define our topic?Connect coverage, authority and supporting sources.
Open the concepts →Identity and trustDo the sources describe the right entity?Verify identities, relationships and evidence.
Open the concepts →Agentic discoveryCan a system find and use our information?Distinguish access, retrieval and permitted actions.
Open the concepts →Choose the right lens
| Approach | Focus | Decision |
|---|---|---|
| SEO | Organic search access, relevance and presentation | Can the right page be found and understood? |
| AEO | Extractable direct answers | Does the content answer the question clearly? |
| GEO | Presence and evidence in generated answers | Are we mentioned accurately, and what gets cited? |
| Agentic ASO | Discoverable information and permitted actions | Can an agent evaluate the offer and complete an authorized step? |
These are working distinctions, not competing standards. SEO remains relevant to AI search. In this glossary, ASO refers to agentic search, not app stores. Read GEO vs SEO →
One sample, three different metrics
Hypothetical example, not site performance: 20 valid answers, 8 naming Brand A, 3 linking to its domain. Across a fixed set of brands, there are 32 answer-level brand mentions, counting each brand once per answer.
| Metric | Counting rule | Example |
|---|---|---|
| Mention rate | Answers naming Brand A / valid answers | 8 / 20 = 40% |
| Domain citation rate | Answers linking to Brand A’s domain / valid answers | 3 / 20 = 15% |
| Competitive mention share | Brand A mentions / all tracked brand mentions | 8 / 32 = 25% |
This competitive-share formula is the convention used in this example, not a universal vendor definition. Keep prompts, language, market, platform, dates and competitor set fixed when comparing rounds. Count failed requests as missing data. Track exact-page citations and factual accuracy separately.
From access to an answer
- CrawlingFetch a page
- IndexingProcess information for later retrieval
- RetrievalSelect relevant information
- Generation + groundingCompose an answer using reference context
No definition matches that search. Try a broader term such as “visibility”, “entity”, or “trust”.
Visibility and measurement
GEO, Generative Engine Optimization #
The discipline of improving how content and entities surface inside generative-engine answers. In business practice, GEO measures whether a brand is mentioned, accurately represented, and cited across platforms such as ChatGPT, Google AI Overviews, Perplexity, and Gemini.
Primary source: Aggarwal et al., “GEO: Generative Engine Optimization” (2023).
Example: A team audits which answers mention its brand and which cite its research.
Boundary: A mention alone does not establish accuracy or a citation.
AEO, Answer Engine Optimization #
Structuring content so an answer engine can extract and return it as a direct answer, including featured snippets, voice assistants, and AI summaries. AEO overlaps with GEO, but focuses on answer extraction rather than the wider problem of entity visibility and citation.
AI Visibility #
How often and how prominently a brand appears across a defined set of AI-generated answers. A defensible score separates mentions, citations, answer position, prompt coverage, and platform coverage instead of blending them into one opaque number.
Example: Record the prompt set, language, platform and date before comparing rounds.
Boundary: Changing the panel changes the scope of the comparison.
AI Mention #
An appearance of a brand or entity by name inside an AI-generated answer. A mention is not a citation: the engine may name the brand while sourcing the claim from a third-party domain.
AI Citation #
A linked or otherwise attributable reference to a source inside an AI-generated answer. A citation is distinct from a brand mention: it exposes the evidence behind the answer and may create a path to verification or traffic.
AI Share of Voice (SOV-AI) #
A competitive share metric whose meaning depends on the counting rule. One explicit convention is a brand’s answer-level mentions divided by the total answer-level mentions of a fixed set of brands. Record the competitor set and count each brand once per answer. Keep this separate from mention rate and citation rate.
Zero-Click Search #
A search that produces no click to an external website within the measurement scope. The reason may be an on-page answer, an abandoned search or another behavior; define the dataset and session boundary before interpreting a rate.
AI Overviews #
Google Search summaries generated with AI for queries where the feature is shown, with supporting links. Appearance and source selection vary by query; inclusion is not guaranteed.
Example: An answer summary can include links that let the reader inspect supporting pages.
Boundary: Source selection varies; there is no fixed citation slot entitlement.
AI Mode #
Google Search’s conversational AI experience for exploration and follow-up questions. It differs from an AI Overview shown within a results page. Models, techniques and supporting links may differ between the two experiences.
Example: Use AI Mode for a multi-part exploration; an Overview may summarize an eligible search query.
Boundary: Neither feature guarantees inclusion of your website.
Topical authority and sources
AI Topical Authority #
Consistent association of an entity with a subject across a defined sample of questions and sources. Assess coverage, accuracy and repeated presence; no finite prompt panel captures every possible question.
Related keynote: How to Own a Topic in ChatGPT.
Topical Gravity #
Semrush’s topic-led strategy framework: define relevant topics, build coverage and corroboration, measure competitive presence, then expand to adjacent topics. Here it guides editorial question mapping; it is not an official search-engine ranking factor.
Example: Map definition, comparison and implementation questions before commissioning more pages.
Boundary: The question map is an editorial hypothesis until supported by audience or search data.
Topic Ownership #
An operational claim of sustained topic leadership against a defined competitor set. Declare the prompts, platforms, markets, repetitions and success threshold before measuring. It is not a universal platform-issued badge.
Example: Set a repeated cross-platform panel and a competitor benchmark before declaring leadership.
Boundary: A single mention or a site-owned citation does not establish ownership.
Citation Authority #
In Fernando’s editorial framework, the ability of a source to be repeatedly used as supporting evidence for a topic. Assess exact cited URLs and context; a citation does not necessarily mean endorsement or recommendation of the author.
Example: Check whether an answer links to the original research or to a third-party summary.
Boundary: Separate citation frequency from factual accuracy and recommendations.
Citation Core #
A working label for sources repeatedly cited within a defined topic and answer sample. Identify it from retained URLs, prompts, platforms and dates rather than assuming a fixed set of influential domains applies everywhere.
Cross-Lingual Citation Authority (CLCA) #
Fernando’s editorial framework for connecting evidence and identity references across languages. Test each language and market separately; visibility and citations do not transfer automatically.
Identity and trust
Entity Density #
Fernando’s editorial framework for maintaining consistent identity facts, connected identifiers and verifiable corroboration. Entity Density is not a standardized score, a ranking factor or a guarantee of AI citations.
Example: Maintain the same verified role and identity across native English and Spanish biographies.
Boundary: Equivalent facts do not require identical wording.
Entity SEO #
Work that clarifies the people, organizations, products and concepts described by content, and their relationships. It complements keyword and semantic work; accurate identity alone does not ensure visibility.
Knowledge Graph #
A structured network of entities and their relationships, used by search and AI systems to connect and disambiguate facts. Inclusion can strengthen entity resolution, but no single knowledge graph determines whether a model recognizes a brand.
Wikidata #
An open, collaboratively edited knowledge base used across Wikimedia projects and many data products. Its Q-identifiers provide stable public IDs that can help systems disambiguate an entity; they are useful infrastructure, not proof or a visibility guarantee.
Example: Check that a Q-identifier describes the intended person, not a namesake.
Boundary: An encyclopedia entry is not a prerequisite for the identity audit.
Schema.org & JSON-LD #
Schema.org is a vocabulary for describing entities and relationships. JSON-LD is a format for expressing linked data using that vocabulary. Choose types that fit the visible content; Google does not require special AI schema for AI Overviews or AI Mode.
Example: Describe a person with Person and refer to that same subject through a stable @id.
Boundary: Markup must reflect visible facts; it cannot manufacture evidence.
@id #
A JSON-LD keyword that identifies a node. Reuse a stable identifier when referring to the same subject across pages; it is not itself a verification badge or an external knowledge-graph entry.
E-E-A-T #
Google’s framework of Experience, Expertise, Authoritativeness and Trustworthiness for assessing content quality. Quality-rater scores do not directly determine rankings, and E-E-A-T is not a single ranking factor or a documented universal rule for AI citations.
Primary source: Google Search Central, people-first content guidance.
Example: A reader checks who produced the research and how its claims are supported.
Boundary: Do not treat a quality framework as a citation guarantee.
Experience #
First-hand involvement relevant to the subject, such as using a product or reporting an event. Evidence of experience helps a reader assess a claim; its importance depends on the page’s purpose.
Expertise #
The author's depth of knowledge in the topic, demonstrated by credentials, publications, and the precision of the content itself. Formal credentials matter for YMYL topics; demonstrated craft matters for everything else.
Authoritativeness #
Recognition of a creator or source as a reference for a subject. Relevant independent references can support that assessment; no single citation count establishes authority across every topic.
Trustworthiness #
The most important of the four. Is the page accurate, transparent about its author, safe, and honest about commercial relationships? Low trust caps the value of any expertise or authority you've built.
YMYL, Your Money or Your Life #
Pages that can materially affect a reader's health, finances, safety, or civic life. Held to the strictest E-E-A-T bar by Google's Quality Rater Guidelines.
Quality Rater Guidelines (QRG) #
Google’s guidelines for human search quality raters. They explain evaluation concepts, including E-E-A-T, but are not the ranking algorithm and do not define a universal AI-citation formula.
Author Entity #
The identifiable person or organization responsible for content, connected to a clear biography and relevant evidence of their work. Author markup should match the visible byline; adding it does not establish expertise by itself.
sameAs #
A Schema.org property linking to a page that unambiguously identifies the same entity. Use verified identity references, not an employer’s homepage, a namesake or a merely related article.
Example: A verified professional profile can identify the same person as the official biography.
Boundary: A related company is a different entity.
Disambiguation #
Distinguishing subjects that share a name by checking identifying facts, context and references. Clear descriptions and verified links can express the distinction, but markup does not guarantee correct interpretation by a system.
Retrieval and agents
ASO, Agentic Search Optimization #
In this glossary, ASO means Agentic Search Optimization: preparing information and permitted actions so agents can discover, evaluate and interact with an offer. This emerging usage differs from App Store Optimization and is not a universal standard.
Leadership question: Can an agent understand the offer, verify the evidence, and complete the next action without guessing?
MCP, Model Context Protocol #
An open protocol that standardizes how AI applications connect to external data sources and tools. For marketing leaders, MCP matters because it moves discovery from reading content toward agents accessing structured resources and taking permitted actions.
Primary source: Model Context Protocol specification. Read the brand-visibility analysis →
Answer Engine #
An interface that presents a generated or extracted answer, often alongside links and other search results. The format and sources vary across products and queries.
AI Crawlers #
Automated clients that fetch web content for different purposes, including search and model development. Check each provider’s documentation before changing access. Google-Extended is a robots.txt control token, not a separate HTTP crawler; it controls specified Gemini training and grounding uses, not inclusion in Google Search.
Example: Decide separately how content may be used for training and surfaced in search.
Boundary: Do not generalize one provider’s robots controls to another.
Indexing #
Processing and storing information so it can be considered for retrieval. Crawling a page does not guarantee indexing, and indexing does not guarantee selection or citation in an answer.
Retrieval #
Selecting information relevant to a query from available sources. Retrieval may combine techniques; the selected passages are inputs to an answer, not proof that the answer is correct.
RAG, Retrieval-Augmented Generation #
Retrieval-Augmented Generation combines retrieval of external information with answer generation. It can supply relevant context at response time, but retrieved material and generated claims still need verification.
Grounding #
Connecting a generated response to reference information supplied or retrieved for the task. Grounding supports verification but does not guarantee that every statement accurately reflects the source.
Example: Compare the generated claim with the passage offered as its source.
Boundary: A linked source can still fail to support the claim.
Embedding #
A numerical representation of content that can support similarity-based retrieval. The model and retrieval design affect what counts as similar; vector proximity is not proof of factual accuracy.
Chunking #
Dividing content into smaller passages for processing or retrieval. Clear headings and self-contained paragraphs help readers and can help retrieval systems interpret a passage, but do not guarantee a citation.
llms.txt #
A proposed convention for publishing a curated Markdown index that helps language models locate the most useful pages on a site. It is not a crawling directive like robots.txt, not a formal web standard, and not a guarantee that an AI system will ingest the file.
Example: A site publishes a curated index of its useful pages.
Boundary: This is not robots.txt and is not required by Google’s AI features.
Need your audience to understand what changes after rankings?
Bring the AI-search briefing to your event →