Foundations / Reference
The GEO glossary: AI search terms in plain English
Plain-English definitions of GEO, AEO, citations, grounding, retrieval, entities, prompt sets, and AI visibility metrics.
The short answer
Use this glossary to distinguish the systems that produce answers, the sources that support them, and the metrics used to evaluate them. These working definitions help make a GEO report understandable and testable.
GEO, SEO, and AEO
Generative engine optimization (GEO) is work aimed at improving representation in generated answers. The term is used in the GEO research paper; practical methods and metrics vary across providers.
Search engine optimization (SEO) improves the discoverability and usefulness of pages in search. GEO and SEO overlap; see the comparison guide.
Answer engine optimization (AEO) is a label used for work focused on answer-style discovery. It is not a separate platform certification. Ask which answer environment and outcome a proposal means.
Retrieval, grounding, and query fan-out
Retrieval is finding material relevant to a question. It is a step in some answer workflows, not proof that the material will appear in the final response.
Grounding connects a generated response to supporting information. The presence of sources helps a reader investigate an answer; it does not remove the need to verify claims.
Query fan-out describes a system issuing related searches to address a question. Google discusses grounding and query fan-out in its AI search guide. Do not treat the term as a requirement to publish a separate page for every possible subquery.
Mentions, citations, and source pages
Mention: the business is named in an answer. It may have been named in the user’s question first, so distinguish branded tests from discovery tests.
Citation: a visible reference or link to supporting material. A citation to a directory about a business is different from a link to the business website.
Source page: the page linked or referenced as evidence. Read it to check whether it supports the adjacent claim.
Referral: a visit arriving through a link from another service. A citation can exist without producing a referral.
Entities and business facts
Entity: an identifiable thing, such as a business, place, person, or service. In a local audit, resolving the correct entity means avoiding confusion between similarly named companies or different branches.
Business fact record: our term for an owner-approved list of current services, coverage, contact details, and other operational facts. It provides a comparison point for answer accuracy.
NAP: shorthand for name, address, and phone number. Consistency matters operationally, but an audit should also consider service areas, hours, and public-address rules.
Crawlability, indexability, and canonical URLs
Crawlability: whether a crawler can fetch a page. Network rules and site directives can affect access.
Indexability: whether a page is eligible to be included in a search index. Eligibility and actual inclusion are different.
Canonical URL: the preferred URL for a page or substantially equivalent content. Canonical tags communicate that preference to search systems.
Structured data: machine-readable information describing content. It should reflect the facts visible on the page. See the technical guide.
Prompt sets and observation metrics
Prompt set: a saved collection of exact questions used in an evaluation. Keep versions so a comparison does not quietly change its scope.
Observation: one recorded response under documented conditions. Failed requests should remain in the collection log.
Observed mention rate: the proportion of valid answers in a specified sample that name a business.
Observed site citation rate: the proportion of valid answers in a specified sample that link to the business website.
Grounding query: in platform reporting, a query associated with retrieving supporting information. Use the provider’s definition; Bing’s AI Performance help explains its reporting context.
Baselines and causal claims
Baseline: the starting evidence against which later observations are compared. It should include collection conditions and raw records.
Correlation: two things change together. It does not establish that one caused the other.
Causal claim: a statement that an intervention produced an outcome. A before-and-after screenshot alone does not establish this, especially when platforms, competitors, and sources also change.
Use the measurement guide for a worked example with explicit denominators.
Check the evidence
Sources & review notes
Primary references checked for this edition on . Platform documentation can change. Recommendations and illustrative examples are Smoketown GEO’s editorial guidance, not guaranteed outcomes.
- Aggarwal et al.: GEO — Generative Engine Optimization (KDD 2024)
- Google Search Central: Optimizing for generative AI search
- Bing Webmaster Tools: AI Performance
How we source and update these guides · Suggest a correction