The vocabulary of AI search
Every metric here is one we publish a formula for. Where a definition involves a judgement call, such as which brands belong in a share-of-voice denominator, the entry says so rather than glossing it.
The practice of getting a brand named and cited inside answers written by AI assistants, rather than ranked in a list of links.
A near synonym for GEO, usually emphasising direct-answer surfaces such as AI Overviews and featured snippets rather than conversational assistants.
The share of AI answers that mention your brand, weighted so that higher-demand questions and higher-reach engines count for more.
Your brand's proportion of all brand mentions detected across a set of AI answers, measured against the competitors appearing alongside you.
The proportion of AI answers that cite at least one page on your own domain as a source.
The pattern where search impressions hold steady while clicks decline, because AI answers resolve the question without the user visiting a site.
The hidden sub-questions a generative engine generates from one user prompt, each retrieving its own sources before the answer is written.
A proposed plain-text file at a site's root that offers language models a curated map of its most useful pages.
Google's generated summary that appears above the traditional results for many queries, citing a small set of sources.
Google's conversational search surface, powered by Gemini, which answers in dialogue rather than returning a ranked list.
Where your brand appears in the order of brands named within a single AI answer, counting from the first mention.
A system that responds to a question with a written answer rather than a list of links, naming sources inside or beneath it.
The pattern where a model searches for relevant documents first, then writes its answer using what it found.
Tying a generated statement to a specific retrieved source, so the claim can be attributed rather than invented.
A confident statement from a model that is not supported by any source, including invented facts about brands.
A search where the user gets what they needed on the results page and never visits a website.
Google's framework for assessing content quality: experience, expertise, authoritativeness and trust.
Optimising so machines understand what your brand is, what it does and how it relates to other things, rather than which keywords it targets.
Machine-readable code that tells a crawler what a page is and what each part of it means.
The user agents AI companies use to fetch web pages, either for training or to answer a question in the moment.
A structured map of entities and the relationships between them, used to reason about what a thing is rather than what words describe it.
An estimate of how often a tracked question is asked, used to weight some questions more heavily than others.
A group of tracked questions covering the same buying topic, scored and acted on together rather than individually.
The type of site a citation comes from: editorial, corporate, user-generated, reference, or your own domain.
The editorial stance a cited page takes: cost, depth, authority, freshness, or written for a named audience.
Whether a mention of your brand in an AI answer is positive, neutral or negative.
Looking for the arithmetic rather than the definition? Every formula the product computes is documented at /metric-docs, including the engine weights.