Glossary
AI search glossary
Plain-language definitions for the ideas behind AI search visibility. Each term explains what the idea means, why it matters when ChatGPT, Gemini, Perplexity, Claude, or Google AI features answer a question, and where to read more on Kollavo.
A
AI citation
An AI citation is a source an answer engine shows for a statement in its response, usually as a link or a named page. It is evidence the engine attached to the answer, not a guarantee the claim is correct. A brand mention with no source is not a citation.
AI crawler
An AI crawler is a bot an AI company uses to fetch pages. Public examples include OpenAI’s GPTBot, Anthropic’s ClaudeBot, and PerplexityBot. Sites allow or limit them in robots.txt. A fetch by one of these bots is not the same event as a citation in a consumer answer.
AI Overviews
AI Overviews are Google’s AI-written summaries that can appear above classic search results. They answer the query in the results page and can show supporting links. They use Google Search systems. Google does not ask sites to add special markup just for these summaries.
AI referral traffic
AI referral traffic is visits that analytics attributes to an AI product, usually because the browser sent a referring host or a link carried a tag you recognize. It counts sessions that arrived. It does not count citations where nobody clicked, and it misses visits that arrive with no referrer.
answer engine optimization (AEO)
Answer engine optimization, or AEO, is the work of earning the answer itself, not only a ranked link. It started with featured snippets and similar result features, and it now includes AI answers that resolve a question on the results page or inside a chat. AEO and GEO overlap, and neither one guarantees placement.
B
C
citation decay
Citation decay is a lasting drop in how often a URL is cited for a fixed prompt set, after you account for ordinary run-to-run variation. One missing citation on one day is not decay. Decay is a decline that remains when you keep the prompts, engines, and counting rules the same.
citation gap
A citation gap is a buyer prompt where AI answers cite competitors, or other sources, and do not cite you. It can also mean your brand is missing while rivals are named. A gap is specific to a prompt and an engine. It is not a single site-wide score.
citation share
Citation share is your portion of the sources AI answers cite for a defined prompt set. Divide citations that point to you, or to coverage you have agreed to count, by all citations recorded, and keep each engine separate. It is a share of sources, not a share of brand mentions.
E
E-E-A-T
E-E-A-T stands for experience, expertise, authoritativeness, and trustworthiness. Google uses the idea in its search rater guidelines to describe quality. It is not a ranking factor you can mark up, and it is not an LLM setting. For AI visibility, it is a reminder to show real experience and sources a person can check.
earned media
Earned media is coverage a brand earns from an independent publisher: a review, a news story, a comparison, or a roundup the brand does not control. It is distinct from ads and from pages on your own site. In AI search, earned pages are often the sources an answer cites.
entity SEO
Entity SEO is the work of helping search systems and language models recognize a specific brand, product, person, or place, and understand how it relates to a category. It relies on consistent names, descriptions, and corroboration across pages you control and pages you do not. It is clarity, not a schema trick.
G
generative engine optimization (GEO)
Generative engine optimization, or GEO, is the practice of making a brand easier for AI answer engines to mention and cite. It covers the questions buyers ask, the pages that answer them, and the independent sources those engines already use. GEO sits beside ordinary search work, and no checklist guarantees a citation.
Google AI Mode
Google AI Mode is Google’s conversational search experience inside Search. People can ask follow-up questions, and the answers can include links to supporting pages. It draws on Google Search. It is a different surface from a classic list of blue links and from AI Overviews.
grounding
Grounding is connecting an AI answer to source material retrieved for that question, so statements can be checked against pages rather than left to the model alone. Citations are the visible form of grounding when an engine shows them. Grounding lowers the chance of invention. It does not make every sentence true.
H
L
listicle placements
A listicle placement is inclusion in a roundup article, such as a “best tools” list, on a publisher’s site. The page may be independently edited or commercially arranged. The placement is a mention on that URL. It becomes an AI citation only if an engine later links to it.
LLM visibility
LLM visibility is how readily a brand shows up when people ask large language models questions in its category. It includes named mentions, linked citations, and the competitors that appear instead. It is an observation across prompts and engines, not a single score those engines publish.
llms.txt
llms.txt is a proposed text file, usually at /llms.txt, that gives language-model tools a short Markdown overview of a site and links to important pages. It is a convention, not a replacement for robots.txt or a sitemap. Publishing it does not guarantee that any engine will crawl or cite you.
P
prompt tracking
Prompt tracking is repeating a fixed list of buyer questions in AI engines and logging what comes back: mentions, citations, sentiment, and competitors. The list stays stable so a later change reflects the answers, not a new set of questions. It is monitoring, not a rank the engines publish.
prompt volume
Prompt volume is the size of the prompt set you track, or the practical demand for a question based on how often buyers ask it. AI assistants do not publish a reliable search-volume number for chat prompts. A bigger tracked list is not automatically a better measurement.
Q
R
S
schema markup for AI
Schema markup for AI means structured data, usually JSON-LD, that describes a page’s entities and facts for machines. It can support ordinary search understanding. It is not a special tag that adds you to ChatGPT or to Google’s AI answers. Google has said its AI search features do not require markup beyond normal Search eligibility.
sentiment in AI answers
Sentiment in AI answers is whether a brand mention is favorable, neutral, mixed, or negative. It is a reading of the wording in that response, not a score the model publishes. The same brand can be praised for one use case and cautioned for another in a single answer.
share of voice in AI search
Share of voice in AI search is your brand’s portion of the brands an assistant names for a fixed set of prompts. It counts appearances, whether or not the answer links to you. It is a competitive comparison, and it is not the same metric as citation share.
source authority
Source authority, in AI search, is the tendency of answers in a category to cite a particular domain. You judge it by seeing the same sites appear across repeated prompts and engines. It is not a universal score, and a famous domain can be irrelevant to your buyers’ questions.
T
topical authority
Topical authority is the depth and consistency of a site’s coverage of a subject, so it can answer the cluster of questions a buyer asks, not a single keyword. It is a description of a body of work. Engines do not publish a topical authority score you can optimize against.
training data vs live retrieval
Training data is material a model learned before it answers. Live retrieval is pages fetched at answer time to inform that reply. An answer can use either, or both. A missing link does not prove the model is using training data, and a link does not reveal the training set.