--- title: "GEO glossary" slug: "glossary" canonical_path: "/glossary" meta_title: "GEO Glossary — every term Citorum uses, defined — Citorum" meta_description: "A working glossary of Generative Engine Optimization terms — share of citation, recommendation rate, the Citorum GEO Index, retrieval-augmented generation, upstream source attribution, and more." h1: "GEO glossary" date: "2026-05-03" last_updated: "2026-05-03" --- > **Title:** GEO Glossary — every term Citorum uses, defined — Citorum > > **Canonical URL:** https://trycitorum.com/glossary > > **Description:** A working glossary of Generative Engine Optimization terms — share of citation, recommendation rate, the Citorum GEO Index, retrieval-augmented generation, upstream source attribution, and more. > > **Published:** 2026-05-03 # GEO glossary **Direct Answer:** A working glossary of Generative Engine Optimization terms. Each entry has the canonical Citorum definition, the operational measurement where one applies, and a link to the article that explores the concept in depth. If you are reading a Citorum dashboard, a Monday report, or an article and you hit a term you have not seen, the definition lives here. ## Answer engine A large-language-model-based system that returns a single generated paragraph (with optional citations) in response to a buyer's question. The six engines Citorum tracks daily are ChatGPT, Perplexity, Gemini, Claude, Copilot, and Google AI Overviews. See [the homepage](https://trycitorum.com/) for the engine list. ## Citation A named mention of a brand, product, or entity inside an answer-engine response. Counted at the entity level, not the string level — *"Citorum,"* *"trycitorum,"* and the brand domain all roll up into one citation. See [share of citation, explained](https://trycitorum.com/articles/geo/share-of-citation-explained). ## Citorum GEO Index A 0–100 composite score per brand, per engine, per day. Weighted sum of share of citation (50%), recommendation rate (30%), and sentiment (20%), normalized against the active competitive set in the prompt category. See [the GEO Index methodology](https://trycitorum.com/articles/geo/the-citorum-geo-index-methodology). ## Daily fan-out The Citorum pipeline of running every tracked prompt against every supported engine on a 24-hour cadence. We do not sample; we run the full set every day so trend lines are real. See [how it works](https://trycitorum.com/how-it-works). ## Direct answer The 35-to-70-word paragraph at the top of a page that resolves the buyer question before any preamble. The single highest-leverage piece of editorial real estate on a GEO page. See [how to write content ChatGPT will quote](https://trycitorum.com/articles/geo/how-to-write-content-chatgpt-will-quote). ## Engine breakdown Per-engine reporting of every Citorum metric — share of citation, recommendation rate, sentiment, and the Index — for ChatGPT, Perplexity, Gemini, Claude, Copilot, and Google AI Overviews. The same brand routinely sits at very different scores on two engines for the same prompt set in the same week. See [the metrics page](https://trycitorum.com/metrics). ## Entity resolution The normalization step that resolves brand mentions to canonical entities so *"Acme"*, *"Acme, Inc."*, and *"acme.com"* are one count, not three. Schema.org `Organization` markup with `sameAs` is the cleanest signal a brand can give the engines. See [schema.org Organization documentation](https://schema.org/Organization). ## FAQPage schema A `schema.org` JSON-LD type that exposes question/answer pairs as structured data. Pairs of `**Question?**\nAnswer paragraph.` in the source markdown roll up into a `FAQPage` block on every Citorum article. Cited at high rates by Perplexity, Copilot, and Google AI Overviews when the questions mirror real buyer prompts. ## Generative Engine Optimization (GEO) The practice of structuring brand, content, and citations so that large-language-model answer engines name and recommend you when buyers ask category questions. See [what is Generative Engine Optimization?](https://trycitorum.com/articles/geo/what-is-generative-engine-optimization). ## llms.txt A machine-readable index for LLM crawlers, modeled on `robots.txt`, that tells answer engines which URLs to read for which kinds of questions. Citorum publishes an [llms.txt](https://trycitorum.com/llms.txt) and an expanded [llms-full.txt](https://trycitorum.com/llms-full.txt) at the site root. See [llmstxt.org](https://llmstxt.org/) for the proposal. ## Prompt set The list of buyer prompts a brand defines for tracking. Each prompt is run against every engine, every day. Versioned in Citorum so historical comparisons stay honest when prompts change. See [how it works](https://trycitorum.com/how-it-works). ## Recommendation rate Of answers where the model recommends a specific tool or vendor, the percentage that recommend you. Stricter than share of citation: a citation can be incidental, a recommendation is the model picking sides. See [the metrics page](https://trycitorum.com/metrics). ## References (article) The end-of-article block listing the publications, primary documents, and reports a Citorum article sourced its claims from. Every Citorum article carries a `## References` section with absolute URLs. ## Retrieval-augmented generation (RAG) An architecture in which the model retrieves passages from a live index at query time and uses them to generate the answer. Perplexity, Copilot, and Google AI Overviews are heavily retrieval-augmented; ChatGPT and Claude use a mix of retrieval and trained weights. RAG engines refresh in hours; closed-weight engines refresh on training cadences. ## Sentiment The framing the model attaches to your brand when it does cite you — positive, neutral, hedged, or negative. Sentiment skew is what catches narrative drift before it becomes the consensus answer. See [the metrics page](https://trycitorum.com/metrics). ## Share of citation The percentage of relevant AI answers that name your brand at least once across a defined set of buyer prompts and engines, in a defined window. The closest GEO equivalent to share of voice. See [share of citation, explained](https://trycitorum.com/articles/geo/share-of-citation-explained). ## Upstream source The URL, publication, or thread (often Reddit, G2, a named publication, or primary documentation) that an answer engine most likely leaned on when generating a citation. Citorum's reporting pipeline is built around upstream-source attribution because that is where the editorial response lands. See [why Reddit is a GEO surface](https://trycitorum.com/articles/geo/why-reddit-is-a-geo-surface). ## Read more - [What is GEO?](https://trycitorum.com/articles/geo/what-is-generative-engine-optimization) - [Metrics & methodology](https://trycitorum.com/metrics) - [Compare Citorum to other AI search trackers](https://trycitorum.com/compare/ai-search-trackers)