// Glossary

Speak the language of the answer layer.

Every term defined answer-first — the first sentence is the definition — because that is exactly the structure an engine lifts. The glossary practises what the product preaches.

Start free audit

GEO: GEO is the practice of being named, cited and described accurately inside AI-generated answers, rather than ranked in a list of links.; Where SEO optimises for a results page, GEO optimises for the single answer an engine composes. The levers are different: machine-readable structure, verifiable claims, and presence in the sources engines retrieve from.; See alsoAEO LLMO Share of model
AEO: AEO is structuring content so that an answer engine can resolve a question directly to it — answer-first, with the facts an engine needs stated plainly.; Synthic's AEO site audit grades a page on access, structured data, meta and content: the things that decide whether an engine can read a page and lift an answer from it.; See alsoGEO Answer-first structure
LLMO: LLMO is shaping how large language models represent a brand across both their training data and their live retrieval.; Training-based mentions are slow to change and unverifiable; retrieval-based mentions can change the week you publish. Synthic labels every mention as one or the other.; See alsoRetrieval vs training basis GEO
Visibility Score: The Visibility Score is Synthic's 0–100 measure of how present a brand is in engine answers, computed as 55 × mention rate + 30 × average prominence/100 + 15 × citation rate over one audit's stored answers.; The formula is published and every input is a stored, verbatim answer. There is no baseline, panel or model behind it.; See alsoMention rate Prominence Citation rate
Mention rate: Mention rate is the share of an audit's answers in which the engine named the brand.; Six prompts across five engines is thirty answers; if the brand appears in twelve, the mention rate is 40%.; See alsoVisibility Score Prominence
Prominence: Prominence is a 0–100 judgement of how early and how centrally a brand appears in an answer that mentions it.; First and recommended scores high; a passing mention in the last paragraph scores low. It is judged by an analyst model reading the stored answer.; See alsoVisibility Score
Citation rate: Citation rate is the share of answers that cite a URL on the brand's own domain.; It measures whether the engine looked you up, not just whether it remembered you — which is why it is weighted into the score separately.; See alsoRetrieval vs training basis Visibility Score
Share of voice: Share of voice is the count of times each brand is named across the answers to a category's prompts — a leaderboard of who owns the answer today.; Synthic tallies both declared competitors and the rivals the engines surface on their own, and links every count to the answers it came from.; See alsoShare of model
Share of model: Share of model is the percentage of category answers in which a given brand is the one recommended.; The GEO analogue of market share. It moves per engine and per persona, which is why Synthic reports it that way.; See alsoShare of voice
Retrieval vs training basis: A mention's basis is retrieval when the engine backed it with a live citation, and training when it came from the model's memory with no source.; Retrieval-based mentions are what GEO can move quickly. Training-based ones are unverifiable and slow; Synthic marks both in the evidence line under every answer.; See alsoLLMO Citation rate
Hallucination: A hallucination is a confident, specific claim an engine makes about a brand that the brand's own site contradicts or does not support.; Synthic checks each claim against ground truth crawled live from the site and returns a verdict — accurate, inaccurate or unverifiable — with a severity and a correction.; See alsoGround truth
Ground truth: Ground truth is the brand's own published content, crawled live at scan time and used as the reference every engine claim is checked against.; It is never cached from a previous scan, so a correction you published this morning counts this afternoon.; See alsoHallucination
llms.txt: llms.txt is a plain-text file at a site's root that hands language models a canonical, machine-readable summary of what the site is and where its key pages are.; An emerging convention, not yet a standard. Synthic generates it from crawled pages and checks for its presence in the site audit.; See alsoJSON-LD AEO
JSON-LD: JSON-LD is the structured-data format — Organization, WebSite, Product, FAQPage — that tells machines what a page is about in a form they can parse without guessing.; Synthic generates it strictly from page content and leaves out any field the page doesn't state.; See alsollms.txt Answer-first structure
Answer-first structure: Answer-first structure puts the direct answer to a question in the first sentence under a heading that is the question.; Engines lift first sentences. Synthic's optimizer rewrites headings and FAQ blocks into this shape from the page's own content.; See alsoAEO JSON-LD
AI crawler: An AI crawler is a bot that fetches web pages to feed an answer engine — for training, for a search index, or live while a person waits for an answer.; Synthic tracks sixteen, classed as training, search or live, and checks each one's access against your robots.txt.; See alsoLive user agent robots.txt
Live user agent: A live user agent (ChatGPT-User, Claude-User, Perplexity-User) is a fetch made on behalf of a person in the middle of a conversation.; A hit from one in your logs means your page was just pulled into an answer. Agent Analytics separates these from index and training traffic.; See alsoAI crawler
robots.txt: robots.txt is the file at a site's root that tells crawlers which paths they may fetch, per user agent.; A blanket Disallow written for one bot commonly blocks every AI agent by accident. Synthic evaluates it live and shows the exact rule that matched.; See alsoAI crawler
Buyer persona: A buyer persona is the lens an audit prompt is phrased through — general user, in-market buyer, startup founder, developer or enterprise evaluator.; The phrasing changes the engine's recommendation, so Synthic can run the same keyword set through each lens and compare.; See alsoPrompt set
Prompt set: A prompt set is a fixed list of questions re-run across engines over time — the GEO equivalent of a rank tracker's keyword list.; Movement between runs is movement in what the engines said, never in how the question was asked.; See alsoBuyer persona
Zero-click: A zero-click query is one resolved on the results page or inside an answer, with no visit to any site.; Where GEO wins and classic SEO cannot reach: being the brand named inside the answer is the whole prize.; See alsoGEO
Citation graph: The citation graph is the map of which pages and domains engines actually pull from when they answer questions in a category.; Synthic aggregates it from stored citations across your audits — Perplexity's native sources and URLs surfaced in every other engine's answers.; See alsoCitation rate

Now measure it on your own brand.

Run a live audit and see exactly where you stand — before your competitor does.

Start free audit See pricing

✓Free first audit✓No credit card✓Live engine data✓Your data stays yours✓Cancel anytime