Getting started
Concepts
A glossary of the terms used across Rixe — from run and sample to share of voice, citation, market pack and fan-out.
Rixe uses a small set of terms consistently on every screen, in reports and in the API. This page defines each one in plain words, grouped by what it describes. When a term has its own page, the definition links to it.
What is measured
Engine
An AI answer engine Rixe asks questions on — for example ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI Mode, Microsoft Copilot, Claude or Grok. Eleven engines are measured today. Your plan decides how many are measured for you; your market decides which ones. See Engines and markets.
Search-grounded engine
An engine that runs a live web search before answering and reports the pages it used. ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI Mode, Microsoft Copilot and Grok work this way. Only these engines produce citations, and only these count toward the composite visibility score.
Chat model
An engine that answers from what it learned in training, without searching the web — DeepSeek, Mistral, Claude and Qwen. Their answers are still measured and shown per engine, but they never cite sources and they do not count toward the composite score, because a model without search rarely names a niche or local brand however well it ranks.
Prompt (question)
One question a customer might ask an AI engine, such as "best CRM for a small agency". Prompts are what Rixe asks on your behalf. In the Turkish interface they are called soru. See Prompts.
Prompt type
Whether a prompt names your brand. A Category prompt does not ("best CRM for a small team"); a Brand prompt does ("is Acme CRM good?"). Category prompts show who gets discovered; brand prompts show what engines say about you. The type is worked out from the prompt's text.
Intent
The buyer-journey stage a prompt belongs to: Awareness, Consideration, Evaluation or Purchase. Every prompt is classified automatically; you can override it. Engines rarely name brands on awareness questions and often do on purchase ones, so a healthy set covers every stage.
Topic
A group of related prompts, such as pricing or integrations. Topics are proposed during setup and let you filter and compare results by subject.
Persona
An optional audience label you can assign to prompts (for example SMB owner or Marketing manager) to filter results by who is asking.
Conversation (chain)
Two or three follow-up questions asked as one conversation on each engine that supports it — "best CRM?", then "for a small team?", then "which is cheapest?". See Conversations.
Fan-out
The searches an engine runs on its own before answering a prompt. When you ask about "best CRM", the engine may search for "CRM pricing comparison" and "CRM for startups" behind the scenes. Those searches show you questions the engine already asks about your category. Only engines that disclose their searches are counted; the others are marked as not disclosing, never guessed. Shown on the Engine searches tab. See Coverage and fan-out.
Coverage
How evenly your prompt set covers the buyer-journey stages and your topics, so that part of your category does not go unmeasured without anyone noticing.
How it is measured
Run
One prompt asked once on one engine. A run either completes (and produces an answer) or fails (and is recorded with the reason).
Answer
What the engine returned for a run, stored in full together with the pages it cited. Every number in Rixe can be traced back to answers.
Sample
The number of answers a figure is based on. It is shown next to rates — as Runs in tables or as "1,240 answers · 48 prompts · 3 engines" under a page title. A rate on 12 answers and the same rate on 400 answers do not mean the same thing.
Not enough data
What Rixe shows instead of a number when a rate rests on fewer than 4 runs. Some screens say insufficient sample. It clears on its own as daily measurement adds answers.
Confidence interval
The range a rate's true value probably falls in, shown next to the rate ("likely between 18% and 34%"). Rixe uses the Wilson interval at 95% confidence, which stays accurate on the small samples where the product makes its claims. Narrow when the sample is large, wide when it is small. When two intervals overlap, the difference between them is not statistically meaningful.
Significant change
A change Rixe reports as real: both periods have at least 4 runs, the difference is statistically significant (p < 0.05), and it is at least 10 percentage points.
Missing data
Runs that were planned but could not be completed — an engine was down, timed out or refused the request. They are left out of rates rather than counted as "not mentioned", and the screen says so with a data missing badge or a banner. A broken engine never quietly lowers a score.
Pooled answer
An answer shared between workspaces that asked exactly the same question, in the same country, on the same engine within 24 hours. Marked pooled on the prompt's page. It is still a real answer from that engine.
Daily measurement
Rixe plans measurements every day at 02:00 UTC on every plan. See How measurement works.
What the numbers mean
Mention
An answer names a brand. Matching follows your market's language rules, and brand aliases count as the brand.
Mention rate (Visibility)
The share of answers that name a brand: answers naming the brand ÷ answers collected. Shown as Visibility in tables. A visibility of 25% means the brand was named in one of every four answers.
Position
Where a brand appears in an answer: 1 means it was named first. Shown as the average position across the answers that named the brand. Being named first matters because readers stop early.
Citation
An engine lists a page from your own domain as one of its sources. Being named is good; being named and used as the source is better. Only search-grounded engines cite.
Source
A website or page an engine cited in an answer. The Sources view shows which sources the engines rely on for your questions and whether your brand appears on them. See Sources.
Share of voice
A brand's mentions as a share of all brand mentions in the same answers. If your brand was named 30 times and all brands together 200 times, your share of voice is 15%. Unlike visibility, shares of voice across brands add up to 100%.
Sentiment
How an answer portrays a brand: positive, neutral or negative.
Recommended
Of the answers that named a brand, how many actually recommended it. Being mentioned and being recommended are not the same thing.
Visibility score
The single 0–100 number on the AI visibility dial. It combines mention rate, citation rate and position across the search-grounded engines, weighted by how much each engine is used in your market. See Visibility score.
Unbranded
A filter that recalculates results from category prompts only, leaving out prompts that name your brand — so you see who the engines reach for when they are not led by a name.
Difficulty and volatility
Two labels on each prompt. Difficulty says how winnable it is: Open ground (nobody wins it), Entrenched (one brand owns it), Rotating (brands take turns) or Contested (several brands in play); it needs at least 10 answers. Volatility says how often the result changes between runs: Sticky, Settled or Unstable.
Drop-off
In a conversation, the share of conversations that named your brand at the first step but no longer at the last.
How your account is organised
Market
A country together with a language, such as Türkiye · Turkish or Canada · French. A market is measured on its own questions, in its own language. Markets are shown side by side, never averaged together.
Market pack
The data that describes one market: which engines matter there and how much (engine weights), the language rules used to recognise brand names, the sources that feed AI answers in that country, and more. Rixe ships 98 market packs across 48 languages. A new language is a new pack, not new software.
Engine weight
How much an engine counts in a market, taken from the market pack. It decides which engines your plan measures and how much each one contributes to the visibility score.
Workspace
One brand measured in one market. Each market you open is its own workspace, with its own prompts, competitors and results. Your plan sets how many workspaces you can have.
Competitor
A brand you track alongside your own, so every rate can be compared. Your plan sets how many competitors each workspace can track. See Competitors.
Credit
On agency plans, the unit measurement is paid in: one credit pays for one measurement of one prompt on one engine. See Credits.
Related pages
Still stuck? Write to [email protected] and a person will answer.