> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fixaeo.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Glossary
> Every FixAEO term defined once, with the page that owns it.
Twenty-odd words, each used in one specific way.
A glossary entry here tells you what FixAEO means by a term, not what the industry means by it. Where a word is contested — "citation" is the worst offender — this page states the product's definition and links the page that owns it. Start with the first pair. Most confusion in AI search reporting traces back to those two words.
## Mention and citation
These are the load-bearing pair. They're counted separately because they're separate events, and a brand can have one without the other.
### Mention
A mention is your brand named in the text of an AI answer. No link required. It counts when the answer names you by brand name or by domain. Mentions are what your [Visibility Score](/metrics/visibility-score) is built from. You read the raw ones on the Mentions tab — see [Understanding AI answers](/understanding-answers).
### Citation
A citation is a link to a website that the engine put on screen. A citation chip, a footnote, or a row in the sources panel. If a reader could see it and click it, it's a citation. A citation to your domain is not a mention of your brand, and the two are tracked on different pages. See [Citations](/citations).
All four combinations happen. Mentioned and cited. Mentioned but not cited. Cited but not mentioned. Neither. Read the two numbers side by side to see which one you're in.
## The metrics
Four headline numbers, plus one that lives on the Sources page. [Metrics overview](/metrics-overview) covers how they fit together.
### Visibility Score
The percentage of captured AI answers that mention your brand. This is the core metric and the headline number on your [dashboard](/dashboard). [Understanding AI answers](/understanding-answers) covers where those answers come from. See [Visibility Score](/metrics/visibility-score).
### Share of voice
Your share of every brand mention in the window, yours and your rivals' together. It's the only headline metric that accounts for what competitors are doing. See [Share of voice](/metrics/share-of-voice).
### Position
How early an answer names you, averaged across the answers that named you. `#1.8` means you're named second on average. Lower is better. See [Position](/metrics/position).
### Sentiment
How an answer describes you once it names you, scored positive, neutral, or negative. Only mentions are classified, so an answer that ignored you can't drag it down. See [Sentiment](/metrics/sentiment).
### Cite rate
A per-domain number on [Sources](/sources): when an engine pulls a domain, how often it actually shows it on screen.
A low cite rate on your own domain means engines read your site and rarely credit it. That's a different problem from not being read at all.
## What you set up
The inputs you control. Get these right and everything downstream follows.
### Prompt
One question FixAEO asks the AI engines on your behalf, in your brand's region. Your tracked prompt list is the denominator behind every metric, so it decides what you're measuring. See [Setting up your prompts](/setting-up-prompts).
### Prompt pool
Your plan's prompt allowance, shared across every brand on the account rather than granted per brand. On Lite you get 15 prompts in total, not 15 each. This is the limit people misread most often. See [Plans and limits](/plans-and-limits).
### Topic
A theme your prompts get grouped into. FixAEO uses the word twice. Setup topics are the ones you pick to steer prompt generation. Topic clusters are what an AI model derives from your live prompt list. The two drift apart over time, on purpose. See [Topics and labels](/topics-and-labels).
### Label
A tag you define and attach to prompts yourself, then filter by. Unlike topic clusters, labels never regroup themselves. There are two kinds, tags and personas. See [Topics and labels](/topics-and-labels).
### Persona
The second kind of label, meant for the audience behind a prompt rather than its subject matter — "startup founder" instead of "pricing". A prompt can carry any number of both. See [Topics and labels](/topics-and-labels).
### Brand profile
The structured description of your brand that FixAEO stores and reuses. It holds what you do, your key products, your factual claims, your differentiators, your target audience, and your tone. It grounds prompt generation and the [Improve](/improve) recommendations. See [Your brand profile](/brand-profile).
### Competitor
A rival brand, either detected automatically in an answer or pinned by you. Pinned competitors appear in the ranking even at zero mentions, so you can watch one before it shows up. See [Tracking competitors](/competitors).
## How is the data collected?
The measurement machinery. [Understanding AI answers](/understanding-answers) is the long version.
### Engine
One AI product FixAEO asks your prompts. There are nine: ChatGPT, Claude, Gemini, Perplexity, Copilot, DeepSeek, Grok, AI Overviews, and AI Mode. Which ones run for you is a plan limit — paid tiers below Enterprise get a six-engine set. See [Plans and limits](/plans-and-limits).
### Model channel
One engine viewed as a distribution channel for your brand. The Model page reports Visibility Score per channel, so you can see that Perplexity runs fifteen points above AI Overviews on an identical prompt list. See [Model](/model).
### Region
The country a prompt is asked in. Every answer FixAEO stores carries the region it came from. It has to. The same prompt run for the US and for Germany can name entirely different brands. New brands default to the United States, in English. See [Your brand profile](/brand-profile).
### Scan
One pass of your tracked prompts across your plan's engines, for your brand's region. Scans are scheduled by plan, and free accounts get none. See [Plans and limits](/plans-and-limits).
### Snapshot
One day's aggregated numbers for one brand and region — mentions, answers captured, sentiment counts, and the per-engine split. Snapshots are what the trend chart on your [dashboard](/dashboard) reads. A day with no captured answers shows as a gap in the chart rather than a false zero.
### Fan-out
The sub-questions an engine implicitly answers inside one complex query. Ask for "best CRM for 10-person SaaS startups" and the engine is really answering four or five smaller questions, each backed by different sources. See [Query fan-outs](/fanouts).
Fan-outs are a best guess, not ground truth. Most engines don't expose their reasoning, so FixAEO infers the breakdown. Treat it as a strong hint about why a competitor won a sub-question, not as a transcript.
### Network source
A URL the engine's page fetched in the background while drafting and then never displayed. It's the complement of a citation: everything an engine read, minus what it put on screen.
The gap between the two is the difference between what an engine considered and what it was willing to show. See [Sources](/sources).
## What counts as traffic from AI?
Two different things arrive at your site from AI, and they need separate tools.
### Agent traffic
Requests from AI crawlers, browsing agents, and MCP clients — GPTBot, ClaudeBot, and the rest. These don't run JavaScript, so tag-based analytics can't see them at all. FixAEO reads them from your server logs or a tag, depending on plan. See [Agent analytics](/agent-analytics).
### Referral traffic
Humans who read an AI answer and then clicked through to your site. These are real sessions and normal analytics does see them. FixAEO reads them from your Google Analytics connection. See [Attribution](/attribution).
When someone asks "how much traffic is AI sending us", find out which of these two they mean. Agent traffic is bots reading you. Referral traffic is people arriving. They move independently, and a healthy crawl rate with flat referrals is a real and common pattern.
Where the answers come from, and how often they're collected.
The four numbers and which one to watch first.
The raw answers every term on this page is counted from.
The mention-versus-citation split in full.