Glossary
Plain definitions of every word Trakkr uses, in the sense the product uses it, from Accuracy to Work type.
Every word below is used here the way the product uses it, which is not always the way the industry uses it. Where a term has a formula behind it, the entry links to Metrics. If a word you met in the app is missing, email [email protected] and it gets added.
A
Accuracy
The surface that checks what AI models say about your brand's facts and grades each claim. Rows read correct, wanders, unverified, wrong point-blank, wrong every way or not checked. Learn more →
Actions
Trakkr's desk for work and its measured outcomes. Work moves through Suggested, This week, Queued and In progress; once it is done, its result moves through Measuring to a verdict. Done says the work happened, not that it earned anything. Learn more →
Active prompt
A prompt Trakkr checks daily and that counts toward your plan limit. Pause one to stop the checks without losing its history.
Agency mode
The agency workspace: portfolio views, the Pitches generator, and the white-label portal. It comes with Scale. Learn more →
Agent
Trakkr's in-app AI assistant. Ask questions about your data in plain English and get answers grounded in your real numbers. Permission to prepare or make changes is granted per work area and is off by default. Learn more →
AI crawler
Any AI provider's bot that fetches pages from your site. Trakkr recognizes dozens of them, including GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot and ChatGPT-User.
AI model
A system that generates text answers. Trakkr tracks eight: ChatGPT, Claude, Gemini, Perplexity, DeepSeek, Grok, Meta AI and AI Overviews.
AI Overviews
Google's AI-generated summary above the search results. One of the three surfaces Trakkr reads citations from, alongside Perplexity and ChatGPT Search.
AI Pages
Trakkr's proxy layer that serves AI-optimized HTML to crawlers without changing your real site. Detect, transform, serve, track. Learn more →
AI visibility
How often and how prominently your brand appears in AI-generated answers. The thing the product exists to measure.
AI volume
The estimated number of times people ask AI platforms about a topic each month, shown with a confidence tier. See the formula →
Alias
An alternative name for your brand, like "AWS" for "Amazon Web Services". Aliases improve detection in AI answers.
Audience
A buyer segment such as "Enterprise IT Buyer", classified from your prompts. It unlocks once a brand has 10 or more active prompts. Learn more →
Automation
A standing instruction that checks something for you and responds within the permission you set. It can tell you, suggest work, prepare a draft, or make an allowed change. Learn more →
Automation rung
The permission level an Automation holds in one work area. The five rungs are Tells you, Suggest only, Prepare drafts, Prepare for approval and Handle it.
Available
The first Journey stage. It passes when the page returns a successful response and allows indexing.
Average position
Where you typically sit in a list-style AI answer, averaged across the answers that mentioned you. Lower is better.
B
Bottleneck
The Journey stage that loses the largest share of measured pages, shown at site level. Relevant is inferred, so it is never named as the bottleneck.
Brand
A company, product or entity you track. Each brand carries its own prompts, scores, citations, competitors, aliases and history, and nothing crosses between brands by default.
Brand group
A color-coded grouping for organizing many brands, used most often by agencies to bucket brands by client.
Branded prompt
A prompt that names your own brand, like "Is Stripe good for SaaS?". Keep prompts category-level instead, so the model has to choose whether to recommend you.
C
Check
One time an Automation evaluates its rule and records what it found or did. A check can end with nothing new, a suggestion, or something that Needs you.
Checks score
The 0 to 100 summary of the checks in Optimize, shown as a quiet chip because the findings and the fix queue matter more than the headline. See the detail →
Citation
A URL an AI model returns as a source for its answer. Trakkr captures citations from Perplexity, ChatGPT Search and Google AI Overviews. Learn more →
Citation feed
A chronological log of citation events: New citation, Lost citation, Sentiment changed and Competitor appeared. Read it weekly on the 7-day window.
Citation gap
A source where competitors are cited and you are not. Closing gaps is some of the highest-return work available.
Citation sentiment
Whether a citing page discusses your brand positively, neutrally or negatively, judged with your brand as the subject. See the detail →
Client portal
The read-only, agency-branded experience for white-label clients, served on your own domain. Different from a one-off shared report link.
Coincided with a drop
A result used when a measured metric fell during the window and cleared every guard. It reports timing, never cause.
Competitive gap
The difference in visibility points between you and one competitor. Positive means you are ahead.
Competitor
Another brand AI puts next to yours when answering your tracked prompts. Trakkr discovers competitors from real answers, so there is no Add button. Learn more →
Conversations
Crawler requests that matched a user-triggered fetcher such as ChatGPT-User or Claude-User, meaning a live answer was being assembled while your page was read. See the detail →
Couldn't measure
A result used when Trakkr lacked the page-level data to judge the work. It is a first-class answer, not an error, and it is excluded from Earn rate.
Crawlers
The Traffic page where every AI bot visit is logged and grouped into Total Visits, Conversations, Agents, Indexing and Training. Learn more →
D
Demand score
How likely people are to ask a given question in AI chat, on a 0 to 100 scale. See the formula →
Diagnose
A single-prompt deep dive. It asks one query of the live models and writes a brief explaining why the answer looks the way it does. Learn more →
Discovery prompt
A prompt about finding a solution, like "What's the best tool for X?". These reveal whether you are in the consideration set.
Domain Rating
Ahrefs' 0 to 100 estimate of a domain's backlink authority, shown wherever a citing domain appears. A supporting signal, not a ranking. See the detail →
E
Earn rate
The share of judged results that Earned: earned divided by earned plus no movement. Couldn't measure is excluded and always shown beside it. See the formula →
Earned
A result used when at least one measured metric cleared the bar fixed at the start of the Measurement window. It means the metric moved, not that the work caused it.
Earned Media
News and editorial citations. One of the eight source types, and usually the class worth chasing hardest.
Estimated
A stage state based on inference rather than direct observation. Relevant is always marked Estimated.
Exact rule
The deterministic layer inside Automations: when a stated trigger matches, perform the stated response. Learn more →
Extra prompts
The add-on that raises the active prompt limit by 50, 100 or 200, also called a prompt pack. It applies to all your brands, and the price is per brand owned.
F
G
Gap prompt
A prompt where competitors appear and you do not. Strong candidates for Suggested actions.
I
Ideas
The prompt brainstorming tool at the bottom of the Prompts page. Three modes: brand, tag and keyword.
Impact
In Outreach, a 0 to 100 measure of how much tracked-prompt visibility a publisher is costing you: how many rivals it cites, across how many of your prompts, weighted by how often AI answers pull the page. The default order in every Outreach lens. See Impact.
Indexing
Crawler requests from search-class bots such as OAI-SearchBot and PerplexityBot. Access can support discovery, but a request never predicts indexing or citation.
J
Journey
The six stages a page moves through: Available, Reached, Understood, Relevant, Selected and Visited.
L
llms.txt
A convention for telling AI crawlers which content to read first. It sits at /llms.txt, like a sitemap written for models rather than search engines.
M
Measurement window
The fixed before-and-after period used to judge an Action, from 14 days for a technical fix to 70 for a campaign. It is frozen when the work starts, so the bar cannot move afterwards. See the table →
Measuring
The result state after work is done and before its Measurement window closes.
Mention
An AI model including your brand name in an answer. Mentions can be positive, neutral or negative.
Model performance
How you score on one AI model. Models differ enough that 80 on Claude and 30 on Perplexity is normal.
N
Narrative
A topic you monitor across models over time, with keywords, an optional comparison brand, and a 0 to 100 score per model. Used to track corrective messaging.
Needs you
Work an Automation cannot finish without your approval, answer, safety review or decision. Nothing in this state proceeds until you respond.
No movement
A result used when Trakkr had the data to judge the work, but no measured metric cleared the earned bar or the guarded drop threshold.
Not measured
A Journey stage state used when the source is not connected or the check has not run. It does not mean the page failed that stage.
O
Optimize
Trakkr's site checking surface. Its Findings view carries a verdict, a fix queue and the full check list. Learn more →
Outreach
A working queue built from citation gaps, ranked by Fit and Difficulty, with generated pitch angles and a status per target. Learn more →
Overall perception
Your headline perception number: a 0 to 100 score across 20 attributes in five categories. See the formula →
P
Page
A URL Trakkr tracks through the six-stage Journey. Learn more →
Page record
The full view of one Page: its Journey, bottleneck, current work, measurements, history and Results.
Perception
How AI describes your brand's qualities across 20 attributes in five categories: trust, quality, value, market position and innovation. Learn more →
Phrasing analysis
For one prompt, Trakkr asks a set of semantic variants and shows where you rank on each. It surfaces wording blind spots, where you win on one phrasing and lose on another.
Pitch
A branded report an agency generates for a prospect, built from real visibility data before the prospect is a customer. Learn more →
Position
Your rank in a list-style AI answer. Position 1 means you were named first.
Presence rate
The share of tracked prompt and model requests where you appear at all, whatever the position. Breadth, not depth. See the formula →
Prompt
A question or request you track. Prompts simulate what real users ask AI assistants, and they are the unit AI visibility is measured in.
Q
Quick Wins
The Outreach quadrant holding high-fit, low-difficulty targets. Start there.
R
RAG (retrieval-augmented generation)
A technique where a model searches the web or a database before answering. Perplexity leans on it heavily.
Reached
The second Journey stage. It passes when AI crawlers fetched the page at least five times in 30 days, including at least one search-type crawler.
Reddit monitoring
Finds Reddit threads discussing your brand or competitors, scores each for citation potential, and helps you draft graded replies. Learn more →
Relevant
The fourth Journey stage. It estimates whether the page's detected queries match prompts you track, so it is always marked Estimated.
Report
A snapshot of your visibility data at a point in time, used for history and comparison.
Research
The discovery tool that generates category prompts, asks each one once, and surfaces which are worth tracking. A one-shot scan, not a daily tracker. Learn more →
Results
The measured outcomes of finished Actions. It keeps Measuring, Earned, No movement, Couldn't measure, Reverted and Coincided with a drop separate from whether the work was done. Learn more →
Reverted
A result used when a change was put back after its window coincided with a guarded drop. It does not claim the change caused the drop.
Roles
Team members hold one of three roles: Owner (account creator, full control), Admin (full data and team access) and Viewer (read-only).
S
Selected
The fifth Journey stage. It passes when an AI answer cited the page in the last 30 days.
Share of voice
Your visibility as a proportion of total visibility across tracked competitors, shown as two donuts: Recommended First and Mention Share. See the formula →
Shared report link
A public URL exposing a read-only snapshot of your dashboard. Different from the white-label portal, which is a full authenticated client experience.
Source type
How Trakkr classifies a citing domain: Owned, Earned Media, Reviews, Social, Institution, PR Wire, Competition or Other. See the table →
Suggested
An Action Trakkr found in your data that nobody has committed to yet. Suggested actions stay separate from planned work and Results.
T
Tag
A label you apply to prompts. Tags let you filter and score prompt groups by intent, topic, competitor, audience or market.
Topic snapshot
A narrowed Research check on one slice of your market. Unlike the main Research scan, which is replaced each time, snapshots persist.
Training
Crawler requests from bots documented or classified for training or bulk collection, such as GPTBot and ClaudeBot. A matching request never proves the content entered a dataset or a model.
Training data
The content models learn from during training. Being present in it is foundational for visibility on models that do not retrieve live.
Trend
The direction a metric is moving. The 30-day trend is usually a better signal than 7-day movement. See the windows →
U
Understood
The third Journey stage. It passes when a non-rendering bot gets the same useful text a browser gets.
Unbranded prompt
A category-level question that does not name your brand, like "Best CRM for B2B SaaS". The shape Trakkr is built around.
V
Verdict
A plain-language judgment based on measured evidence. A page verdict names the first Journey stage that needs work; a result verdict says what moved without claiming causation.
Visibility score
How prominently AI models mention you, weighted by position and divided by every successful response in the slice. Your headline metric. See the formula →
Visited
The sixth Journey stage. It passes when people arrive from an AI answer, or when connected search data shows clicks above the brand's page median.
Visitors
The Traffic page showing humans who landed on your site from an AI referrer, read from your connected Google Analytics 4 or PostHog property. Learn more →
W
White-label portal
A client portal on your own domain with your branding, used by agencies to give clients an ongoing experience instead of one-off links. Learn more →
Win rate
How often you rank above one competitor in the answers where you both appear. A head-to-head measure, not a measure of leadership overall. See the formula →
Work type
The kind of Action being done, which sets its Measurement window. The types are Technical fix, Page improvement, Content refresh, New content, Outreach draft, Reply draft, Campaign, Setup and Watch.