Reference
Metrics reference
The canonical definition of every number in Analyze AI — visibility, sentiment, position, narrative strength, authority, and the Risk Score — including how each is calculated.
How measurement works
Every metric below is computed from the same base data: your tracked prompts, run on a recurring schedule against the AI engines, with each answer parsed for brand mentions, rank, sentiment, and citations. Nothing is estimated or modeled from third-party panels — if a number exists, real AI answers produced it.
Metrics are windowed (7, 14, 30, or 90 days, or a custom range) and can be filtered by engine and by brand throughout the app.
Visibility
The percentage of AI answers that mention your brand. If your prompts produced 100 answers in the window and 62 mentioned you, visibility is 62%.
- Counted per answer, not per prompt: a prompt that runs daily contributes every answer to the window.
- Filterable per engine — the blended number is your headline; per-engine visibility is your diagnosis.
- Mentions are entity-resolved (name variants and product names count), so casual phrasing doesn't slip through.
Position
Your average rank inside the answers that mention you. When AI lists or orders brands, being named first matters; position averages where you appear. Lower is better — #1 means you're typically the first recommendation. A brand can have high visibility and poor position: mentioned everywhere, recommended first nowhere.
Sentiment
How positively AI talks about your brand, 0–100, averaged over the answers that mention you. Each mention is scored in context, with the supporting text span recorded — so any score can be traced to the exact sentence that produced it. As a rule of thumb: 60+ reads positive, 40 and under reads negative, between is neutral. Scores come with the reasoning preserved, which is why sentiment in Analyze AI is quotable, not just chartable.
Mentions & share of voice
Mentions count how many times each brand appears across the window's answers — the raw material behind visibility. Share of voice compares visibility across every tracked brand per engine: your slice of the conversation, engine by engine.
Narrative strength
How strong your story is when AI tells it, 0–100. Perception analysis reads every answer for the themes attached to each brand and how favorably, specifically, and consistently each brand is framed. High visibility with low narrative strength is the "known but mixed" pattern — mentioned often, described weakly. This is the vertical axis of the perception quadrant.
Authority score
How much weight a cited source carries, 0–100. Estimated from domain strength and organic footprint (traffic and ranking keywords). A mention on one authority-85 page typically does more for your AI visibility than ten mentions on authority-20 blogs. Shown for every URL in citation intelligence.
Risk Score
Analyze AI's own measure of how much competitive threat a cited source represents, 0–100. This is not a spam or quality flag — it answers a sharper question: of all the pages AI engines are using in my space, which ones are actively working against me?
Five weighted components:
| Component | Max | What it measures |
|---|---|---|
| Authority & traffic | 25 | Domain strength plus the page's organic traffic and keyword footprint — how powerful the page is |
| Citation impact | 25 | How many engines cite it, and how often — how much AI actually leans on it |
| Competitive stance | 25 | How many competitors are framed positively on the page, and how favorably |
| Coverage gap | 15 | Whether you're absent (or weakly framed) while competitors shine there |
| Recency & repetition | 10 | Whether its citations are accelerating right now |
Bands: below 35 is low, 35–64 is medium, 65 and above is high.
Read it as a priority list: a high-risk source is a powerful page that engines trust, that presents competitors well, where you're invisible — and it's gaining momentum. That's the page to win a mention on, counter with better content, or watch. Full workflow in source authority & risk.
Content score
The Content Optimizer's 0–100 editorial score for a page, split into Argument & Flow (does the piece make a case, in the right order) and Clarity & Polish (sentence-level quality). Rewrites show the before/after score, and a rework only earns QA Passed after link checks and claim-by-claim verification — see optimizing a page.
Traffic metrics
AI Traffic Analytics reports standard analytics measures — sessions, engagement rate, bounce rate, session duration, conversions — restricted to sessions referred by AI platforms, as recorded by your connected Google Analytics 4 property (setup). Assistants don't always pass referrer data, so treat AI-traffic numbers as a floor.
Answers vary run to run — that's the nature of AI engines, and it's why every metric here is presented as a windowed trend rather than a single reading. Judge direction and level together, and be suspicious of any conclusion drawn from one day.