Glossary — Model Research

Scores & Metrics

Authority Score: The headline composite metric measuring how AI language models perceive and recommend a brand. Combines mention rate, sentiment, citation quality, and cross-model consistency. Called 'Model IQ Score' in Phase 2.
Confidence Tier: Indicates the statistical reliability of a score. Directional (grey) = single observation from a Quick run. Reliable (blue) = 3+ repeats from a Standard run. Verified (green) = 10+ repeats from a Deep run.
Mention Rate: Percentage of AI model answers that mention a specific brand when asked about a topic. Higher = more visible.
Share of Voice: A brand's mention count as a proportion of all brand mentions in the same run. Shows competitive standing.
Avg Sentiment: Mean sentiment score across all sentences mentioning a brand. Range: -1 (very negative) to +1 (very positive).
Citation Share: Percentage of all source citations that reference a specific domain or URL. Higher = more authoritative in AI training data.
R / T / I Scores: Relevance (how relevant the source is to the answer), Trust (how trustworthy the source is), Influence (how much the source shaped the answer). Each scored 1-10.

Visibility: How often and prominently AI models mention a brand. Derived from mention rate, cross-model rate, and topic breadth.
Reputation: How positively AI models describe a brand. Derived from sentiment, recommendation language, and trust signals.
Distinction: How uniquely AI models position a brand vs competitors. Derived from differentiation mentions and unique attributes.
Endorsement: How strongly AI models recommend a brand. Derived from recommendation rate and citation authority.
Risk: Negative signals AI models associate with a brand. Higher = more risk. Inverted in composite: a low risk score is good.
Composite Score: Overall brand intelligence score: (Visibility + Reputation + Distinction + Endorsement + (100 - Risk)) / 5.
Quadrant: 2x2 positioning: Brand Signal (Visibility + Distinction) × Brand Substance (Reputation + Endorsement), with Risk overlay.
Momentum: Trend direction of composite score: Rising, Stable, Declining, or New (no historical data).

Quick Run: Single pass, 1 model, ~25 prompts. Produces Directional confidence. Good for rapid checks.
Standard Run: 3 repeat passes, up to 2 models, ~50 prompts. Produces Reliable confidence. The default for most analysis.
Deep Run: 10+ repeat passes, all models, ~200 prompts. Produces Verified confidence. Pro tier only.
Boost: Upgrade an existing run's confidence tier by adding more repeat passes. 1 run = Reliable, 2 runs = Verified (Pro only).

Watchlist: A monitored topic area (e.g. 'AI in Healthcare'). Contains threads, alerts, and intelligence feed items.
Topic: An organisational grouping within a watchlist. Contains threads.
Thread: A monitoring intent within a topic. Defines what to track and generates prompts for periodic AI model queries.
Thread Run: A single execution of a thread's prompts against AI models. Captures visibility, sentiment, and mention data.
KOL: Key Opinion Leader — an influential person identified in research content for a watchlist topic.
Alert: An automated scan that runs on a schedule (daily/weekly) to discover new content matching a watchlist's search queries.
Insight: An AI-generated analysis article synthesising recent intelligence from a watchlist.

Citation Health: Overall score (0-100) measuring the quality of sources AI models cite. Combines diversity, authority, coverage, and consistency.
Source Diversity: How evenly citations are distributed across domains. Measured using Shannon diversity index.
Authority Tier: Domain authority classification: Elite (8+ avg R/T/I), High (6+), Medium (4+), Low (below 4).
Cross-Model Consistency: Percentage of cited domains that appear in answers from multiple AI models. Higher = more reliable signal.
Domain Category: Classification of a cited domain: professional, news, reference, blog, social, review, documentation, etc. 15+ categories.

Wilson Confidence Interval: Statistical method used to calculate reliable confidence bounds on mention rates, accounting for sample size.
PSR: Prompt Success Rate — percentage of prompts where a brand was mentioned at least once.
Statistical Significance: Whether observed differences between brands are likely real vs due to random variation. Based on sample size and effect magnitude.