Glossary
All important terms and concepts around AI Answer Intelligence — clearly explained.
AI Answer Intelligence
The systematic capture, analysis, and evaluation of AI-generated answers about brands, products, and topics. Encompasses visibility, accuracy, evidence, and action recommendations.
Visibility-Scan
The public unit of consumption for AI answer monitoring: 1 Visibility-Scan = 1 monitored question per run, regardless of how many AI providers your plan includes. A run with 12 active questions consumes 12 Visibility-Scans.
Website-Scan
A heuristic, AI-free crawl of your website covering structure, metadata, crawlability, and citability. Website-Scans have their own monthly quota and currently do not call any AI provider.
Competitor benchmarking
Tracking which competitors get mentioned, cited, or described more favorably in AI answers — and which sources support those answers. Surfaces who is recommended instead of your brand.
Presence
Whether a brand is mentioned by name in an AI answer. Measured through deterministic string matching against configured brand names and aliases.
Citation
When an AI system includes a URL as a source link in its answer. Owned citations are particularly valuable as they signal authority and generate traffic.
Snapshot
A complete record of an AI answer at a specific point in time: answer text, citations, metadata, and raw payload. Snapshots are the primary evidence foundation.
Finding
A structured result of deterministic analysis of a snapshot: brand mention detected, owned citation present, competitor mentioned, etc. Findings are versioned and traceable.
Diagnostic
An actionable issue derived from findings and historical comparisons: e.g., brand absent, owned citation missing, new competitor appeared.
Severity
The urgency of a diagnostic, calculated from the combination of issue type and question priority. Levels: critical, high, medium, low.
Diff
The comparison between two consecutive snapshots of the same question and provider. Shows text, citation, and finding changes.
Critical Facts
User-defined key facts about the brand (product names, prices, core claims). The system detects when these facts are missing or divergent in AI answers.
Review Debt
The number of unreviewed diagnostics that require human review. High review debt indicates insufficient monitoring attention.
Recommendation Confidence
The confidence in a recommendation based on the quality and quantity of underlying evidence. Promptalyze clearly distinguishes between deterministic and AI-derived results.
Answer Stability
How consistently an AI answer stays the same over time. Frequent changes may indicate unstable sources or contested topics.
Collection Run
A data collection where all configured questions for a project are sent to selected AI providers. Can be triggered manually or on a schedule.
Provider Adapter
A standardized interface through which Promptalyze communicates with different AI providers. Each adapter implements the same contract and normalizes provider-specific responses.