Skip to main content
Use these definitions when reading Qwairy dashboards and this documentation.

Monitoring vocabulary

  • Prompt: a natural-language question tracked across one or more AI models.
  • Topic: a thematic group assigned to a prompt. A prompt has one topic or none.
  • Tag: a reusable label. A prompt can have multiple tags.
  • Funnel stage: the prompt’s position in the buyer journey: TOFU (discovery), MOFU (evaluation), BOFU (conversion), or Post-purchase.
  • Provider: the company or product family behind an AI experience, such as OpenAI or Google.
  • Model: the specific AI experience or model selected for generation, such as ChatGPT (Core) or GPT-6.1 Sol.
  • Country: the market in which a prompt runs. A workspace has a default country and can list additional countries. A prompt runs in the default country unless you select other countries for it.
  • Persona: an optional audience profile assigned to prompts. A prompt runs once for each assigned persona.
  • Answer: the output captured for one prompt, model, country, and persona in a run.
  • Monitoring frequency: how often a prompt is scheduled. A prompt may inherit the workspace default or override it with Daily, Weekly, Monthly, or No monitoring.
  • Boost: additional answers generated for the same prompt and model in one run. A boost can be set from 2 to 20 answers, and the workspace limits how many prompts can be boosted.
  • Credits: the unit consumed by answer generation and selected product operations. Creating or importing prompt definitions is free; each generated answer costs the credits of its model. See Providers and credits.

Brand relationships

Qwairy detects the brands mentioned in answers and classifies each one by its relationship to the monitored brand:
  • Self: the monitored brand.
  • Direct: a direct competitor. Direct competitors enter competitive metrics.
  • Indirect: a related brand that is not treated as a direct competitor. Indirect brands appear under Related Brands in Monitor > Competitor Mentions and do not enter competitive metrics.
You can also mark a brand as a favorite or ignore it. Ignored brands are excluded from competitive analysis. In formulas, exports, and the API, these relationships appear as SELF, DIRECT, and INDIRECT. Review these classifications when a metric looks unexpected.

Mentions and citations

A mention is an occurrence of a classified brand in an answer. A citation is a detected source associated with an answer. A link in the answer and a citation used for analysis are related signals, but they are not interchangeable on every page.

Overview metrics

All formulas below use the active date range and filters. If a denominator is zero, the interface displays the empty or zero state defined by that view.
The REST performance endpoint has its own Coverage definition: answers with a SELF mention divided by all answers. Do not use that REST formula to reproduce the prompt-level Coverage card on the dashboard.

Scope and comparability

Metrics change with provider, country, topic, tag, funnel, persona, prompt, list, and date filters. Compare periods only after checking that the scope, models, countries, personas, and prompt set are equivalent. For the source data behind a metric, use Response Analysis, Competitor Mentions, and Source Explorer.