Glossary

Project: the workspace for one brand. Website, brand settings, competitors, monitors, prompts, and analytics all live inside a project.
Monitor: a panel of prompts that runs on a repeating schedule across your chosen AI models, scoped to one country and language. Multiple markets means multiple monitors.
Prompt: a question sent to AI models inside a monitor, like a keyword in classic SEO but phrased the way a person asks. Each prompt has a type, an optional intent, and can carry tags and topics.
Response: one model’s answer to one prompt on one scheduled run. The response stores the full answer text plus everything analysis extracts from it: mentions, citations, sentiment, enrichments.
Brand: your own brand or a competitor, tracked at the project level. Competitors are set once per project and apply to every monitor in it.

Mention: your brand discussed in the text of a response. One response counts as one mention per brand, however often the name appears. See Citations vs. mentions.
Citation: a source URL the AI attached to its answer. Recorded with its domain and position among the answer’s sources.
Visibility score: 0-100 for how prominently your brand appears in a response, based on placement, share of attention, and depth. The period average counts responses without your brand as zero, so it reflects both prominence and coverage. See Visibility Score.
Sentiment score: 0-100 for how positively the response discusses your brand, scored independently of how visible it is.
Share of Voice: your share of brand appearances measured against your competitors: your mentions divided by the total mentions across the compared brands. See Share of Voice.
Position in answer: where your brand first substantively appears in a response, counted from 1. Being the first brand named is worth more than being the ninth.
Citation rank: the position of your domain among an answer’s cited sources, counted from 1. Lower is better.
Self-citation: a citation pointing at your own domain, as opposed to third-party pages.

Prompt Type: Organic (no brand named, tests pure category visibility), Brand Specific (your brand named), or Competitor Comparison (head to head). Filtering by type keeps branded prompts from inflating your category numbers.
Intent Type: where the prompt sits in the buyer journey: Informational, Branded, Navigational, Commercial, or Transactional.
Tag: a project-scoped label you attach to prompts for filtering and grouping, with an optional color.
Topic: a project-scoped theme label attached to prompts, used to group visibility by subject area.

Query fanout: the sub-queries an AI model searched while composing its answer. Fanouts show what the model actually looked up, which often differs from the prompt you asked.
Persona: an audience profile attached to a monitor so prompts run from that buyer’s perspective. The persona wraps the prompt at run time; your stored prompt text stays unchanged.
Content slot: one scheduled piece in the Content Agent’s content schedule, moving from planned through generating and review to published.
Action item: a recommendation on the Action Board (Actions in the left rail), drawn from content gaps, sentiment issues, offsite opportunities, and site health. Moves from Suggested through To do and In progress to Done.
Enrichments: structured extras a model returned alongside its answer: images, shopping results, tables, local businesses, news, or ads.

Crawler logs: HTTP requests AI bots (GPTBot, ClaudeBot, PerplexityBot, and others) make to your website, captured through your CDN. They show what AI companies actually read on your site.
AI visitor analytics: site visits that arrived from an AI assistant, detected by referrer. The human traffic your AI visibility produces.
Sitemap: your site’s structure as Promptwatch understands it, the foundation for content recommendations. See Sitemap or Search Console.

Model (or platform): one AI answer engine Promptwatch can track: ChatGPT, AI Overview, AI Mode, Copilot, Gemini, Perplexity, Grok, Claude, Alexa, DeepSeek, Mistral, and more. Models are selected per monitor.
Live search: models that search the web while answering and attach sources to their answers. Only live-search models produce citations reliably.