Core concepts
The objects and evidence model used throughout the Sight workspace.
Sight is organized around durable evidence objects. Organizations control access and billing, projects represent a website and brand, and scans capture a time-bounded set of crawl and AI-answer work. Prompts, answers, sources, actions, and reports remain connected to that project and scan context.
Understanding these relationships prevents common reporting mistakes. A metric is never independent of its answer set, an issue is never independent of its affected URLs, and a recommendation should retain the evidence that produced it. Use the definitions below when comparing dashboard views, exports, API responses, or client reports.
The same object can appear in several workflows without becoming a different record. A source cited by an answer can later support a gap, an action, and a report block, while still retaining the original answer and scan relationship. Following those links is the safest way to understand why a recommendation exists and whether later work changed the underlying evidence.
| Concept | Meaning | Where you use it |
|---|---|---|
| Organization | The billing and access boundary for users, projects, API keys, and webhooks. | Account and plans |
| Project | One website and brand workspace. The app sometimes calls this a site. | Project switcher and project navigation |
| Scan | A time-bounded collection of crawl, SEO, performance, and GEO work. | Scan history and scan detail |
| Prompt | A question sent to selected AI surfaces to measure how a brand appears. | GEO Prompts |
| AI answer | A stored response, including brand mentions, position, sentiment, features, and source citations. | AI answers and GEO overview |
| Source | A domain or URL retrieved or cited by an AI answer. | Domains, URLs, and gap analysis |
| Action | An evidence-backed opportunity that can be accepted, dismissed, tracked, and reviewed. | Earned, Owned, and Impact |
| Report | An editable, shareable narrative assembled from current scan evidence. | Reports and report builder |
Evidence first
Recommendations should always lead back to the page, response, citation, check, or measurement that produced them.
Evidence should remain inspectable after a metric is summarized or a recommendation is edited. Preserve the original scan, affected URL, answer, source, captured value, and filters. If a claim cannot be traced to one of those records, label it as interpretation or business context rather than a measured Sight result.
- SEO findings link to audited URLs and explain the practical fix.
- GEO metrics are computed from stored model responses and their citations.
- Source gaps include competitor evidence before they become actions.
- Reports should preserve the source date, scan, project, and selected filters.