Documentation

Sources & Citations

The Sources tab shows where AI assistants pull their evidence from when answering your prompts. It combines normalized citations with prompt metadata so you can audit coverage across channels.

Filters

  • Prompt – Focus on a single prompt or view all prompts together.
  • Root domain – Limit results to a specific domain. Useful for competitive deep dives or owned-property audits.
  • Window – 7, 14, 30, 60, or 90-day lookbacks. Defaults to 30 days.
  • Domain limit – Number of domains to highlight in the leaderboard (default 5, max 25).
  • Page size – Controls pagination size for the prompt run table.

Filters sync to the URL. Share the link with stakeholders to give them the same view.

Overview section

  • Total mentions – Count of citations within the window.
  • Distinct domains – How many unique domains AI assistants referenced.
  • Owned share – Percent of mentions pointing to your tracked domains.
  • Sentiment change – Difference in sentiment velocity compared to the previous window.
  • Window summary – Displays the exact date range the data covers.

Domain leaderboard

  • Ranks domains by total mentions and shows daily sparkline trends.
  • Use search to filter for specific domains or keywords.
  • Sorting options include mentions, trend, and alphabetical.
  • Each domain card surfaces its share of voice percentage.

Prompt run table

  • Lists the individual runs contributing to the selected filters.
  • Columns include prompt label, AI platform, answer excerpt, cited domain, and timestamps.
  • Click Inspect prompt to open the prompt editor with the run preselected.
  • Pagination uses keyset cursors; the “load more” button fetches additional rows without re-running the entire query.

Exporting & refreshing

  • The Refresh button re-pulls data on demand. A spinner shows when new data is being fetched.
  • Copy table rows for quick exports. For larger data pulls, contact support@chatobserver.com.

Troubleshooting

  • Empty state – Appears if no prompts ran during the selected window or if all runs failed analysis. Shorten the window or re-run prompts.
  • Overview error – If the top cards fail to load, check your filters or try again; short-lived cache issues usually resolve quickly.
  • Prompt unavailable – A run may reference a deleted prompt. Reactivate or recreate the prompt if you need ongoing coverage.

Best practices

  • Pair this view with the Prompts tab to ensure each strategic question has the citations you expect.
  • Add top competitor domains to your tracked list so the leaderboard highlights them automatically.
  • Use smaller windows (7–14 days) during launches to gauge immediate impact, then zoom out to 60–90 days for long-term trends.