Documentation

Prompts Overview

Prompts are the heart of Chatobserver. Each prompt defines an instruction that is sent to multiple AI assistants (ChatGPT, Perplexity, Gemini) on the cadence you choose. Chatobserver captures the raw answers, normalizes citations, and updates your dashboards automatically.

Lifecycle

  1. Draft – Create or duplicate a prompt. The editor validates required fields before allowing you to save.
  2. Active – The prompt appears in the active tab and follows its schedule. Active prompts consume credits whenever they run.
  3. Inactive – Temporarily pause scheduled runs without losing history. Manual runs are still allowed.
  4. Archived – Remove prompts from day-to-day views while preserving historical runs. Archive is reversible from the inactive tab.

Key fields

  • Label & excerpt – Displayed in tables, filters, and notification emails.
  • Prompt text – The exact instruction delivered to each AI channel.
  • Target platforms – Choose one or many. Chatobserver parallelizes runs per platform.
  • Schedule – Determines when Chatobserver refreshes the prompt. See Scheduling & Cadence Controls for details.
  • Analysis policy – Toggles enrichment pipelines such as ranking and sentiment scoring.
  • Personas & groups – Metadata for segmenting results and sharing context.
  • Visibility – Private or team-wide.

Collaborating on prompts

  • Owners and admins can edit any prompt. Members can edit prompts they created or have been granted access to through groups.
  • Each edit writes to the audit log. Revert by reopening the editor and restoring previous values.
  • Use the “duplicate prompt” action to create variants for different personas or regions without starting from scratch.

Monitoring health

  • Status pills in the Prompts table show run health (pending, queued, running, complete, failed, credit hold).
  • Next run indicates if a prompt is scheduled. Prompts with manual cadence display “Run manually.”
  • Last run links directly to the Runs tab with the prompt pre-filtered.

Deleting prompts

Deleting removes the prompt and its historical runs. This action cannot be undone and should be reserved for sandbox or test data. Archiving is the safer option for production prompts.