Atlassian Apps / Documentation

Documentation for Confluence AI Hygiene.

This page is the public documentation URL for AI Readiness & Content Hygiene Inspector for Confluence. It explains where the app appears, how the scan and review flow works, how the policy presets behave, and where the boundaries are.

  • Last updated: April 15, 2026
  • Confluence Cloud only
  • Space admin workflow

Requirements

  • Atlassian Confluence Cloud
  • an active Marketplace subscription where licensing data is present
  • Confluence space administrator access for the space being reviewed

Where The App Appears

The app is opened from Confluence Space Settings in the space being reviewed. The workflow is space-scoped rather than a site-wide dashboard.

The current user interface presents a scan status area, findings tabs, and a review policy form in the same app view.

Scan Workflow

  1. Open the app from the target Confluence space.
  2. Review the current policy preset or tune the policy fields.
  3. Start a scan.
  4. Let the scan continue in the background while progress is persisted.
  5. Review findings by status and decide whether to ignore, approve, or reopen each page.

The app can continue background work after the settings screen is closed. If a scan encounters retryable rate limits, the worker retries. If a scan fails, the scan status is marked as failed and the app shows a user-facing failure message.

Signals The App Uses

The current scoring model can use signals such as:

  • Stale content
  • Thin content
  • Outdated or draft markers
  • Weak titles
  • Obvious placeholder pages
  • Macro-heavy stubs
  • Deprecated or archive labels

Findings are scored and assigned a severity. The current UI exposes scan summary cards for pages scanned, no issues detected, needs review, ignored, and approved, plus a top-signal summary and recommendation per flagged page.

Findings Tabs And Review States

The current UI exposes four findings tabs:

  • Needs review for open findings
  • Ignored
  • Approved
  • All findings

The review actions currently available in the UI are Ignore, Approve, and Reopen.

Ignoring a page applies the rovo-ignore label. Approving or reopening removes that label through the app workflow.

No issues detected means a page did not cross the current policy threshold. It is not the same as an explicit Approved review state.

Review Policy

The policy form currently allows admins to set:

  • preset, with the shipped presets Balanced, Strict, and Project Space
  • stale threshold in days
  • minimum plain-text characters
  • minimum score to open a finding
  • whether to detect outdated or draft markers
  • whether to detect weak titles
  • whether to detect macro-heavy stubs
Preset Description Stale days Min text chars Min score
Balanced Good default for most handbook and team spaces. 365 300 45
Strict Aggressive review posture for high-trust knowledge bases. 180 500 35
Project Space Tuned for fast-moving project documentation. 120 200 40

Retention Notes

Category Data Retention
Policy settings Per-space preset selection and review thresholds such as stale days, minimum text length, and minimum score. Retained until changed by the customer or removed with the app installation data.
Per-page review decisions Decision status and update timestamps needed for admin review history. Retained until changed by the customer or removed with the app installation data.
Scan artifacts and findings Scan metadata, checkpoints, findings, statuses, and derived review records used to continue and review background scans. Configured with a 30-day Forge TTL in the current implementation.
App installation data after uninstall Forge-hosted persistent storage associated with the installation. Retained by Atlassian Forge for a limited post-uninstall period documented by Atlassian as up to 28 days, with relink possible if requested in time.

Important Boundaries

  • Only Confluence space administrators can use the app.
  • The app validates that label actions target pages in the current space before applying them.
  • rovo-ignore is a governance convention, not a documented Atlassian-native Rovo exclusion control.
  • No issues detected means not flagged under the current policy, not human-approved.
  • Do not assume every action appears in native Confluence audit logs in the same way.