Editorial plan
Articles linked to intent, keyword, market, language and CTA. Checks first whether it is better to update an existing page.
Complete list by area. Items that depend on an external service report the real connection status inside the app.
From planning to publication, with persistent versions and human approval.
Articles linked to intent, keyword, market, language and CTA. Checks first whether it is better to update an existing page.
Workspace timezone, rescheduling, priority, pause and resume. A job already started is not duplicated after a refresh.
Research → outline → draft → media and sources → competitor review → revision → polish → final checks → approval → publication.
Headings, lists, tables, links, images, allowed video, undo/redo, autosave and version history.
On a selection or the whole article, with a diff preview and the option to undo. No silent overwrite of human work.
Off by default, enabled per site. Blocks unverified critical facts, insufficient sources and failed checks.
Measurements with numerator, denominator and coverage always visible. What is unavailable stays N/A.
Neutral prompts and branded queries kept separate. The exact text, version and localisation method are stored for every observation.
Mention rate, share of voice, citation share, average position and category rank: numerator, denominator, engines included and coverage always shown.
Full answer, actual citations, sources only retrieved, observed queries and AI suggestions kept distinct.
Clicks, impressions, CTR and position by page, query, country and device, with the data delay declared by Google.
Comparison of content, queries and cited sources, with the origin of the suggestion.
If prompts, engines or sample change, the warning is explicit and the comparison is limited to the stable subset.
Questions where you are not visible, with the rationale next to every priority.
Manual and CSV import, AI candidates, provider enrichment. Variant deduplication and clustering by intent.
Editorial opportunities with reason, URL, brands already present, suggested action and known or unknown cost. A link is “verified” only after checking the real redirect and HTML.
Relevant threads, useful reply drafts, per-channel mode (draft only, approval, automatic with rules, paused). Default: approval required.
HTTP status, redirects, canonical, sitemap, robots, titles and descriptions, heading hierarchy, broken links, orphan pages and heavy images. Performance only if actually measured.
Suggestions based on relevance and real pages, with destination and anchor checks.
Isolation between customers, durable jobs and programmatic access with explicit permissions.
Owner, administrator, editor, viewer. Every query is checked on the server: isolation between customers is tested.
Queue on PostgreSQL, job lease, retry with backoff and jitter, resume after crash. Closing the tab does not stop the work.
Answers from authorised workspace data and links every number to the source record. No shell or arbitrary SQL access.
Versioned API with scoped tokens, rate limit, revocation and audit. Read-only MCP on summaries, keywords, content, citations and activity.
Areas that depend on an external provider stay declared as not connected until you configure it: no sample numbers in place of a measurement.