Skip to content
Docs/AI Agent

Topical Maps

A topical map clusters your site's GSC queries and crawled pages into topic pillars that an LLM identifies. It's the visual answer to "what am I actually about, and what's missing?". Maps are persistent (saved to your tenant DB), editable (rename clusters, reassign pages, add gap topics), and can auto-refresh on a cadence.

How a map gets built

  1. Pull the last N days of GSC queries for the site (the seed range you choose).
  2. An LLM clusters the queries into named topic pillars with descriptions, ratings (strong / moderate / weak), and insights.
  3. Each crawled page is auto-assigned to the closest cluster via cosine similarity over the embeddings.
  4. Gap topics are surfaced per cluster: queries the LLM thinks the site should cover but doesn't yet rank for.

You can regenerate a map any time, or set an auto-refresh cadence (every 7 / 14 / 30 / 90 days). Each regeneration creates a new map row alongside the old one. Your history stays intact.

What you can do on /topical-maps

  • Browse clusters - Each cluster shows its name, authority rating, query and page counts, top queries, assigned pages with similarity scores, and gap topics.
  • Edit - Rename clusters, set a pillar page (the page that "owns" the topic), reassign pages, add custom gap topics, dismiss gaps you've handled.
  • Regenerate - Pull a fresh map from current GSC data (creates a new map version; old maps stay accessible).

Agent integration

The agent has list_topical_maps, get_topical_map, get_cluster_context, and regenerate_topical_map tools. Common questions it can now answer from real data:

  • "What are my strongest topic clusters?"
  • "Which clusters have the most gap topics?"
  • "What's my pillar page for keyword X?"
  • "Where does my page on Y fit in my topical map?"

Per-tier access

Available on every tier; auto-refresh is Pro+ (Starter creates maps manually).