The keyword clustering tool that turns keyword lists into a content plan

Keyword clustering is grouping keywords that one page could rank for, so you write one strong page per group instead of one thin page per keyword. Two searches belong together when the searcher wants the same thing and would be happy with the same answer.

Clustering usually happens at two levels. The first groups keywords into pages. The second groups related pages into a topic: a hub page for the broad search and spoke articles for the specific ones, each linking back to the hub. LogNorm does both and keeps one topic list per website, so keywords, articles, AI visibility prompts and moves all point at the same topics.

LogNorm clusters by asking, pair by pair, whether one page could rank for both searches: same intent, same answer. Keywords that pass form a page with a main (pillar) keyword, and then every member is checked against that pillar.

That second check matters. Simple clustering can chain keywords together: A is close to B, B is close to C, and soon C sits on a page where it doesn't belong. Checking each member against the pillar stops a run of loosely related keywords from drifting into one page.

1. 1
   Join existing topics

   A keyword that clearly fits a topic you already keep joins it. A keyword that fits a hidden topic stays out of every topic.

2. 2
   Group into pages

   The rest are paired by intent and answer, grouped into pages, and checked against each page's pillar keyword.

3. 3
   Group pages into topics

   Pages are paired again: would one pillar page naturally link to both as supporting articles? Pages that pass form a topic.

4. 4
   Pick the hub

   The hub keyword is chosen from the pages' pillars and names the topic. Every other keyword in the topic is a spoke.

Grouping runs on LogNorm's own decision model. If it isn't available, an AI model from your connected providers groups the keywords and a second pass re-checks the groups.

Keywords your buyers wouldn't search stay out. Before grouping, LogNorm rates every keyword for relevance, intent, format and buyer fit against your business profile, and that read decides who takes part.

• Off-topic keywords, competitors' own brand searches and keywords rated tangential to your business never join a topic.
• Searches for your own brand go into one navigational topic.
• Keywords you skipped stay out of clusters for good.
• Keywords you targeted always take part, because you chose them.

A page that joins nothing becomes a topic of its own only if it has at least two keywords and real search volume, or includes a keyword you targeted. At most 15 new topics are created per run, highest volume first; the remaining keywords stay unclustered and still appear in Discover. See keyword research for how each keyword is judged.

A topic holds up to 10 pages and 60 keywords, with one hub keyword. The cap keeps a topic small enough for one hub page to cover and for a team to finish.

Yes, once a person has touched them. A topic becomes locked the moment someone creates it, renames it, edits its keywords, picks a new hub keyword, keeps it or hides it, and research never replaces or re-clusters a locked topic. It can only add new spoke keywords.

Each run rebuilds only the untouched topics it created before. An active research topic is reused when two or more of its keywords still fit, and keeps its name and status. Hidden topics, topics you made, topics AI suggested and topics that articles or AI prompts are filed under are left alone. Once you have curated your list, new research topics arrive as Suggested for you to keep or hide, so your plan never changes under you.

Yes, in three ways, and you don't have to wait for research. Any topic you make is locked from the start.

• New topic cluster: give it a name, a hub keyword, an intent, an optional reason it matters and the keywords in it. Those keywords are targeted, and LogNorm fetches their volume and difficulty.
• New cluster from selection: select keywords in Discover and make them a cluster. The keyword with the most searches becomes the hub.
• Suggest topic clusters: LogNorm proposes 3 to 10 new topics from your product, buyers, competitors and researched keywords that aren't in a topic yet, with real search data. This needs an AI model and uses credits.

Inside a topic, each keyword row lets you target, skip, make it the hub, move it to another cluster or remove it. Details are in Topics and clusters.

Build hub and spoke plan on a topic writes the plan: one hub page and typically 3 to 10 spoke articles. The topic needs at least two keywords that aren't skipped, and the plan runs in the background.

Each card names the keyword it targets plus close variants, its format (guide, how-to, comparison, listicle, template, glossary, landing page or case study), searches per month, difficulty, why it's worth writing, an angle and up to three competitor pages to beat. A card marked Rework existing page means you already have a page ranking for it, so improve that page instead of writing a new one. Under the plan, LogNorm lists the competitor pages that rank for the topic's keywords, grouped by competitor with the themes each covers.

Plan article, or Plan all, turns cards into ideas in your blog planner, with target keyword, topic and angle, scheduled into the next free slots at your publishing cadence.

The plan creates moves for up to eight of its articles, ranked against everything else in your backlog. A keyword list on its own doesn't tell you what to write this week; a ranked move does.

Articles built on targeted keywords get higher confidence, and those built on unreviewed suggestions get lower confidence. Articles for keywords you already rank 1 to 3 for get no move, and hidden topics produce none. When a move becomes a brief, LogNorm pulls the live Google results for its keyword and collects the cluster's keywords for the article. A connected Claude Code or Codex agent can also build topic plans and write the articles, and its work shows on the dashboard under its name. See the AI SEO agent page.