Topic clusters
The single structure that keywords, tracked prompts and content all hang from in RankX AI. What a good cluster looks like, and why it is worth getting roughly right.
A topic cluster is a subject your business wants to be known for. It is RankX AI's single source of truth for how keywords, tracked prompts and content group together, which makes it the join between the three halves of the product: what you research, what you measure, and what you publish.
Get them roughly right rather than exactly right. Clusters are editable, nothing is lost by renaming one, and the cost of not having them is much higher than the cost of imperfect ones.
What hangs off a cluster
| Object | How it uses the cluster |
|---|---|
| Tracked keywords | Attached to a cluster, so ranking movement can be read by subject rather than one query at a time |
| AI visibility prompts | Attached to a cluster, so visibility can be read by subject too |
| Content briefs and articles | Written against a cluster, which is where the brief's supporting keywords come from |
The payoff is that all three become comparable. "We rank well on this subject, the assistants never name us on it, and we have published nothing about it" is a sentence you can only construct if one structure spans all three, and it is usually the most useful sentence available.
Where they come from
RankX AI generates a candidate set during onboarding from your brand profile and your competitor set, which takes under a minute. That set is a starting point, and it is better than starting from a blank screen, but it is generated from what your site says rather than from what you sell.
Editing it is expected. Prune the clusters that describe the industry rather than your business, and add the ones you sell into that your site does not talk about yet, which are usually the most valuable.
What a good cluster looks like
A subject, not a keyword. "AI visibility tracking" is a cluster. "best ai visibility tracking tool 2026" is a keyword that belongs in it.
A subject you sell into. A cluster you have no offer for produces prompts you cannot win and briefs nobody should write.
Distinguishable from its siblings. If two clusters would take the same keywords and the same prompts, they are one cluster with two names, and having both makes every by-cluster reading useless.
Named the way you would say it aloud, not the way the category describes itself. The generated set skews to industry vocabulary because it reads industry pages.
Why cluster structure is not editable from an assistant
There is no MCP tool that creates a cluster, deliberately.
Cluster structure is a decision about how the business describes itself, and it is the one place where an agent generating "something reasonable" produces lasting mess: every keyword, prompt and article filed under an invented cluster inherits the invention. Attaching an existing keyword to an existing cluster is a tool, because that is bookkeeping. Deciding what the clusters are is not.
Reading by cluster
Once keywords and prompts are attached, the readings worth taking are:
Where you rank but are not named. Good search presence and no assistant mentions on the same subject usually means the assistants cannot reach or read those pages. Start with AI Readiness.
Where you are named but do not rank. The assistants know you for this and Google does not show you. That is often a genuine content gap on a subject you already have authority for, which is the cheapest kind to fix.
Where you have neither. A cluster with no ranking and no mentions and no published content is not a failure, it is an unstarted piece of work.
Where you have content and neither. The most useful bad news in the product: you have published and it has not landed. Read the AI Chat Feed and AI Answer Citations for that cluster's prompts and see who is being named instead.
Keeping them current
A reasonable rhythm: review clusters when your offer changes, not on a schedule. Renaming one is free and non-destructive; splitting one that has grown two distinct halves is worth doing the moment the by-cluster reading stops being useful.
Pausing the prompts under a cluster that no longer matters is better than deleting the cluster, because the history stays readable.
Driving this from an assistant
list_topic_clusters reads them with their linked counts, and
link_keyword_to_cluster attaches a tracked keyword.
"Show me my topic clusters with how many keywords and prompts each has, and tell me which ones have keywords but no prompts."
See the tool reference.
Where to go next
- Content briefs, written against a cluster.
- AI Prompts, which attach to one.
- Keyword research, for filling them.
Keyword research
Discovering keywords in RankX AI, what each figure means, how the cache makes repeats free, and how discovery feeds rank tracking and topic clusters.
Content briefs
Turning a keyword into a researched brief in RankX AI. What a brief contains, why the research step is free, and what approving one means.