Patnick
Sign In
Demand Intelligence · Deep Dive

Query Clustering.

Google clusters queries internally. If your strategy treats them as separate keywords, you're fighting the algorithm — not with it.

What is it?

Query Clustering, defined.

Query Clustering is the process of grouping semantically related search queries into thematic clusters using sentence embeddings (384-dim vectors via sentence-transformers), HDBSCAN density-based clustering, and Claude-generated labels — producing a topical map that mirrors Google's internal query-network representation and aligns directly with the pillar + cluster content architecture that semantic SEO research recommends for maximum topical authority.

Google's internal models cluster queries semantically; your strategy should too. Patnick groups your query universe into 5-15 thematic clusters using sentence embeddings, then labels each cluster via Claude. The output is a topical map — exactly the structure semantic SEO research recommends for pillar + cluster content architecture.

Why it matters

Four concrete outcomes.

Sentence embeddings

Queries are embedded into vector space using a dedicated embedding model (not a full LLM). Fast and cheap.

Automatic cluster labels

Each cluster gets a human-readable label generated by Claude based on the member queries.

5-15 themes per site

Patnick targets 5-15 clusters per site — enough granularity for strategy, not so many you drown in themes.

Drives topical authority

Clusters map directly to pillar + cluster page architecture for topical authority SEO.

How it works

The 4-step process.

  1. 01

    Embed queries

    Every query in the universe is converted to a 384-dimensional vector via sentence-transformers.

  2. 02

    Compute similarity matrix

    Pairwise cosine similarity between all query vectors.

  3. 03

    Cluster

    HDBSCAN density-based clustering groups related queries into 5-15 thematic clusters.

  4. 04

    Label with Claude

    Each cluster's top 10 queries are passed to Claude, which generates a descriptive theme label.

Inside Patnick

See it in the dashboard.

This is how query clustering surfaces inside the real Patnick dashboard. Enter the your audit to click through it.

patnick.com/dashboard
SEO toolsSearch visibilitySchemaTrends
Explore more

Related capabilities in Demand Intelligence.

Google Search Console Integration

Historical engagement data is a first-party signal you can't fake. Your GSC already has it — Patnick just connects the wiring.

Read→

Google Trends Signals

Contextual search is real search. Momentum is the signal that says 'this query is about to matter'.

Read→

Keyword Volume

Volume matters, but only as one input. The 30-website lexical study proves that semantic variation carries most of the real demand.

Read→
People also ask

Frequently asked questions.

What is query clustering?
Query clustering is the process of grouping semantically similar search queries together so you can develop strategy at the topic level rather than the keyword level. Instead of tracking 500 individual queries, you track 10 clusters, each containing the related queries.
Why cluster queries instead of targeting them individually?
Because Google's algorithm already clusters them internally. When you publish content targeting 'best running shoes', Google's ranking systems will also match it against 'top running shoe brands', 'best sneakers for runners', etc. If your strategy treats these as separate keywords, you'll waste effort and miss the aggregate opportunity.
What technique does Patnick use for clustering?
Sentence embeddings (all-MiniLM-L6-v2 from sentence-transformers) to convert each query into a 384-dimensional vector, then HDBSCAN density-based clustering to group them. HDBSCAN handles clusters of varying density and automatically determines the optimal number of clusters — no need to pick K in advance.
How are cluster labels generated?
For each cluster, Patnick takes the top 10 queries by frequency and passes them to Claude with a prompt asking for a descriptive 2-4 word label. The result is human-readable like 'Running Shoe Recommendations' or 'AI SEO Tool Comparisons', not cryptic category IDs.
Can I manually override cluster assignments?
Yes. The dashboard has a cluster editor where you can reassign queries, merge clusters, split clusters, or rename labels. Manual overrides persist across re-clustering runs — Patnick respects your decisions.
How does clustering feed topical authority SEO?
Each cluster becomes a candidate for a pillar page on your site, and the member queries become cluster articles underneath it. This mirrors Google's recommended topical authority architecture (pillar + cluster model). Patnick surfaces cluster data explicitly so you can build pillar pages aligned with your actual user demand.

SEO Implementation

$499/mo

I build the roadmap. Your writer executes.

Start →

Full SEO Management

Most Value

$799/mo

I handle everything. You focus on your business.

Let me handle it →

14-day money-back guarantee. 6-month commitment recommended.

See it live.

Log into the demo dashboard and click any block to learn exactly what it does.

See Our Plans→← Back to Demand Intelligence