Topic Clusters

Topic Clusters is an AI-powered feature that groups semantically related search queries together, allowing you to analyze their collective performance and identify content opportunities at scale. By leveraging the Embedding Database, this feature goes beyond simple keyword matching to understand the actual meaning and intent behind queries.

Why Use Topic Clusters?

Traditional keyword analysis looks at queries individually, missing the bigger picture of how related searches contribute to your overall performance. Topic Clusters solves this by:

• Revealing Hidden Patterns: Discover groups of queries that share semantic meaning but use different words

• Measuring Topic Performance: See aggregated metrics (clicks, impressions) for entire topic areas

• Identifying Content Gaps: Find clusters of queries where you're underperforming

• Streamlining Optimization: Focus on improving entire topics rather than individual keywords

• Understanding User Intent: Group queries by what users actually mean, not just what they type

Prerequisites

Before you can create topic clusters, you need to set up the embedding system:

Step 1: Enable Embedding Database

1. Navigate to Settings → Embedding in the left sidebar

2. Toggle on "Enable Embeddings" master switch

3. Under Google Search Console Queries, ensure it's enabled

4. Select an embedding model (see Choosing the Right Model)

5. Click "Update settings"

Step 2: Generate Query Embeddings

1. Go to Google Search Console > Properties in the left sidebar

2. Select your property

3. Navigate to the Settings tab

1. Scroll down to the "Query Embeddings" section

2. Click "Generate Embeddings" to process all your queries

The embedding process runs in the background. Processing time depends on:

• Number of queries (1,000 queries ≈ 1-2 minutes)

• Selected model (local models are slower but free)

• Your computer's specifications

Creating Topic Clusters

Once embeddings are generated, you can start creating clusters:

Step 1: Access Topic Clusters

1. Navigate to your GSC property's Insights page

2. Scroll down to the Topic Clusters card

3. Click "Create Cluster" to begin

Step 2: Search for Related Queries

1. Enter Topics: Type one or more seed topics to find similar queries

   • Single topic: Finds queries similar to that specific topic

   • Multiple topics: Finds queries similar to the centroid (average) of all topics

2. Set Similarity Threshold: Adjust the threshold to control how closely queries must match

   • Higher values (0.8-0.9): Stricter matching, fewer but more relevant results

   • Lower values (0.6-0.7): Broader matching, more results with varying relevance

   • See Understanding Similarity Scores for details

3. Click Search: SEO Utils will return all semantically related queries

Step 3: Select Queries

Review the search results and select queries to include in your cluster:

• Individual Selection: Click checkboxes for specific queries

• Select All on Page: Use the header checkbox to select visible queries

• Select All Results: Click "Select all X queries" to include all matching queries across pages

• Review Similarity Scores: Higher scores indicate stronger semantic relationships

Step 4: Configure Cluster Details

After selecting queries, click "Next" to configure your cluster:

1. Cluster Name: Give your cluster a descriptive name (e.g., "SEO Best Practices", "Local Coffee Shops")

2. Description (Optional): Add notes about what this cluster represents

3. Color: Choose a color for visual identification in charts and reports

Click "Create Cluster" to save your new topic cluster.

Managing Topic Clusters

Viewing Cluster Performance

Once created, clusters appear in the Topic Clusters card, showing:

• Total Clicks: Aggregated clicks from all queries in the cluster

• Total Impressions: Combined visibility across all cluster queries

Editing Clusters

Click on any cluster to:

• Add/Remove Queries: Refine your cluster by adjusting included queries

• Update Details: Change name, description, or color

• Re-run Search: Find new related queries with a different threshold

Deleting Clusters

To remove a cluster:

1. Click on the cluster to open edit mode

2. Click the "Delete" button

3. Confirm deletion (this only removes the cluster, not the underlying queries)

Exporting Cluster Data

Export your clusters for further analysis:

• Download CSV: Click the download button to export cluster metrics

• PDF Reports: Clusters appear in exported PDF Insights reports

Advanced Features

Multi-Model Flexibility

The embedding database stores vectors from different models separately, allowing you to:

• Experiment with Models: Try different embedding models to find the best for your content

• Compare Results: See how different models group your queries

• Switch Without Loss: Change models without losing previously generated embeddings

To switch models:

1. Go to Settings → Embedding

2. Select a different model for Google Search Console Queries

3. Generate new embeddings with the selected model

4. Create clusters using the new embeddings

Managing Embeddings

Control your embedding data from the GSC Settings page:

• View Status: See how many queries have embeddings

• Delete Embeddings: Remove embeddings for a specific model to free space or start fresh

Best Practices

Choosing Seed Topics

• Be Specific: "coffee brewing methods" works better than just "coffee"

• Use Natural Language: Write topics as users would search

• Combine Related Terms: Use multiple seeds to capture topic variations

• Consider Intent: Mix informational, commercial, and navigational terms

Setting Similarity Thresholds

Start with these recommended thresholds:

• Tight Clusters (0.85-0.95): For very specific topics or branded queries

• Balanced Clusters (0.75-0.85): Good for most content topics

• Broad Clusters (0.65-0.75): For exploratory analysis or finding opportunities

Organizing Clusters

• Avoid Overlap: Check that queries don't appear in multiple similar clusters

• Create Hierarchies: Build parent topics with broader themes, child clusters for specifics

• Use Consistent Naming: Develop a naming convention for easy identification

• Document Purpose: Use descriptions to explain each cluster's optimization goal

Use Cases

Content Gap Analysis

1. Create clusters for topics you want to rank for

2. Identify clusters with high impressions but low clicks

3. Analyze which queries need better content

4. Develop content strategies for entire topic areas

Competitor Comparison

1. Build clusters around competitor brand terms

2. Find topics where competitors are mentioned

3. Identify opportunities to create comparison content

4. Track performance improvements over time

Seasonal Planning

1. Create clusters for seasonal topics

2. Monitor performance trends throughout the year

3. Plan content calendars based on cluster seasonality

4. Optimize existing content before peak seasons

User Intent Mapping

1. Group queries by search intent (informational, transactional, navigational)

2. Ensure content matches the dominant intent in each cluster

3. Identify intent gaps in your content strategy

4. Optimize conversion paths for each intent type

Troubleshooting

No Queries Found

• Lower the threshold: Try 0.6-0.7 for broader matching

• Use different seed topics: Try synonyms or related terms

• Check embeddings: Ensure queries have been embedded with the current model

Too Many Irrelevant Queries

• Increase threshold: Use 0.85+ for stricter matching

• Refine seed topics: Be more specific with your search terms

• Manual curation: Remove irrelevant queries after initial search

Embeddings Not Generating

• Check model configuration: Ensure the embedding model is properly selected

• Verify API keys: For cloud models, check API key validity

• Local model issues: Ensure Ollama is running and the model is downloaded