About Services Tools Blog Contact us

Search Query Topic Grouping

Groups a Google Search Console spreadsheet into named topics, so a content plan follows demand instead of a long list of near-duplicate queries.

  • Query Clustering & SEO
  • Cross-sector
  • Pipeline
  • Productised
  • Local/private
  • 2026

The problem

A raw query export is thousands of near-duplicate lines. Nobody can see which topics deserve a page, which guide types are hiding in the list, or which queries are only noise.

What it does

The tool turns each query into a numerical fingerprint of its meaning and saves it, then groups similar queries. Because the fingerprints are saved, changing the settings does not repeat the slow step. Gemma names each group and writes a short brief with clicks and impressions, and queries that fit nowhere go to a separate file. The same job has been run on a retail spreadsheet and on an equipment spreadsheet.

Outcomes

  • Thousands of queries collapse into a set of named topic briefs.
  • Clicks and impressions stay attached, so the plan follows demand.
  • Each query's meaning is saved, so changing the settings does not repeat the slow step.
  • The same process works on a retail spreadsheet or a services spreadsheet.

Simplified mockup

Interactive preview with sample data. Nothing is sent to a server.

Query export

Queries
1,842
18 shown below
Impressions
86,410
clicks stay attached
Vector cache
Cold
first embed writes it
Sample lines
QueryClicksImpr.Pos.
children shoe size chart4206,1008.2
how to measure a child’s foot2103,4006.1
child shoe length in cm881,90011.4
toddler shoe sizes641,5009.8
waterproof walking boots3105,2007.4
boots for wet grass1202,8009.1
country boots for mud481,40010.5
wide fit school shoes1904,1005.8
school shoes wide width771,8008.4
uk to eu shoe size2403,9004.2
shoe size converter1802,7003.9
half size shoes explained511,10015.1
how to polish leather shoes952,1008.8
leather boot care4090011.2
shoes128,00022.0
near me890018.4
sale152,40019.1
red260035.0
Cluster
Cache empty · this run will embed

Named topics

click a topic
Impressions by topic
Member queries length
QueryClicks
children shoe size chart420
how to measure a child’s foot210
child shoe length in cm88
toddler shoe sizes64

Uncategorised pile

37 lines

Outliers fail the minimum size or sit too far from every group. They stay in their own file so they do not invent a topic.

QueryClicksImpr.Why it was left out
shoes128,000too broad
buy31,200no subject
near me8900local, not a guide
sale152,400too broad
red2600one word
size61,500no subject

Topic brief

length.md sorted by clicks

Shoppers want a centimetre length, not another copy of the same chart. Lead with how to measure at home, then a short UK to EU table. 12,900 impressions, 782 clicks.

Highest line: “children shoe size chart”, 420 clicks, average position 8.2. A page can move that.

QueryClicksPos.
children shoe size chart4208.2
how to measure a child’s foot2106.1
child shoe length in cm8811.4
toddler shoe sizes649.8

How it works

  1. Read the query spreadsheet, keeping clicks, impressions and position beside each line.
  2. Process the queries in batches and save each one's meaning for reuse.
  3. Group queries with similar meaning, using a minimum group size and a similarity limit.
  4. Write queries that fit no group to a separate file.
  5. Have Gemma name each group and write one brief, with queries sorted by clicks.
  6. Repeat the run from the saved data when the similarity limit changes.

Pipeline

flowchart LR
  Export[Query export] --> Embed[Embed and cache]
  Embed --> Cluster[Density clustering]
  Cluster --> Noise[Uncategorised pile]
  Cluster --> Name[Name each group]
  Name --> Brief[Markdown brief]
  Brief --> Plan[Content plan]