Compares what people search for with the pages on a site, shows the closest page for each search and marks the topics that no page answers.
Query Clustering & SEO
Cross-sector
Pipeline
Productised
Hybrid
2026
The problem
Search reports and page lists sat in separate spreadsheets, so nobody could tell which page should answer a given search. A services site could not see which searches landed on the wrong page or on no page at all.
What it does
Page text is stripped of formatting and compared with the searches. Each search or topic is then marked as a match or a miss against its closest page. A dashboard ranks pages for a selected search or a typed phrase. A miss gets a short drafted answer, so the gap is not left as a blank row.
Outcomes
A search shows the pages that already sit closest to it.
Topics with impressions and no matching page are listed on their own.
A miss comes with a drafted answer and not an empty cell.
The comparison can be cleared and rebuilt from the exports.
Simplified mockup
Interactive preview with sample data. Nothing is sent to a server.
Closest page
4 queries · 186 pages
Queries Clicks
Ranked pages Similarity
Driveway channel drainsMatch0.91
Slot drain installation0.74
Garden soakaway crates0.41
Garden soakaway cratesMatch0.88
Crate size guide0.71
Driveway channel drains0.33
No page reaches 0.55. Nothing is marked as a match.
Tank servicing notesBelow line0.34
Tanker call-out areas0.22
Jetting a blocked gullyMatch0.86
Drain rods and access0.62
Driveway channel drains0.29
Gap · septic tank emptying costNo green page180 clicks · 3,400 impressions
A short page on call-out fees, typical tank sizes and how often a household tank is emptied would answer this search. No page on the site covers prices. Closest page sits at 0.34.
Topic files
Clicks by topic
Topic
Clicks
Impressions
Closest page
Flag
Driveway drainage
420
6,800
Driveway channel drains
Matched
Soakaway crates
260
4,100
Garden soakaway crates
Matched
Septic emptying prices
180
3,400
Tank servicing notes
Gap
Blocked gully
95
1,200
Jetting a blocked gully
Matched
Vector index
Query vectors
0
Page vectors
0
Topics with a match
0
Built from the April export
Queries and page copy are embedded together. Click counts stay on the query vectors. Clearing drops both sets so the index can be built again from the exports.
Index is live
Query vectors
0
Page vectors
0
Index empty. Load the query export and the page export to build it again.
How it works
Convert the search and page exports into a form that can be compared by meaning, and store them with click counts.
Strip the formatting from pages written as PHP files and treat them the same way.
Rank pages by how close they are in meaning when a search is picked or a phrase is typed.
Mark each topic as matched or unmatched against the closest page.
Draft an answer for each unmatched topic with a language model.
Write the results to a review table or one file per topic, with clicks, impressions and the match flag.
Pipeline
flowchart LR
Pages[Page copy] --> Index[Embed pages and queries]
Queries[Query export] --> Index
Index --> Rank[Rank closest page]
Rank --> Match{Page answers it?}
Match -->|Yes| Keep[Matched row]
Match -->|No| Gap[Draft a gap answer]