About Services Tools Blog Contact us

Search Topics to Page Matching

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
Gap · septic tank emptying cost No green page 180 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
TopicClicksImpressionsClosest pageFlag
Driveway drainage4206,800Driveway channel drainsMatched
Soakaway crates2604,100Garden soakaway cratesMatched
Septic emptying prices1803,400Tank servicing notesGap
Blocked gully951,200Jetting a blocked gullyMatched

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

How it works

  1. Convert the search and page exports into a form that can be compared by meaning, and store them with click counts.
  2. Strip the formatting from pages written as PHP files and treat them the same way.
  3. Rank pages by how close they are in meaning when a search is picked or a phrase is typed.
  4. Mark each topic as matched or unmatched against the closest page.
  5. Draft an answer for each unmatched topic with a language model.
  6. 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]