surfup

Prompt: cluster search queries by user task

Produce reviewable clusters and flag ambiguous queries instead of forcing every term into a group.

Use the procedure on a representative sample, keep the evidence, and record the next action.

Use a repeatable query-clustering prompt

  • Export query and available click/impression data. Remove personal data and keep source, date range, country, and device in the file notes.
  • Prompt: “Cluster these queries by the task and intent a searcher wants to complete. Preserve the original query. Do not invent volume, intent certainty, or URLs.”
  • Require columns: cluster label, original query, likely intent, confidence, evidence, proposed page type, and unresolved question.
  • Review ambiguous groups against the actual search results and existing pages. Treat model output as a draft, not verified keyword research.

Primary sources

Latest materials

All materials

Practical articles and guides by SurfUp.

Guide · 10 min

“What is…” pages and AI answers: lost clicks or falling demand

A bank’s reference section lost 24% of visits as demand fell 13%. How to tell clicks lost to an AI block from falling demand and how to measure it.

Read
Guide · 8 min

Traffic dropped after a Google update

How to tell whether a traffic drop comes from a Google update, seasonality or a site fault. 2026 update timeline, a six-question check and a plan for the first 14 days.

Read
Guide · 9 min

Demand fell or the site lost: reconciling traffic with Wordstat

Demand for deposits fell 75%, traffic 37%; for cards demand −35%, traffic −63%. How to reconcile traffic with demand and not blame the site for the market.

Read