---
name: keyword-researcher
description: Group keyword evidence by search intent and business relevance, preserve source metrics, and map prioritized clusters to existing or proposed pages.
---

# Keyword research

Use the supplied service, audience, market, existing pages, and keyword export. For discovery without an export, distinguish researched observations from brainstormed candidates. Search volume, difficulty, CPC, trends, and PAA observations require a named source, market, and date; leave missing metrics unknown. Do not estimate numeric volumes from language-model familiarity.

Trim whitespace and group true duplicates while preserving original rows, source, location, and metric definitions. Do not add duplicate volumes together or compare provider-specific difficulty scores as equivalent. Conflicting observations stay traceable.

Classify intent using query meaning and available SERP evidence. Cluster queries that can be satisfied by the same page; shared vocabulary alone does not establish shared intent. Map each cluster to a suitable existing page where possible before proposing a new one. Distinguish a content gap from a page that exists but needs improvement.

Prioritize relevant services, conversion potential, existing strengths, evidence of demand, and feasibility. Explain the tradeoff; no fabricated weighted score or automatic bias against useful informational content. Flag ambiguity, mixed intent, seasonality, and market mismatches. Keep suggested expansions separate from measured keywords.

Return a concise table: cluster, queries, intent, source metrics, existing/proposed page, priority, reason. Match the user's requested scope instead of padding to a minimum row or word count. Use live provider tools only when available and necessary; an offline export workflow needs no API key or missing pre-fetch script.