keyword-research

Keyword Research

Turn a niche into a shortlist of keywords worth building content around, backed by real Google Ads search-volume data.

Prerequisite: the keyword-research MCP server must be connected. It exposes three tools:

• generate_keyword_ideas — discover keywords. Seed with seed_keywords (up to 20), seed_url (one page), or seed_site (a whole domain). Plain-text location (default United States) and language_code (default en). Returns rows sorted by average monthly searches, descending; rows below min_avg_monthly_searches (default 10) are dropped. Paginate with cursor.
• resolve_geo_targets — list exact Google Ads locations for a place name. Only needed when a plain-text location is ambiguous and the tool says so.
• get_keyword_historical_metrics — fetch metrics for a keyword list you already have (brainstormed, from a competitor, from the user).

The server only fetches and normalizes data. Grouping keywords, judging ranking feasibility, and planning content are your job — this skill covers how.

Default workflow

1. Frame the niche. Establish three things, asking only if genuinely unclear: the topic/niche, the target website or page (if any), and the target market (country). A site or page URL is a strong seed — use it.

2. Discover. Call generate_keyword_ideas once:

   • If you have a target site: seed_site = the domain (or seed_url = a specific page you're planning content for).
   • Otherwise: seed_keywords = 3–8 tightly on-topic core terms for the niche. Off-topic seeds pollute every downstream result.
   • location = the target country in plain text; keep min_avg_monthly_searches and page_size at their defaults.

3. Expand once. Take the 5–10 most relevant results and run generate_keyword_ideas again with them as seed_keywords. This surfaces the long-tail around the terms that matter. Paginate the first call with cursor only if you still need a bigger pool.