keyword-research

name	keyword-research
description	Use when needing to find content opportunities, identify keyword gaps, or research what competitors rank for. Uses web search to do manual keyword research without paid SEO tools.

Keyword Research (Manual — No Paid Tools)

Research keywords and content gaps using web search. Ahrefs/Semrush charge $99-199/mo for search volume data they get from Google Ads API + clickstream panels. You can get 80% of the value with web search analysis.

Process

1. Define Target Queries (5-10)

Read .agents/product-marketing-context.md for ICP and product positioning. Generate queries your target audience would search:

• "What is [concept]?" — awareness
• "[Framework] alternative" — consideration
• "[Framework A] vs [Framework B]" — evaluation
• "How to [task] in [language]" — implementation
• "Best [category] for [use case]" — decision

2. Search Each Query

For each query, use WebSearch and analyze:

• Who ranks? — note domain names, content types, word counts
• What's missing? — angles nobody covers, outdated info, gaps
• What format wins? — listicles, tutorials, comparisons, guides
• Our angle — what unique perspective can we offer?

3. Assess Competition

• Enterprise/high-authority domains (Google Cloud, AWS, IBM) = hard to beat head-on
• Medium blogs, personal sites, niche tools = beatable with better content
• Nobody ranking = blue ocean keyword, high priority

4. Score and Prioritize

For each keyword opportunity:

Factor	Weight	Score 1-5
Relevance to our ICP	40%
Competition beatable?	30%
Unique angle available?	20%
Content already exists?	10%

5. Save Findings

Write to .agents/keyword-research.md with:

• Keyword, who ranks, gap identified, recommended action
• Priority order for content creation
• Competitors to watch

Tips

• Google autocomplete = free keyword suggestions (search partial queries)
• "People Also Ask" boxes = related questions to target
• Reddit/HN threads = real language engineers use
• Check if our pages appear at all — if not, note it
• Long-tail beats head terms for new domains (target "ReAct pattern Python tutorial" not "AI agent")