Keyword Research
Build a keyword map grounded in search intent. Without paid API access, rely on SERP analysis, autocomplete patterns, and competitor pages. With seo-data-provider env keys, pull real volume/KD.
When to use
- New site / new market content plan
- Expanding clusters or languages (FR/EN/AR)
- Prioritizing which pages to create or refresh
When not to use
- Single-page title tweak (use
on-page-seo-optimizer)
- Fabricating a "volume report" with no API or export
Intent types
| Intent |
Signal |
Page type |
| Informational |
how, what, guide |
Blog / docs |
| Commercial |
best, vs, review, pricing |
Comparison / landing |
| Transactional |
buy, demo, quote, brand+product |
Product / contact |
| Navigational |
brand names |
Home / feature pages |
Workflow
- Seed list: products, services, pains, jargon from the user (FR + EN + AR when relevant).
- If MCP
search-console is connected, pull live queries/pages (impressions, CTR, position) for the property before estimating; treat GSC as ground truth over SERP guesses.
- If
seo-data-provider is available, pull volume/KD/related; else expand via web_search (variants, PAA themes, related searches).
- For each candidate, inspect the SERP: who ranks (giants vs niche), format (listicle, landing, video).
- Estimate difficulty qualitatively without API: Low / Medium / High from domain strength + content depth needed. Never invent numeric KD or monthly volume.
- Cluster by topic; assign one primary keyword + secondaries per target page.
- Save
seo/keyword-map-<topic>-<date>.md (and CSV if useful).
Keyword map format
| Cluster | Primary keyword | Intent | Difficulty | Volume | Target page | Priority | Notes |
|---------|-----------------|--------|------------|--------|-------------|----------|-------|
Volume: integer from API/export, or n/a (requires API data)
Difficulty: Low/Medium/High (qualitative) or API KD if sourced
Priority: P1/P2/P3 by business value × winnability
Rules
- Prefer MCP
search-console live data over CSV or estimates when the preset is connected (Settings → MCP).
- Never invent volume numbers; say estimate or requires Ahrefs/Semrush/DataForSEO/GSC.
- One primary keyword per page - no cannibalization.
- Prefer money-intent clusters when the goal is leads/sales.
- Pair with
seo-content-writer for execution and competitor-seo-analysis for gaps.
Anti-patterns
- 500-keyword dumps with no clusters or page owners
- Copying competitor keyword lists without SERP checks
- Treating brand navigational queries as content opportunities
1---2name: keyword-research3description: Find keywords, estimate difficulty and intent, and map opportunities to pages. Use when planning content, launching a site, or expanding into new topics or languages.4metadata:5 source: https://skillmd.com/skills/navinspire-ia/keyword-research6 navin:7 emoji: 🔑8 category: seo9---1011# Keyword Research1213Build a keyword map grounded in search intent. Without paid API access, rely on SERP analysis, autocomplete patterns, and competitor pages. With `seo-data-provider` env keys, pull real volume/KD.1415## When to use1617- New site / new market content plan18- Expanding clusters or languages (FR/EN/AR)19- Prioritizing which pages to create or refresh2021## When not to use2223- Single-page title tweak (use `on-page-seo-optimizer`)24- Fabricating a "volume report" with no API or export2526## Intent types2728| Intent | Signal | Page type |29|--------|--------|-----------|30| Informational | how, what, guide | Blog / docs |31| Commercial | best, vs, review, pricing | Comparison / landing |32| Transactional | buy, demo, quote, brand+product | Product / contact |33| Navigational | brand names | Home / feature pages |3435## Workflow36371. Seed list: products, services, pains, jargon from the user (FR + EN + AR when relevant).382. If MCP `search-console` is connected, pull live queries/pages (impressions, CTR, position) for the property before estimating; treat GSC as ground truth over SERP guesses.393. If `seo-data-provider` is available, pull volume/KD/related; else expand via `web_search` (variants, PAA themes, related searches).404. For each candidate, inspect the SERP: who ranks (giants vs niche), format (listicle, landing, video).415. Estimate difficulty **qualitatively** without API: Low / Medium / High from domain strength + content depth needed. Never invent numeric KD or monthly volume.426. Cluster by topic; assign one primary keyword + secondaries per target page.437. Save `seo/keyword-map-<topic>-<date>.md` (and CSV if useful).4445## Keyword map format4647```markdown48| Cluster | Primary keyword | Intent | Difficulty | Volume | Target page | Priority | Notes |49|---------|-----------------|--------|------------|--------|-------------|----------|-------|50```5152- `Volume`: integer from API/export, or `n/a (requires API data)`53- `Difficulty`: Low/Medium/High (qualitative) or API KD if sourced54- `Priority`: P1/P2/P3 by business value × winnability5556## Rules5758- Prefer MCP `search-console` live data over CSV or estimates when the preset is connected (Settings → MCP).59- Never invent volume numbers; say estimate or requires Ahrefs/Semrush/DataForSEO/GSC.60- One primary keyword per page - no cannibalization.61- Prefer money-intent clusters when the goal is leads/sales.62- Pair with `seo-content-writer` for execution and `competitor-seo-analysis` for gaps.6364## Anti-patterns6566- 500-keyword dumps with no clusters or page owners67- Copying competitor keyword lists without SERP checks68- Treating brand navigational queries as content opportunities