A four-mode SEO intelligence skill for use with any MCP-compatible LLM. One prompt replaces 12+ hours of analyst work. Powered by live SEMrush data — no exports, no spreadsheets, no tab switching.
Built by iPullRank.
- An LLM that supports MCP tool calling (Claude, GPT-4o, Gemini, or equivalent)
- A SEMrush account with API access
- The SEMrush MCP server connected to your LLM
Claude.ai
- Create a Claude Project
- Go to Project Knowledge → upload
SKILL.md - Start a conversation — the skill is ready
Anthropic API / any LLM
- Pass the contents of
SKILL.mdas your system prompt - Add the SEMrush MCP server to your tool configuration
- Send your request as normal
Invoke with a plain English prompt. No special syntax required.
/semrush-seo-intelligence build an AIO industry dashboard for mortgage — use bankrate.com as the seed domain
/semrush-seo-intelligence find keyword gaps for skechers.com/women/shoes/ vs competitors
/semrush-seo-intelligence run a decay audit on engadget.com subfolders
/semrush-seo-intelligence build a topic cluster for "handheld gaming console"
Tracks AI Overview keyword exposure across a competitive domain set over 12 months. Discovers top competitors from a seed domain, pulls FK52/FP52 columns from domain_rank_history, computes AIO % by month, and renders an interactive dashboard with trend lines, a 3-month gainers/losers table, and a current snapshot ranking.
Trigger phrases: "AIO dashboard," "AI Overview by industry," "AIO trends," "which domains are winning AI Overviews"
Cost: ~900 SEMrush units for a full 30-domain, 6-industry run
Surfaces keywords that competitor pages rank for in the top 10 that your target URL is absent from. Pulls keyword sets for the target and up to 4 competitors, identifies gaps, scores by volume ÷ difficulty, and categorizes by intent. Output is a filterable gap table prioritized by opportunity.
Trigger phrases: "keyword gaps for [url]," "what am I missing vs competitors," "find keyword opportunities"
Maps 12 months of organic keyword count and traffic estimates across a domain's top subfolders. Classifies each subfolder as growing, stable, or declining. Flags specific pages and ranking positions at risk before they drop. Includes a domain-level keyword trend chart and page triage table.
Trigger phrases: "decay audit," "which pages are losing traffic," "subfolder health," "diagnose traffic drop"
Takes a seed keyword and builds a full content cluster. Pulls related terms, question variants, and exact-phrase variants from three SEMrush reports, then classifies every keyword by intent and cluster role: pillar variants to target on the cornerstone page, supporting pages to build around it, FAQ targets, and high-KD terms that warrant their own future pillar. Output is a filterable cluster map with priority scoring and content type recommendations.
Trigger phrases: "topic cluster," "pillar page strategy," "keyword cluster map," "content plan for [topic]"
This skill is engineered around SEMrush's domain_rank_history report for AIO data — the correct aggregate source — rather than per-keyword row scanning, which burns 50–200× more units and returns less accurate results.
| Approach | Units (30 domains) |
|---|---|
| This skill (Mode A) | ~900 |
| Naive per-keyword scan | ~1,050,000 |
- All modes use the
usdatabase by default. Modify the skill file to change region. - URL-level keyword data (
url_organic) is unavailable for many domains due to SEMrush indexing. The skill falls back todomain_organicwith topic filters automatically. - Keyword classifications (intent, cluster role) are model-inferred from volume, difficulty, and phrasing patterns — not pulled from SEMrush directly.
MIT. Use it, fork it, build on it.
iPullRank is a relevance engineering and AI search strategy agency. We build proprietary tooling to solve real problems in client work, then open-source what's useful to others.
Issues and PRs welcome. Questions: patrick@ipullrank.com