SEO methodology for auditing, planning, and fixing organic search performance — backed by 60+ authoritative sources...
A complete SEO methodology distilled from 60+ authoritative sources. Use this skill to audit, plan, or improve any website's organic search performance.
Identify the user's job. Read ONLY the matching reference(s) — never read all of them.
| User's situation | Read | Why |
|---|---|---|
| General audit / "look at my site" | phase-1-technical.md + phase-2-content.md + pitfalls.md |
Covers ~80% of issues. |
| Speed / CWV problem (LCP/INP/CLS) | phase-1-technical.md §1.2 |
Optimization patterns. |
| Indexation / crawl / sitemap issues | phase-1-technical.md §1.1 + phase-4-programmatic.md §4.2 |
Crawl budget + canonicals. |
| Multi-language / hreflang | phase-3-international.md |
Return-link rule (#1 silent killer). |
| Programmatic / templated pages at scale | phase-4-programmatic.md + pitfalls.md |
March 2024 scaled-content. |
| Affiliate / monetization | phase-5-affiliate.md |
rel=sponsored, value-add. |
| CRO / CTAs / mobile-first | phase-6-conversion.md |
Placement, trust, mobile. |
| GSC analysis / quick wins | phase-7-monitoring.md |
Quick-win formula. |
| Traffic drop / "we got slapped" | phase-7-monitoring.md (§7.4) + pitfalls.md |
Diagnostic order. |
If unsure which references apply, ask the user one clarifying question — don't fan out and read everything.
Before guessing at someone's setup, run the bundled audit script when a URL is available:
python3 ~/.claude/skills/seo-growth-engine/scripts/audit.py https://example.com
Outputs JSON with HTTP status, robots.txt presence/contents, sitemap reachability, title/meta length, canonical, hreflang cluster, JSON-LD count, h1 count, image-without-alt count. Run this first whenever the user gives you a URL — it gives you facts instead of asking them to read their own HTML to you.
It is pure stdlib (no install needed), single-page only, and respectful (one HEAD + one GET). Don't crawl with it.
Lead with the diagnosis and the next action; keep any checklist below them, for readers who want depth.
Structure every response like this:
If the user asks a short focused question ("is X correct?"), give a short focused answer. Don't dump the workflow. Match length to question.
| Metric | Good | Poor | Source |
|---|---|---|---|
| LCP | ≤ 2.5s | > 4.0s | web.dev / CrUX field data |
| INP | ≤ 200ms | > 500ms | web.dev (replaced FID Mar 2024) |
| CLS | ≤ 0.1 | > 0.25 | web.dev |
| #1 result avg word count | ~1,447 | — | Backlinko 11.8M study |
| Pages with 0 organic traffic | 96.55% | — | Ahrefs |
| 100ms speed improvement | +8.4% conv | — | Deloitte |
| Top 25% landing pages convert | > 5.31% | avg 4.02% | Unbounce |
These principles override generic-advisor instincts. They exist because most SEO advice fails by being non-specific.
Full list in references/sources.md (60+ entries: Google official, industry research, conversion/ROI, technical deep dives). Cite from this list when making claims.