Google 搜索 SEO Skill

---
name: google-search-seo
description: >-
  Audit and advise on Google Search SEO: crawl/index, ranking/SERP presentation,
  and SEO fundamentals (People-first, E-E-A-T, AI content, generative search).
  Use when the user asks about Google SEO, Googlebot crawl, indexing, robots.txt,
  sitemap, canonical, noindex, Search Console, rankings, snippets, sitelinks,
  Discover, AI Overviews, core updates, or wants a site SEO/crawl/rank readiness
  audit.
---

# Google Search SEO (Crawl · Index · Rank · SERP)

## Core Rule

Treat **Google Search Central** as the source of truth. This skill distills official guidance into an actionable audit; it does **not** guarantee crawl, index, or ranking.

Before a serious audit, refresh official Google docs when internet access is available. If live Google documentation conflicts with this skill, **live Google docs win**.

Every **site audit** must evaluate every checklist ID in `references/seo-requirements.md`. Mark each as `Pass` / `Fail` / `Unknown` / `N/A`. Do not sample only “likely” issues.

## Required Reference

Read `references/seo-requirements.md` before auditing. It contains the full SEO-* checklist and severity mapping.

For invocation examples and reusable prompts, read `references/usage-prompts.md` when the user asks how to use this skill or needs an audit request template.

## Non-negotiable Facts (never contradict)

1. **Crawl ≠ Index ≠ Serving** — fixing one layer does not fix the others.
2. **Crawl is not a ranking factor.** Paying for ads does not buy crawl or organic rank.
3. **`robots.txt` does not reliably hide pages** from search results; use crawlable `noindex` or auth/delete.
4. **`noindex` requires the URL to be crawlable** — `Disallow` + `noindex` often means Google never sees `noindex`.
5. **Sitemap / canonical / “follows guidelines” do not guarantee** crawl, index, or SERP features.
6. **Canonical is a hint**, not a force.
7. **Titles, snippets, site name, sitelinks, featured snippets, Discover, AI Overviews** are largely automated; you supply signals, Google may ignore them.
8. **AI Overviews need no special SEO markup** beyond being indexed, snippet-eligible, and technically sound.
9. **E-E-A-T is not a single ranking score**; People-first content beats search-engine-first spam.
10. **Reject** “guaranteed #1”, secret Google relationships, link schemes, and unverified AEO/GEO tool claims.

## When to Use Which Mode

| Mode | Trigger | Focus |
| --- | --- | --- |
| `full-audit` | Pre-launch / big redesign / full-site SEO audit | All SEO-* IDs |
| `crawl-index` | Not indexed, crawl errors, robots, sitemap, canonical | Section C IDs |
| `content-quality` | Thin/AI pages, spam risk, traffic drops after quality issues | Section B IDs |
| `serp-presentation` | Bad titles/snippets, no favicon/sitename, Discover | Section D IDs |
| `generative-ai` | AI Overviews / AI Mode visibility | Section E IDs |
| `diagnose` | User describes a symptom | Diagnostic table below, then relevant IDs |

## Diagnostic Table

| Symptom | Check first |
| --- | --- |
| Not crawled | Host availability, robots.txt, login wall, infinite URL space, Crawl Stats |
| Crawled but not indexed | Quality, `noindex`, soft 404, duplicate/non-canonical, Page Indexing |
| Indexed but not ranking / not showing | Relevance, quality, intent match, robots meta blocking snippets |
| Wrong URL in results | Canonical / 301 / hreflang / HTTP↔HTTPS / parameter variants |
| Want page hidden | Crawlable `noindex` or password — **not** robots.txt alone |
| Want less crawl budget waste | Shrink URL inventory, robots on non-critical, short redirects — **not** expect ranking boost |

## Audit Workflow

1. **Identify target**: live URL and/or local repo; site type (tool, blog, ecommerce, etc.); mode.
2. **Gather evidence**:
   - Fetch homepage, key templates, `robots.txt`, sitemap, sample canonicals, HTTP status.
   - In repo: routes, `robots`/`sitemap` generators, metadata/`hreflang`/`canonical`, SSR vs client-only content.
   - Sample content pages for People-first quality and template duplication.
3. **Classify**: `Blocker` / `High` / `Medium` + Pass/Fail/Unknown/N/A per ID.
4. **Report** using the output format below.
5. **Completeness gate**: every ID in `references/seo-requirements.md` appears once.

## Implementation Guidance

Prefer concrete findings:

- `robots.txt Disallow: /en/docs` instead of “crawl issue”
- `tool pages share near-identical FAQ templates` instead of “thin content”
- `canonical points to HTTP while site is HTTPS` instead of “canonical wrong”
- `meta robots nosnippet on money pages` instead of “snippet problem”

Do not promise rankings. Prefer Search Console (URL Inspection, Page Indexing, Crawl Stats, Sitemaps, Performance) over third-party “rank predictor” tools.

## Output Format

```markdown
**Decision**
Healthy / Needs work / Critical issues

**Mode**
full-audit | crawl-index | …

**Blockers**
- `SEO-…`: Issue. Evidence. Fix. (official Google Search Central basis if cited)

**High**
- …

**Medium**
- …

**Exhaustive Checklist**
| ID | Status | Evidence | Next action |
| --- | --- | --- | --- |
| SEO-… | Pass/Fail/Unknown/N/A | … | … |

**Completeness Check**
- IDs in reference: N
- IDs in report: N
- Missing: none | list
```

## Completeness Gate

Count IDs in `references/seo-requirements.md` and ensure the report includes every ID. Incomplete audits must not give a final Decision.