Skills

google-news-seo

Installation
SKILL.md

Google News SEO Diagnostic Engine

Language / 语言：Detect the user's language and respond in the same language throughout. 用户用中文提问则全程用中文回复；用英文提问则全程用英文回复。

Disclaimer / 声明：本 skill 为启发式审计框架，非 Google 官方结论；评分与清单不能替代 Search Console、Publisher Center 或法律顾问对内容政策的判断。

HTTP fetch · 抓取工具约定

凡需拉取线上页面 HTML 的步骤，本文件统一写为 web_fetch。在运行时请映射到你环境中实际可用的等价能力（例如 Cursor 内置网页抓取、MCP fetch 等），使用工具面板中的确切名称，避免因大小写或别名不一致导致未调用。

Priority rubric · 优先级分级（对齐业界）

  • P0（Blocking / Compliance）：直接影响抓取/收录，或涉及误导性标注与内容政策风险。例：robots 阻断、文章不可访问、datePublished 明显错误、作者署名存在明显欺骗。
  • P1（High impact）：显著影响 Google News 可见性、稳定收录或排名竞争力，但通常不构成单点阻断。例：News sitemap 信号弱、Schema 关键字段缺失、作者页不完整。
  • P2（Best practice / Optimization）：优化点击率、可读性或质量信号，属于增益项。例：标题长度区间、关键词前置、heading 层级微调。

Industry note / 行业口径：依据 Google Search Central 与 Publisher Center 文档，Article/NewsArticle 结构化数据是推荐增强，并非 Top Stories/Google News 的绝对硬性门槛。只有当 Schema 与页面事实明显冲突或构成误导时，才可升级为 P0/P0-candidate。

Google News System Model

Google News operates on a two-layer architecture. This skill evaluates both layers independently.

Layer Name Function
Layer 1 News Index System Determines whether a site enters the Google News index
Layer 2 News Ranking System Determines whether articles rank in topic clusters and appear in Top Stories

Layer 1 must pass before Layer 2 matters. If a site is not indexed, ranking optimization is premature.

Diagnostic routing:

  • Layer 1 fails → focus on index eligibility (publisher trust, crawlability, Schema, sitemap)
  • Layer 1 passes, Layer 2 fails → focus on ranking signals (freshness, content type, cluster compatibility, competitors)
  • Both pass → surface remaining optimization opportunities via competitor gap analysis

0 · Initial Assessment 审计前评估

Before starting any checks, gather context by asking the following questions. Skip any question already answered in the user's prompt.

# Question / 问题
1 Site type / 站点类型 — Is this a dedicated news publisher, a corporate blog with a news section, or another site type?
2 Target topics / keywords / 目标关键词或议题 — Which topics, keywords, or named entities are most important for this audit?
3 Current status / 当前状态 — Any known issues, recent migrations, or Google Search Console alerts?
4 Scope / 审计范围 — Full Google News Diagnostic (Layer 1 + Layer 2 + Competitor Analysis), article-level audit, or a specific area (Schema only / EEAT only / Technical only / On-Page only)?

Skip rule / 跳过规则： If the user's prompt already implies answers (e.g., provides a URL + says "check the Schema"), skip the answered questions and proceed directly.

Full Diagnostic trigger / 全站诊断触发： If scope is "Full Diagnostic" or the user asks "why is my site not in Google News" / "why don't my articles rank" / "why do competitors outrank me", run all Layer 1 + Layer 2 checks and generate the Google News Diagnostic Report (Section 10).

Acknowledgement / 确认语句： Before starting checks, output one line:

Auditing [URL / article / site] — scope: [scope summary]. Starting checks...

0.5 · News Index Presence Detection 索引存在检测

Run this check first when performing a Full Diagnostic or when the user asks why their site is not appearing in Google News.

Methodology limits / 方法局限

通过 WebSearch 构造的 site:<domain> 类查询 不等于 Google 内部「Google News 收录」状态：结果受索引区域、个性化、抓取快照时间与搜索界面模块影响。本节结论必须标注为 启发式（approximate），并建议用户用 Google Search Console、Publisher Center、在搜索工具中选择「新闻」或站点专属报告、以及 Rich Results Test 等交叉验证。

Step 1 — Simulate Google News index query

Use WebSearch to simulate a site:<domain> query in a Google News context. Construct the search as:

site:<domain> news articles

Analyze the returned results to determine whether the site's articles appear in Google News coverage.

Step 2 — Classify index presence

Result Status Definition
0 results detected Not Indexed No evidence of Google News indexing
1–5 results Limited Presence Some articles indexed; inconsistent coverage
6+ results across multiple days Strong Presence Active Google News publisher

Also record:

  • Total detected article count
  • Timestamp of the most recently indexed article
  • Whether results span multiple days/topics (diversity indicator)

Step 3 — Route diagnostic focus

Index Status Diagnostic Focus
Not Indexed Prioritize Layer 1 checks; flag as root cause category
Limited Presence Run both Layer 1 and Layer 2; present findings for each
Strong Presence Note Layer 1 as passing; focus diagnostic output on Layer 2 ranking checks

Output format

News Index Status: [Not Indexed ❌ / Limited Presence ⚠️ / Strong Presence ✅] (heuristic — not official Google News index)
Detected article count: [N]
Latest indexed article: [timestamp or "not detected"]
Diagnostic focus: [Layer 1 / Both Layers / Layer 2]
Suggested verification: [e.g. GSC coverage, Publisher Center, News tab site: search]

1 · Prerequisites 前置要求

Determine input type first / 先判断产物类型：

Input Action
Live URL / 线上 URL Fetch page, extract JSON-LD Schema
Raw Schema JSON Parse directly
URL or HTML + EEAT scan request / URL 或 HTML + EEAT 扫描请求 Run full E-E-A-T analysis → see Sections 7–9 / 执行 EEAT 全维度分析 → 见第 7–9 节

Schema Fetch Protocol / Schema 获取流程

Retrieve JSON-LD in three sequential phases. Only advance to the next phase if the current one yields no JSON-LD blocks.

⚠️ Anti-pattern — do NOT do this: Do NOT report "JSON-LD 无法通过抓取自动提取（前端渲染）" or "client-side rendered" based solely on a failed web_fetch. SSR/SSG sites (Next.js, Nuxt, Hugo, WordPress) pre-render JSON-LD into static HTML — curl can fetch it directly. A fetch failure alone is not evidence of CSR.

Phase 1 — web_fetch（优先）

Use web_fetch to retrieve the page. If the response contains one or more <script type="application/ld+json"> blocks → extract and proceed to Schema analysis. Stop here.

Phase 2 — curl fallback（web_fetch 无结果时）

Note: Attempting curl fallback — SSR/SSG sites pre-render JSON-LD into static HTML; curl can retrieve it directly without JS execution.

sed -i portability / 跨平台：下文对临时 Python 脚本的占位符替换——macOS (BSD sed) 使用 sed -i '' "s|…|…|g" file；Linux (GNU sed) 使用 sed -i "s|…|…|g" file（无中间 ''）。任选与当前 OS 匹配的写法。

If Shell tool is unavailable, skip directly to Phase 3.

# Step A: fetch raw HTML with Googlebot UA
# Note: macOS mktemp doesn't support extensions; use a plain suffix
TMPFILE=$(mktemp /tmp/page_XXXXXX)
curl -sL \
  -H "User-Agent: Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)" \
  --max-time 15 \
  "<URL>" > "$TMPFILE" 2>&1
echo "File size: $(wc -c < "$TMPFILE") bytes"
  • 403 / 429: retry once with Chrome UA: -H "User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36"
  • Timeout / non-zero exit / file size 0: skip to Phase 3
# Step B: extract JSON-LD — write Python script via heredoc to avoid quoting issues
cat > /tmp/extract_jsonld.py << 'PYEOF'
import re, json
html = open('/tmp/page_XXXXXX.html').read()  # replace XXXXXX with actual suffix from $TMPFILE
pattern = re.compile(r'<script[^>]+type=["\']application/ld\+json["\'][^>]*>(.*?)</script>', re.DOTALL | re.IGNORECASE)
blocks = pattern.findall(html)
print(f'Total blocks found: {len(blocks)}')
for i, b in enumerate(blocks, 1):
    print(f'=== JSON-LD Block {i} ===')
    try:
        print(json.dumps(json.loads(b.strip()), indent=2, ensure_ascii=False))
    except Exception as e:
        print(f'Parse error: {e}')
        print(b.strip()[:500])
PYEOF
# Replace the filename placeholder with the actual $TMPFILE path, then run:
sed -i '' "s|/tmp/page_XXXXXX.html|$TMPFILE|g" /tmp/extract_jsonld.py
python3 /tmp/extract_jsonld.py

# Step B fallback — grep (if Python 3 unavailable)
grep -oE '<script[^>]+type="application/ld\+json"[^>]*>.*?</script>' "$TMPFILE" | sed 's/<[^>]*>//g'

# Step C: clean up
rm -f "$TMPFILE" /tmp/extract_jsonld.py

If extraction returns results → use for Schema analysis; note "Schema retrieved via curl fallback".

  • Multiple blocks: extract all; use the block with "@type": "NewsArticle" for Schema analysis
  • Malformed JSON: output raw block, flag as "partially retrievable", continue best-effort analysis

Phase 3 — 🔍 Manual（仅在两阶段均无结果时）

Mark Schema detection as 🔍 Manual. Output exactly:

Schema could not be auto-detected (web_fetch and curl both returned no JSON-LD).
Verify manually: 🔗 https://search.google.com/test/rich-results?url=<URL>

❌ Do NOT say: "前端渲染" / "client-side rendering" / "JavaScript-rendered" ✅ Only say: "could not be auto-detected"

3 hard requirements for Google News inclusion / 三项硬性门槛：

  1. Dedicated news publisher / 专属新闻网站 — the site's core purpose must be news, not a product with a news section
  2. Content policy compliance / 内容政策合规 — no dangerous / deceptive / manipulated-media content; AI-generated content must be transparently disclosed
  3. Technical compliance / 技术合规 — permanent URLs, HTML-rendered content, robots.txt must not block Googlebot-News

1.5 · Publisher Trust Detection 发布者信任度检测

Layer 1 check. Google evaluates whether the site operates as a legitimate news publisher before indexing it.

Step 1 — Verify trust pages

Use web_fetch to check each URL. A page passes if it returns HTTP 200 with non-trivial content (not a redirect to homepage).

Page URL pattern Points
About <domain>/about 8 pts
Editorial Policy <domain>/editorial-policy 8 pts
Contact <domain>/contact 8 pts
Team <domain>/team 8 pts
Authors <domain>/authors 8 pts

Subtotal (pages): 40 pts

Step 2 — Detect newsroom description

Scan the About page text for presence of keywords indicating a professional newsroom: newsroom · editorial team · journalism · reporters · editor-in-chief · press · media organization

  • Keywords present → 30 pts
  • Absent → 0 pts, flag as P1: "No editorial identity statement found on About page"

Step 3 — Detect Organization Schema

Check homepage and About page JSON-LD for @type: Organization or @type: NewsMediaOrganization with fields: name, url, logo.

Condition Points
All three fields present 30 pts
1–2 fields present 15 pts
Schema absent 0 pts

Publisher Trust Score Output

Publisher Trust Score: [0-100]

| Check | Result | Notes |
|-------|--------|-------|
| /about | Pass ✅ / Fail ❌ | |
| /editorial-policy | Pass ✅ / Fail ❌ | |
| /contact | Pass ✅ / Fail ❌ | |
| /team | Pass ✅ / Fail ❌ | |
| /authors | Pass ✅ / Fail ❌ | |
| Newsroom description | Pass ✅ / Fail ❌ | |
| Organization Schema | Pass ✅ / Partial ⚠️ / Fail ❌ | [missing fields] |

1.6 · Author Authority Detection 作者权威性检测

Layer 1 check. Google builds an author credibility graph. Unverifiable or AI-generated author attribution is a strong negative signal.

Step 1 — Verify author presence and schema match

Check both:

  1. Visible byline on the page (author name in HTML)
  2. author.name in NewsArticle Schema
Condition Result
Both present and matching Pass ✅ — 30 pts
Schema author missing Fail ❌ — 0 pts (P1)
Mismatch between Schema and page byline Fail ❌ — 0 pts (P0)

Step 2 — Fetch author profile page

Use web_fetch on the URL from author.url (Schema) or the byline hyperlink.

A complete author profile requires:

  • Author name visible
  • Biography ≥ 50 words
  • At least one of: job title, social media link, publication count
Condition Points
Full profile (name + bio 50w+ + credential) 40 pts
Partial profile (name + bio only) 20 pts
Profile missing or 404 0 pts (P1)

Step 3 — Detect suspicious AI author names

Scan author.name (Schema) and page byline for obvious AI-as-author labels (e.g. role phrases), not arbitrary substrings inside human names:

AI Agent · ChatGPT · GPT-4 · OpenAI (as sole byline) · System Writer · Auto Writer · AI Writer · Gemini (as product credited as author) · Claude only when clearly denoting the model as author (e.g. "Written by Claude"), not when part of a plausible human name (e.g. "Jean-Claude", "Claude Smith").

  • Clear AI-as-author attribution → flag P0 候选，且必须标注 需人工确认；不得仅凭姓氏或常见人名片段判负。
  • Ambiguous / substring-only match → 🔍 Manual，不扣满分项。
  • No concerning pattern → 10 pts

Step 4 — Detect social / credential signals

Check author profile for: Twitter/X link, LinkedIn link, professional email, years of experience mentioned in bio, or domain expertise statement.

  • 1+ signals present → 20 pts
  • No signals → 0 pts

Author Authority Score Output

Author Authority Score: [0-100]

| Check | Result | Notes |
|-------|--------|-------|
| Author byline present | Pass ✅ / Fail ❌ | |
| Schema author.name matches byline | Pass ✅ / Mismatch ⚠️ / Fail ❌ | |
| Author profile page | Full ✅ / Partial ⚠️ / Missing ❌ | |
| No suspicious AI name | Pass ✅ / Flag ❌ | [matched keyword if flagged] |
| Social / credential signal | Present ✅ / Absent ⚠️ | |

2 · NewsArticle Schema Checklist Schema 审查清单

Core checks — high impact / 核心检查（高影响，默认 P1）

Calibration note / 评级口径：以下项默认按 P1 处理（影响可见性与富结果质量）；若出现误导性标注（例如作者/发布时间与页面事实冲突），再升级为 P0 候选并要求人工确认。

- [ ] @context = "https://schema.org"  (not http://)
- [ ] @type = "NewsArticle"  (matches actual content type)
- [ ] dateModified >= datePublished  (must not be earlier)
- [ ] image URL contains no AI-tool markers  (qwen_generated / ChatGPT Image / dall-e / midjourney)
- [ ] author.name matches the byline shown on the page  (Schema ≠ AI agent while page shows human)
- [ ] author is a real person with a verifiable author page  (not a team name or AI Agent)
- [ ] publisher.logo exists and is a valid ImageObject

Recommended — affects rich results / 建议补充项（影响富摘要）

- [ ] mainEntityOfPage points to the article URL
- [ ] description field present  (populated from article summary)
- [ ] articleSection set
- [ ] BreadcrumbList last item has an "item" URL

3 · AI Content Checks AI 内容专项

Issue / 问题 Risk / 风险 Fix / 修复
Image filename contains AI tool name Manipulated media policy violation Rename on upload; strip AI-tool prefixes
Schema author ≠ page byline Deceptive markup Use the same real editor name in both
No human editor byline Insufficient E-E-A-T Establish human editorial attribution
AI Agent listed as author Unverifiable authority Replace with the reviewing editor

Recommended attribution pattern for AI content / AI 内容推荐署名模式：

  • Schema author → real editor's Person node
  • Page display → "Generated by AI, reviewed by [Editor Name]" / "AI 生成，经 [编辑姓名] 审校"
  • Editor's author page must include: real name, professional bio, contact

4 · Schema Fix Template 修复模板

{
  "@context": "https://schema.org",
  "@type": "NewsArticle",
  "mainEntityOfPage": {
    "@type": "WebPage",
    "@id": "https://example.com/news/article-slug/"
  },
  "headline": "Article title / 文章标题",
  "description": "100-200 char summary / 文章摘要",
  "datePublished": "2026-03-03T19:55:26-05:00",
  "dateModified": "2026-03-03T20:10:00-05:00",
  "image": ["https://cdn.example.com/images/article-cover.jpg"],
  "author": [{
    "@type": "Person",
    "name": "Editor Name / 编辑姓名",
    "url": "https://example.com/author/editor-slug/",
    "jobTitle": "Senior Editor"
  }],
  "publisher": {
    "@type": "Organization",
    "name": "Publication Name",
    "url": "https://example.com",
    "logo": {
      "@type": "ImageObject",
      "url": "https://example.com/logo.png",
      "width": 600,
      "height": 60
    }
  },
  "articleSection": "Markets",
  "isAccessibleForFree": true
}

5 · Systemic Bugs 系统性 Bug（批量修复）

If the same issue appears across multiple articles, fix at the template level. 若多篇文章存在相同问题，需从模板层修复，不要逐篇处理。

Bug Root cause / 根因 Fix / 修复
dateModified < datePublished Field assignment order / 赋值顺序错误 Set modified to last-edited timestamp
"http://schema.org" Hardcoded old protocol / 模板写死旧协议 Global replace → "https://schema.org"
publisher.logo missing Not in template / 模板缺字段 Add logo ImageObject to Schema template
description missing Not mapped from summary / 未映射摘要 Auto-populate from summary / excerpt
BreadcrumbList last item has no item Template omission / 模板漏写 Set last item's item = article URL
AI-tool name in image filename No rename on upload / 上传未重命名 Strip AI-tool prefixes in upload pipeline

5.5 · Technical SEO 技术检查

Check / 检查项 Pass Condition / 通过条件 Priority Auto / Manual
robots.txt accessible <domain>/robots.txt returns HTTP 200 P1 Auto
Googlebot-News not blocked No Disallow rule under User-agent: Googlebot-News covering the article path P0 Auto
News Sitemap exists Sitemap URL (from robots.txt or <domain>/news-sitemap.xml) is accessible and contains <news:news> tags P1 Auto
News Sitemap has <news:news> tags Sitemap is valid per Google News Sitemap spec P1 Auto
News Sitemap freshness At least one <news:publication_date> within the last 48 hours P1 Auto
News Sitemap Health Score accessible (30 pts) + valid news namespace (40 pts) + 48h freshness (30 pts) — Auto
Crawlability Score article text in HTML (30 pts) + server-rendered Schema (30 pts) + canonical valid (20 pts) + robots not blocking (20 pts) — Auto
Core Web Vitals — LCP LCP < 2.5s (Good); 2.5–4s (Needs Improvement ⚠️); > 4s (Poor ❌) P1 🔍 Manual
Core Web Vitals — INP INP < 200ms (Good); 200–500ms (⚠️); > 500ms (❌) P1 🔍 Manual
Core Web Vitals — CLS CLS < 0.1 (Good); 0.1–0.25 (⚠️); > 0.25 (❌) P1 🔍 Manual
HTTPS Article URL uses https:// P0 Auto

CWV verification / CWV 验证： Use PageSpeed Insights: https://pagespeed.web.dev/report?url=<url>

News Sitemap Health Score calculation:

Score = (accessible ? 30 : 0) + (news namespace valid ? 40 : 0) + (article within 48h ? 30 : 0)

5.6 · On-Page SEO 文章页检查

Title Tag / 标题标签

Check Pass Condition Priority
Title tag present <title> element exists and is non-empty P0
Length 50–70 characters Title length between 50 and 70 characters P2
Primary keyword near start Key topic/entity within the first 60 characters P2
Unique (not duplicated) Title is not identical to other pages (🔍 Manual) P1
No keyword stuffing Same keyword not repeated > 2 times P2

Meta Description

Check Pass Condition Priority
Meta description present <meta name="description"> exists and non-empty P2
Length 120–160 characters Description length between 120 and 160 characters P2
Contains primary keyword Primary topic/entity appears in description P2

Canonical Tag

Check Pass Condition Priority
Canonical tag present <link rel="canonical"> exists P1
Points to correct URL Canonical URL matches the article URL (self-referencing) or a valid canonical version P1 / P0 if target returns non-200

Heading Structure

Check Pass Condition Priority
Exactly one H1 Page contains exactly one <h1> element P2
H1 matches headline H1 text consistent with article headline P2
Logical hierarchy No heading levels skipped (e.g., no H1 → H3 without H2) P2
H1 contains primary keyword Primary topic/entity appears in H1 P2

5.7 · Crawlability and Rendering Check 爬取与渲染检测

Layer 1 check. Verifies that Googlebot can access and read the article content without JavaScript execution.

Step 1 — Fetch with Googlebot UA

TMPFILE=$(mktemp /tmp/crawl_XXXXXX)
curl -sL \
  -H "User-Agent: Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)" \
  --max-time 15 \
  "<ARTICLE_URL>" > "$TMPFILE" 2>&1
echo "File size: $(wc -c < "$TMPFILE") bytes"

Step 2 — Detect article body in initial HTML

sed -i: macOS 用 sed -i '' "s|…|…|g"；Linux (GNU) 用 sed -i "s|…|…|g"（与 §1 Schema Phase 2 相同）。

cat > /tmp/check_body.py << 'PYEOF'
import re
html = open('TMPFILE_PATH').read()
clean = re.sub(r'<(script|style)[^>]*>.*?</(script|style)>', '', html, flags=re.DOTALL|re.IGNORECASE)
clean = re.sub(r'<[^>]+>', ' ', clean)
clean = re.sub(r'\s+', ' ', clean).strip()
print(f"Visible text length: {len(clean)} chars")
print(f"First 300 chars: {clean[:300]}")
PYEOF
sed -i '' "s|TMPFILE_PATH|$TMPFILE|g" /tmp/check_body.py
python3 /tmp/check_body.py
  • ≥ 200 chars of visible text → article text present → 30 pts
  • < 200 chars → flag: "Article text not detected in initial HTML — recommend verifying with Rich Results Test"

❌ Do NOT label as "client-side rendered" — only say "article text not detected in initial HTML"

Step 3 — Detect server-rendered Schema

Check whether raw HTML contains <script type="application/ld+json"> with "@type": "NewsArticle" before any dynamically loaded bundles.

Result Points Note
NewsArticle Schema in raw HTML 30 pts Server-rendered ✅
JSON-LD found but no NewsArticle type 15 pts Partial
No JSON-LD in raw HTML 0 pts Flag: "Potential hydration-only Schema — verify Next.js/Nuxt SSR is configured to pre-render Schema (use getServerSideProps or static export)"

Step 4 — Verify canonical and robots

  • Canonical tag present and pointing to article URL → 20 pts
  • Not blocked by robots.txt for Googlebot or Googlebot-News → 20 pts

Crawlability Score Output

Crawlability Score: [0-100]

| Check | Result | Points | Notes |
|-------|--------|--------|-------|
| Article text in initial HTML | Pass ✅ / Fail ❌ | 30 / 0 | |
| Server-rendered Schema | Full ✅ / Partial ⚠️ / Absent ❌ | 30/15/0 | |
| Canonical tag valid | Pass ✅ / Fail ❌ | 20 / 0 | |
| Not blocked by robots | Pass ✅ / Fail ❌ | 20 / 0 | |

5.8 · URL Structure Analysis URL 结构分析

Layer 1 check. URL patterns affect Googlebot's ability to recognize and categorize news content.

Recommended patterns ✅

  • /news/<descriptive-slug>
  • /article/<descriptive-slug>
  • /<year>/<month>/<descriptive-slug>
  • /<category>/<descriptive-slug>

Problematic patterns ❌

Pattern Example Priority Recommendation
Query-parameter article ID ?id=123, ?article_id=456, ?p=789, ?postid=1 P1 Migrate to path-based URL; implement 301 redirect
Numeric-only slug /news/12345 (no keywords) P2 Add descriptive keywords to URL slug
Session ID in URL ?session=abc, ?PHPSESSID= P0 Remove session parameters from indexable URLs
Tracking params without canonical ?utm_source= without self-referencing canonical P1 Ensure canonical points to clean URL

Output

URL Structure: [Recommended ✅ / Needs Improvement ⚠️ / Problematic ❌]
Detected pattern: [pattern description]
Issues found: [list or "None"]

5.9 · Freshness Signal Analysis 新鲜度信号分析

Layer 2 check. Google News heavily weights freshness; publication timing is a competitive ranking factor.

Step 1 — Validate datePublished / dateModified

Check Pass Condition Priority Points
datePublished present ISO 8601 timestamp in Schema P0 required
dateModified ≥ datePublished Modified timestamp not earlier than published P0 30 pts if valid
No excessive modification Not modified > 5× within 24 hours P1 deduct 15 if flagged

Step 2 — Estimate publication speed

Compare datePublished with the article's <news:publication_date> in the News Sitemap (if available), or use HTTP Last-Modified / Date response headers as proxy.

Speed Condition Points
Fast Sitemap entry within 30 min of datePublished 40 pts
Moderate 30 min – 1 hour gap 20 pts
Slow > 1 hour gap 0 pts

Step 3 — Flag manipulation signals

  • dateModified updated repeatedly with no detectable content change → flag as "possible freshness manipulation" (P1)
  • datePublished set in the future → flag as P0 error

Freshness Score Output

Freshness Score: [0-100]

| Signal | Result | Points |
|--------|--------|--------|
| datePublished / dateModified valid | Pass ✅ / Fail ❌ | 30 / 0 |
| Publication speed | Fast ✅ / Moderate ⚠️ / Slow ❌ | 40/20/0 |
| No manipulation signals | Pass ✅ / Flagged ⚠️ | 30 / 15 |

5.10 · Content Type Classification 内容类型分类

Layer 2 check. Content type affects cluster placement, ranking duration, and competitive differentiation.

Step 1 — Classify article type

Evaluate based on headline, body length, quote count, and source references:

Type Signals Base Score
Breaking News Published within 2h of event; headline includes "Breaking", "Just In", "Developing"; body < 600 words 90–100
Analysis Body 800+ words; 3+ named sources or citations; analytical headline ("Why", "How", "What This Means For") 70–90
Digest / Roundup 5+ outbound links to sources; "Roundup", "Weekly", "Today's News" in headline 40–60
Aggregation No direct quotes; no named reporter byline; summarizes other sources only 20–40

Step 2 — Detect AI template patterns

Scan H2/H3 headings within the article body for:

Key Takeaways · Pros and Cons · Opposing Views · Summary · FAQ · Related Topics · Bottom Line

  • Any pattern detected → flag as "AI template structure detected" (P1) → deduct 20 pts from base score
  • Recommend restructuring to inverted-pyramid news format

Content Type Score Output

Content Type: [Breaking News / Analysis / Digest / Aggregation]
AI Template Detected: [Yes ⚠️ / No ✅]
Content Type Score: [0-100]

5.11 · Publisher Authority Estimation 发布者权威估算

Layer 2 check. Google evaluates publisher-level authority when ranking articles in topic clusters.

Estimate publisher authority using available signals:

Signal How to check Points
Article volume Count total articles in News Sitemap; 10+ = strong, 3–9 = moderate, < 3 = low 25 pts
Organization Schema @type: Organization or NewsMediaOrganization present on homepage 25 pts
Brand mentions WebSearch "<publisher name>" news — check result count and source quality 25 pts
Internal link structure Article pages link to author pages, topic hubs, related articles 25 pts

Publisher Authority Score Output

Publisher Authority Score: [0-100]

| Signal | Result | Notes |
|--------|--------|-------|
| Article volume | Strong ✅ / Moderate ⚠️ / Low ❌ | [count] |
| Organization Schema | Present ✅ / Absent ❌ | |
| Brand mentions | Strong ✅ / Moderate ⚠️ / Weak ❌ | |
| Internal link structure | Good ✅ / Partial ⚠️ / Poor ❌ | |

5.12 · Topic Cluster Compatibility 话题聚类兼容性

Layer 2 check. Google News groups articles about the same event into topic clusters; these signals determine cluster membership.

Step 1 — Headline entity analysis

Check whether the headline contains at least one named entity (person, organization, location, financial instrument, or event name).

  • Named entity present → 25 pts; note the detected entity
  • Generic headline (no named entities) → 0 pts (P1): "Add specific entity names to headline for cluster matching"

Step 2 — Entity density in body

Count named entities per 100 words in the article body (use recognizable proper nouns as proxy):

  • ≥ 1 entity per 100 words → 25 pts
  • < 1 entity per 200 words → 0 pts (P1): "Increase named entity density — add specific company names, people, or locations"

Step 3 — Timeliness check

  • Article published within 6 hours of the referenced event → 25 pts
  • Published 6–24 hours after → 15 pts
  • Published > 24 hours after → 0 pts

Step 4 — Original reporting signals

Check for any of:

  • Direct quoted speech with attribution ("…" said [Name])

  • Exclusive data, statistics, or documents

  • Bylined reporter with a domain-specific author profile

  • 1+ signals present → 25 pts

  • None detected → 0 pts (P1): "Add original reporting signals — direct quotes or exclusive data"

Topic Cluster Compatibility Score Output

Topic Cluster Compatibility Score: [0-100]

| Signal | Result | Points |
|--------|--------|--------|
| Headline named entity | Present ✅ / Absent ❌ | 25 / 0 |
| Entity density | High ✅ / Low ❌ | 25 / 0 |
| Timeliness | < 6h ✅ / 6-24h ⚠️ / > 24h ❌ | 25/15/0 |
| Original reporting | Present ✅ / Absent ❌ | 25 / 0 |

5.13 · Top Stories Detection Top Stories 检测

Layer 2 check. Determines whether the article topic triggers a Google Top Stories carousel and whether the analyzed site appears in it.

Step 1 — Extract topic keyword

From the article headline, identify the primary entity or event name. Use the most specific named entity (e.g., "Apple WWDC 2026" rather than "tech event").

Step 2 — Search for Top Stories carousel

Use WebSearch to search the topic keyword. Analyze results for:

  • A "Top Stories", "News", or "In the News" carousel module
  • Publisher names and headlines appearing in the carousel
  • Timestamps of carousel articles
Result Output
Top Stories carousel found Topic triggers Top Stories; extract publisher list
No carousel found "No Top Stories carousel detected for this topic"

Step 3 — Compare site against carousel publishers

Check whether the analyzed domain appears in the list of carousel publishers.

Result Status
Analyzed site in carousel Top Stories Presence: Confirmed ✅
Analyzed site absent, competitors present Top Stories Presence: Gap detected ❌
No carousel exists Top Stories Presence: Topic not triggering carousel ⚠️

Output

Top Stories Presence: [Confirmed ✅ / Gap detected ❌ / Not triggering ⚠️]

Competitors in carousel:
| Publisher | Headline | Published |
|-----------|----------|-----------|
| [name] | [headline] | [time] |

6 · Output Report Format 总结报告格式

After analysis, output the report in the following structure. 分析完成后按对应结构输出。

Full Diagnostic mode (Section 10 triggered): System Model → Dual-Layer Scorecard → Executive Summary → Detailed Check Tables → Competitor Gap Analysis → Priority Fix List

Article / Scoped audit mode: Executive Summary → Detailed Check Tables → Priority Fix List

Dual-Layer Scorecard (Full Diagnostic only)

Output at the top of Full Diagnostic reports, before the Executive Summary.

## Google News Diagnostic Report

**Google News SEO Score: XX / 100**  [🟢 Strong / 🟡 Developing / 🔴 At Risk]

| Layer | Score | Status |
|-------|-------|--------|
| Layer 1 — Index Eligibility | XX/100 | Pass ✅ / Partial ⚠️ / Fail ❌ |
| Layer 2 — Ranking Potential | XX/100 | Pass ✅ / Partial ⚠️ / Fail ❌ |

**Diagnosis**: [plain-language statement — see Section 10 for routing logic]

Executive Summary 执行摘要

Output this section first (after Dual-Layer Scorecard if present), before all detailed tables.

Overall Health / 总体健康度：

Condition Rating
0 P0 issues AND ≤ 2 P1 issues 🟢 Good
1–2 P0 issues OR 3–5 P1 issues 🟡 Needs Work
3+ P0 issues OR 6+ P1 issues 🔴 Critical

Top Issues / 优先问题（最多 5 项）：

- [P0] **[Area]**: Brief description of issue
- [P1] **[Area]**: Brief description of issue

If more than 5 issues exist, show the 5 most severe and add: "See full findings below for all issues." If no issues found: "No critical issues found ✅"

Quick Wins / 快速修复（最多 3 项，低成本高收益）：

- **[Fix name]**: One-sentence instruction (e.g., "Add publisher.name to NewsArticle Schema")

Quick Win criteria: fixable in < 30 minutes, no code deployment required (e.g., CMS field edit, meta tag addition, file rename). If none qualify: "No quick wins identified — remaining issues require development work."

Executive Summary template / 执行摘要模板：

### Executive Summary

**Overall Health**: 🟢 Good / 🟡 Needs Work / 🔴 Critical
**Scope**: [what was audited]

**Top Issues**
- [P0] **Schema**: dateModified is earlier than datePublished
- [P1] **On-Page**: Meta description missing on article page
- [P1] **Technical**: No News Sitemap found

**Quick Wins**
- **Fix dateModified**: Set dateModified to the last-edited timestamp in your CMS Schema output
- **Add meta description**: Map the article excerpt/summary field to meta description in your theme template

Detailed Findings / 逐项检查表

Table / 表格：

Check item / 检查项 Result / 结果 Notes / 说明
(item) Pass ✅ / Fail ❌ / Manual 🔍 Issue description or fix suggestion

Priority Fix List / 优先级修复列表

  • P0 — blocks indexing or violates content policy / 影响收录或违反内容政策
  • P1 — affects rich results or EEAT signal strength / 影响富摘要或 EEAT 信号
  • P2 — best practice / 规范性

6.5 · Competitor Gap Analysis 竞争对手差距分析

Run during Full Diagnostic or when the user asks "why do competitors rank higher?" / "为什么竞争对手排名更高？"

Step 1 — Identify top competitors

Use WebSearch to search: "<article topic>" news

Extract the top 3–5 news publisher results: domain, headline, publish timestamp.

Competitors detected:
| Rank | Publisher | Headline | Published |
|------|-----------|----------|-----------|
| 1 | [domain] | [headline] | [time] |

Step 2 — Fetch competitor articles

For each competitor URL, use the three-phase fetch protocol (web_fetch → curl → Manual) to extract:

  • datePublished from Schema or page
  • NewsArticle Schema completeness: count present required fields out of 9 (@type, headline, image, datePublished, dateModified, author, publisher, publisher.logo, mainEntityOfPage)
  • Author authority: named author with linked profile page (Yes / Partial / No)

Step 3 — Compute gap metrics

Publication speed gap:

Your site: datePublished → [timestamp]
Earliest competitor: [publisher] → [timestamp]
Gap: [X minutes / hours] → [Advantage ✅ / Disadvantage ❌]

Full comparison table:

| Publisher | Schema Completeness | Author Authority | Publish Speed |
|-----------|---------------------|------------------|---------------|
| Your site | XX% (X/9 fields) | Full / Partial / None | [timestamp] |
| [Competitor 1] | XX% | Full / Partial / None | [timestamp] |
| [Competitor 2] | XX% | Full / Partial / None | [timestamp] |

Step 4 — Output gap analysis summary

## Competitor Gap Analysis

**Topic**: [search topic]
**Competitors analyzed**: [N]

### Key Gaps

**Publication Speed**
Your site: [X min after event]
Competitors avg: [Y min after event]
Gap: [difference] → [recommendation if gap > 15 min]

**Schema Completeness**
Your site: [XX%]
Competitors avg: [XX%]
Gap: [difference] → [recommendation if gap > 10%]

**Author Authority**
[Competitors provide full author profiles / Your site matches competitor standard]

### Recommendations
- [Specific action to close each identified gap]

7 · EEAT Scan 触发与输入处理

Trigger words / 触发词： EEAT 扫描 / Run EEAT scan / 扫描 EEAT / EEAT audit / EEAT 审计 / 做个 EEAT 扫描

Step 1 — Read the signal checklist / 第一步：读取检查项清单

Before scanning, read eeat-reference.md (same directory as this file) to load all 24 signal definitions.

Read: eeat-reference.md

Step 2 — Determine input type / 第二步：判断输入类型

Input Action
Live URL / 线上 URL Use web_fetch to fetch the page; extract full page HTML, JSON-LD, and visible text
Raw HTML / 原始 HTML Parse provided HTML directly; no fetch needed
URL unreachable / 无法抓取 Mark all signals that require live page inspection as 🔍 Manual; proceed with available information

Step 3 — Execute dimensional scans / 第三步：按维度执行扫描

Run the four dimensions in order: Experience → Expertise → Authoritativeness → Trustworthiness

For each signal in eeat-reference.md:

  1. Check the pass condition against the fetched page content
  2. Record result: Pass ✅ / Fail ❌ / 🔍 Manual
  3. For Fail results, note what was found and what is expected
  4. For Manual signals, note what requires manual verification

8 · EEAT 评分算法

Per-dimension score / 维度评分：

维度分 = floor(该维度通过项数 / 该维度有效项数 × 100)
  • 有效项数 = 总项数 − 🔍 Manual 项数（Manual 项从分母中排除）
  • 四个维度各自独立计算

Total score / 总分：

总分 = floor((经验分 + 专业度分 + 权威性分 + 可信度分) / 4)

Rating labels / 评级标签：

Score range Label
80–100 良好 ✅
50–79 需改进 ⚠️
0–49 差 ❌

9 · EEAT 报告格式

After completing all scans, output the report in the following structure. Use the same language as the user's prompt throughout (Chinese prompt → Chinese report; English prompt → English report).

Report template / 报告模板：

## EEAT 扫描报告

**扫描对象**：https://example.com/article/ （或 "Raw HTML input"）
**扫描日期**：YYYY-MM-DD
**总分：XX / 100**

---

### 维度总览

| 维度 | 得分 | 评级 |
|------|------|------|
| 经验 (Experience) | XX | 良好 ✅ / 需改进 ⚠️ / 差 ❌ |
| 专业度 (Expertise) | XX | ... |
| 权威性 (Authoritativeness) | XX | ... |
| 可信度 (Trustworthiness) | XX | ... |

---

### 经验 (Experience) — XX/100

| 检查项 | 结果 | 说明 |
|--------|------|------|
| 第一手内容标识 | Pass ✅ | 符合要求 |
| 经历日期明确 | Fail ❌ | 未注明具体经历时间 |
| 作者署名可见 | Pass ✅ | 符合要求 |
| 作者简介链接存在 | 🔍 Manual | 需手动核查作者页是否可访问 |
| 原创媒体 | Fail ❌ | 图片文件名含 "dall-e" |

### 专业度 (Expertise) — XX/100

| 检查项 | 结果 | 说明 |
|--------|------|------|
| ... | ... | ... |

### 权威性 (Authoritativeness) — XX/100

| 检查项 | 结果 | 说明 |
|--------|------|------|
| ... | ... | ... |

### 可信度 (Trustworthiness) — XX/100

| 检查项 | 结果 | 说明 |
|--------|------|------|
| ... | ... | ... |

---

### 行动建议

**P0 — 立即修复（影响收录或违反内容政策）**
- **[经验]** 原创媒体 — 将图片文件名中的 "dall-e" 前缀去除，上传流程中禁止保留 AI 工具名称
- **[可信度]** HTTPS — 将站点迁移至 HTTPS 并配置 301 重定向

**P1 — 应尽快修复**
- **[专业度]** 作者资质说明 — 在作者简介中补充职业背景或领域经验描述
- **[权威性]** 发布方名称缺失 — 在 NewsArticle Schema 的 publisher.name 字段填写机构名称

**P2 — 建议跟进**
- **[专业度]** 内容深度 — 文章不足 500 字，建议扩充至覆盖 3 个以上子议题

10 · Google News Diagnostic Report 完整诊断报告

Trigger: Run this section when scope is "Full Diagnostic", or when the user asks why their site is not in Google News / why articles don't rank / why competitors outrank them.

Score Aggregation / 评分聚合

Each Layer 1 and Layer 2 sub-metric below is scored 0–100 first (use the checklists in earlier sections; binary checks map to 100 / 50 / 0 as each subsection defines).

Layer 1 Score (0–100) — weights within Layer 1 sum to 100%:

Layer1 = 0.25×PublisherTrust + 0.25×AuthorAuthority + 0.25×SchemaHealth
       + 0.17×NewsSitemapHealth + 0.08×Crawlability
Sub-check Weight (of Layer 1)
Publisher Trust Score 25%
Author Authority Score 25%
Schema Health (completeness % as 0–100) 25%
News Sitemap Health Score 17%
Crawlability Score 8%

Layer 2 Score (0–100) — weights within Layer 2 sum to 100% (renormalized from 15∶10∶10∶5):

Layer2 = 0.375×Freshness + 0.25×ContentType + 0.25×TopicClusterCompat
       + 0.125×TopStoriesSignal
Sub-check Weight (of Layer 2)
Freshness Score 37.5%
Content Type Score 25%
Topic Cluster Compatibility Score 25%
Top Stories Presence (fixed mapping: Confirmed→100, Gap→0, Not triggering→50) 12.5%

Final Google News SEO Score (0–100):

Google News SEO Score = 0.6 × Layer1 + 0.4 × Layer2

（即：Layer 1 占最终总分的 60% 权重，Layer 2 占 40%；上表百分比是 层内 合成 Layer1/Layer2 时用，勿与 60/40 混淆。）

Rating labels:

Score Label
80–100 🟢 Strong
50–79 🟡 Developing
0–49 🔴 At Risk

Diagnosis Routing / 诊断结论路由

Condition Diagnosis Statement
Layer 1 score < 50 "Primary issue: This site is likely not eligible for Google News indexing. Fix Layer 1 issues before optimizing for ranking."
Layer 1 50–69 (Partial) "Site has partial Google News index presence. Resolve remaining Layer 1 gaps to achieve consistent indexing, then address Layer 2 ranking."
Layer 1 ≥ 70, Layer 2 < 50 "Site is indexed but articles are not competitive in Google News clusters. Focus on Layer 2 ranking improvements."
Layer 1 ≥ 70, Layer 2 ≥ 70 "Site and articles meet Google News baseline requirements. Competitor gap analysis shows remaining optimization opportunities."

Full Diagnostic Report Template / 完整诊断报告模板

## Google News Diagnostic Report

**Analyzed**: [URL or domain]
**Date**: [YYYY-MM-DD]
**Google News SEO Score**: XX / 100  🟢 Strong / 🟡 Developing / 🔴 At Risk

---

### Dual-Layer Scorecard

| Layer | Score | Status |
|-------|-------|--------|
| Layer 1 — Index Eligibility | XX/100 | Pass ✅ / Partial ⚠️ / Fail ❌ |
| Layer 2 — Ranking Potential | XX/100 | Pass ✅ / Partial ⚠️ / Fail ❌ |

**Diagnosis**: [statement from routing table above]

---

### Layer 1 — Index Eligibility

| Check | Score | Status |
|-------|-------|--------|
| News Index Status | — | Not Indexed ❌ / Limited ⚠️ / Strong ✅ |
| Publisher Trust | XX/100 | Pass / Partial / Fail |
| Author Authority | XX/100 | Pass / Partial / Fail |
| Schema Health | XX% complete | Pass / Partial / Fail |
| News Sitemap Health | XX/100 | Pass / Partial / Fail |
| Crawlability | XX/100 | Pass / Partial / Fail |
| URL Structure | — | Recommended ✅ / Issues ⚠️ |

---

### Layer 2 — Ranking Potential

| Check | Score | Status |
|-------|-------|--------|
| Freshness | XX/100 | Fast ✅ / Moderate ⚠️ / Slow ❌ |
| Content Type | [type] / XX pts | Breaking / Analysis / Digest / Aggregation |
| Publisher Authority | XX/100 | Strong / Moderate / Weak |
| Topic Cluster Compatibility | XX/100 | High / Medium / Low |
| Top Stories Presence | — | Confirmed ✅ / Gap ❌ / Not triggering ⚠️ |

---

### Competitor Gap Analysis Summary

[See Section 6.5 output]

---

### Priority Fix List

**P0 — Fix immediately (blocks indexing)**
- [item]

**P1 — Fix soon (affects ranking)**
- [item]

**P2 — Best practice**
- [item]

References 参考资源

  • Google News ranking factors, optimization strategies, AI content policy, News Sitemap examples, two-layer architecture model, Topic Cluster signals: 见 reference.md
  • EEAT signal definitions and priority table: 见 eeat-reference.md
Related skills

More from wghust/stark-skills

Installs
23
Repository
wghust/stark-skills
GitHub Stars
5
First Seen
Mar 4, 2026
Security Audits
Gen Agent Trust HubPass
SocketPass
SnykWarn