Blog Google: Google API Data for Blog Performance

Bundled paths below use host Markdown substitution of ${CLAUDE_SKILL_DIR}; it is not an exported shell variable. Resolve them before execution, quote paths, and refuse nonabsolute overrides. Pass blog_reference_root resolved from ${CLAUDE_SKILL_DIR}/../blog/references, blog_template_root from ${CLAUDE_SKILL_DIR}/../blog/templates, and needed sibling roots to agents. Read the main reference orchestration-details.md before loading project context; pass only its helper-fenced output to downstream agents.

Direct access to Google's SEO APIs for blog performance analysis. Provides real Chrome user metrics, indexation status, search performance, entity analysis, YouTube video discovery, keyword volumes, and PDF/HTML performance reports.

Most integrations have no usage fee within their documented quotas. Cloud Natural Language requires billing and can incur charges after its free monthly tier. Google Ads requires an eligible account and developer token. Never enable billing or make a paid request without explicit user approval.

Cost and consent gate (mandatory, read first)

This skill uses the user's Google credentials, quotas, and sometimes billing.

• What it spends: API quota on the user's Google Cloud project (small buckets: YouTube search.list has its own bucket of 100 calls per day, URL Inspection allows 2,000 per day per site, the Indexing API 200 publish requests per day); Cloud Natural Language bills after 5,000 free units a month; Keyword Planner needs a Google Ads account.
• What it writes externally: index sends a notification to Google and setup writes credential config under ~/.config/claude-seo/. Everything else is read-only.
• Consent: before the FIRST call in a session that reads account data (gsc, inspect, ga4, keywords), spends a small bucket (youtube, batch inspect), can bill (nlp), or writes (index, setup), name the command, the property or URL, the expected call count, and the quota or cost basis, then wait for an explicit yes. Free keyless-style reads such as pagespeed and crux on a public URL only need the call named.
• No uncapped loops: batch files are limited to what the user approved; stop on the first quota error instead of retrying through it.
• Skips: report each skipped or unavailable command in ONE line for the orchestrator's "Fallbacks and skips" summary, such as "Google: no credentials, skipped CrUX".

Persistent runtime directory

CLAUDE_BLOG_RUNTIME_DIR optionally selects one trusted absolute persistent root. This integration uses <root>/blog-google/.venv; NotebookLM also uses <root>/blog-notebooklm/data for its library, authentication and browser state. Without this override, the existing skill-local defaults stay in use. Google's shared credential config/token paths stay unchanged. No state is copied or migrated automatically, and ordinary runs never perform setup.

Preserve an existing validated operator override. For a plugin without one, explicitly pass the host-substituted ${CLAUDE_PLUGIN_DATA}/claude-blog-runtime path on both setup and every later runner call. This is Markdown substitution, not an exported shell variable. Execute only once it is an absolute resolved path. Single quotes keep an unresolved placeholder literal so validation rejects it instead of expanding it to a different shell path:

CLAUDE_BLOG_RUNTIME_DIR='${CLAUDE_PLUGIN_DATA}/claude-blog-runtime' python3 "${CLAUDE_SKILL_DIR}/scripts/setup_environment.py" --check --json
CLAUDE_BLOG_RUNTIME_DIR='${CLAUDE_PLUGIN_DATA}/claude-blog-runtime' python3 "${CLAUDE_SKILL_DIR}/scripts/run.py" google_auth --check --json

Only explicit setup without --check installs dependencies. Standalone users keep the default unless they explicitly choose an absolute runtime root. Relative, unresolved, or linked managed subdirectories fail closed. Changing the root selects separate state; it never authorizes a migration or cleanup of the old root. Plugin persistent data follows the host's uninstall retention rules; preserve it explicitly when uninstalling if it is still needed.

Prerequisites

Always check credentials before running any command:

BLOG_SKILLS_DIR="${CLAUDE_BLOG_SKILLS_DIR:-${CLAUDE_SKILL_DIR}/..}"
case "$BLOG_SKILLS_DIR" in /*) ;; *) echo "ERROR: skills dir must be absolute" >&2; exit 1 ;; esac
python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" google_auth --check --json

Config file: ~/.config/claude-seo/google-api.json (shared with claude-seo)

{
  "api_key": "YOUR_GOOGLE_API_KEY",
  "oauth_client_path": "/path/to/client_secret.json",
  "default_property": "sc-domain:example.com",
  "ga4_property_id": "properties/123456789",
  "ads_developer_token": "...",
  "ads_customer_id": "123-456-7890",
  "ads_login_customer_id": "123-456-7890"
}

If missing, read ${CLAUDE_SKILL_DIR}/references/auth-setup.md and walk the user through setup.

Credential Tiers

Always communicate the detected tier before running commands.

Machine-readable errors

Some legacy helpers emit an error JSON object with process exit 0 in --json mode. Preserve those exit contracts when integrating existing callers, but inspect the parsed result: any error is a failed operation. Batch results also need per-item error checks; do not treat a zero exit status alone as successful API work. A future exit-status migration needs a separately reviewed compatibility change.

Quick Reference

PageSpeed + CrUX

/blog google pagespeed <url>

Combined Lighthouse lab data + CrUX field data for a published blog post.

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" pagespeed_check <url> --json Reference: ${CLAUDE_SKILL_DIR}/references/api-reference.md

Output merges lab scores (point-in-time Lighthouse) with field data (28-day Chrome user metrics). CrUX tries URL-level first, falls back to origin-level.

/blog google crux <url>

CrUX field data only (no Lighthouse run). Faster.

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" pagespeed_check <url> --crux-only --json

/blog google crux-history <url>

CrUX History trends use 25 weekly periods by default and accept up to 40 with --periods. They show whether CWV metrics are improving, stable, or degrading.

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" crux_history <url> --json

Search Console

/blog google gsc <property>

Search Analytics: clicks, impressions, CTR, position for last 28 days.

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" gsc_query --property <property> --json Default: 28 days, dimensions=query,page, type=web, limit=1000.

Includes quick-win detection: queries at position 4-10 with high impressions.

The dedicated Search Console generative-AI reports are a gradual, subset rollout in the Search Console UI. They have separate Search and Discover views; the Search view covers AI Overviews and AI Mode. Do not promise clicks, queries, or API retrieval from these dedicated views. Until Google documents an API, report that capability as SKIPPED or unavailable and point the user to the UI.

Google's July 29 Search Central announcement says Search Console platform properties for Instagram, TikTok, X, and YouTube are globally available. The current Help Center still says gradual rollout. Report this as a Google-source conflict, verify availability in the user's account, and do not claim that /blog google gsc retrieves these platform reports through the current API.

/blog google inspect <url>

URL Inspection: real indexation status from Google.

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" gsc_inspect <url> --json

Returns: verdict (PASS/FAIL), coverage state, robots.txt status, indexing state, page fetch state, canonical selection, mobile usability, rich results.

After a canonicalization fix, Google may retain the URL in a duplicate cluster for up to two weeks. If the implementation is now correct and the fix is within that window, report PENDING_REEVALUATION rather than an immediate failure. Search Console's Request Indexing feature is quota-limited; reserve it for important URLs.

For batch inspection: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" gsc_inspect --batch <file> --json

Indexing API

/blog google index <url>

Notify Google of a URL update through the Indexing API.

Eligibility: Google's Indexing API documentation says it "can only be used to crawl pages with either JobPosting or BroadcastEvent embedded in a VideoObject" (https://developers.google.com/search/apis/indexing-api/v3/quickstart, read 2026-10-12). Ordinary blog posts are NOT eligible. Do not run this command for a blog post, and do not offer it as a way to speed up indexing of articles. For a blog post, use /blog google inspect <url> and, if needed, Search Console's Request Indexing in the UI (quota-limited).

If the user names a URL anyway, first ask whether it is a JobPosting page or a livestream (BroadcastEvent in VideoObject) page. Only if the user confirms that explicitly, say what will be sent and get a second explicit yes, then run:

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" indexing_notify <url> --confirm-eligible --json Reference: ${CLAUDE_SKILL_DIR}/references/api-reference.md

The script refuses to submit without --confirm-eligible; never add that flag on your own. Daily quota: 200 publish requests. Do not present this as a replacement for URL Inspection's Request Indexing feature.

For batch (same eligibility confirmation, applies to every URL in the file): python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" indexing_notify --batch <file> --confirm-eligible --json

GA4 Traffic

/blog google ga4 [property-id]

Organic traffic report: daily sessions, users, pageviews, bounce rate, engagement.

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" ga4_report --property <id> --json Default: 28 days, filtered to Organic Search channel group.

For top landing pages: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" ga4_report --property <id> --report top-pages --json

YouTube (Video Discovery)

YouTube research can add useful, relevant media and distribution context. Any third-party visibility correlation is observational, not a Google ranking or citation requirement. Free, API key only. Used by blog-write and blog-rewrite for video embedding.

/blog google youtube <query>

Search YouTube for videos relevant to a blog topic.

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" youtube_search search "<query>" --json Quota: search.list has its own quota bucket: by default 100 calls per day at 1 unit per call (Google's quota-cost page, read 2026-10-12). Other YouTube endpoints share a separate 10,000 units per day. Run one search per distinct need and stop at the first quota error.

Returns: title, channel, views, likes, duration, description, tags.

For video details + comments: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" youtube_search video <video_id> --json

NLP Content Analysis

Google's entity and sentiment analysis can support topic and editorial review. It does not expose ranking-system scores, and E-E-A-T is not a numeric Google ranking factor.

/blog google nlp <url-or-text>

Full NLP analysis: entities, sentiment, content classification.

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" nlp_analyze --url <url> --json Free tier: 5,000 units/month. Requires billing enabled on GCP project.

For entity extraction only: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" nlp_analyze --url <url> --features entities --json

Keyword Research (Google Ads)

Gold-standard keyword volume data. Requires Google Ads account (Tier 3).

/blog google keywords <seed>

Generate keyword ideas from seed terms for blog topic research.

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" keyword_planner ideas "<seed>" --json

For volume lookup: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" keyword_planner volume "<kw1>,<kw2>" --json

Reports

/blog google report <type>

Generate a PDF/HTML report with charts and tables.

Script: python3 "$BLOG_SKILLS_DIR/blog-google/scripts/run.py" google_report --type <type> --data <json> --domain <domain> --format pdf

Note: PDF generation needs WeasyPrint and its system libraries (Pango and Cairo). The explicit setup script installs only the Python packages, and no run installs system packages. If they are missing, the report falls back to HTML; tell the user and let them decide whether to install the libraries.

Rate Limits

Read ${CLAUDE_SKILL_DIR}/references/rate-limits-quotas.md for detailed quota management.

Blog Workflow Integration

This skill is both user-invocable (/blog google pagespeed) and callable internally by other blog sub-skills:

• blog-seo-check: Runs PSI + CrUX on published post URL for live CWV data
• blog-rewrite: NLP entity analysis to identify E-E-A-T entity gaps
• blog-geo: GSC performance data for real search appearance insights
• blog-audit: Batch CWV + indexation checks across all published blog URLs
• blog-write / blog-rewrite: YouTube search for video embedding

Falls back gracefully when credentials are not configured.

Report Templates

Use the bundled templates when a workflow requests a durable human-readable report. Keep unavailable account data marked SKIPPED; never fill an empty section with estimated metrics.

• assets/templates/cwv-audit-report.md for PageSpeed and CrUX evidence.
• assets/templates/gsc-performance-report.md for Search Analytics exports.
• assets/templates/indexation-status-report.md for URL Inspection evidence.

Technical Notes

• INP replaced FID on March 12, 2024. Never reference FID.
• CLS values from CrUX are string-encoded (e.g., "0.05"). Scripts handle parsing.
• CrUX 404 = insufficient Chrome traffic, not an auth error.
• Search Analytics data has 2-3 day lag.
• Indexing API is officially for JobPosting and BroadcastEvent-in-VideoObject pages only; blog posts are ineligible.
• Most integrations have no usage fee within quota. Cloud Natural Language requires billing and can incur charges; Google Ads requires account and developer-token access.
• Read ${CLAUDE_SKILL_DIR}/references/search-currentness.md before diagnosing a named update, canonical change, Discover visibility, Google generative-AI reporting, platform properties, Preferred Sources, AMP, or crawler byte-limit issue.
• A named update's dates do not prove what caused an individual site's change. Wait one full week after rollout before comparing data, and separate Web, Image, Video, and News performance.
• Googlebot processes only the first 2MB of supported files and first 64MB of PDFs. Keep critical metadata and primary content before the cutoff.

Error Handling