Google Updates Sentinel

Google Updates Sentinel — Algorithm Monitoring and Recovery

Purpose

SYNTHEX runs a Sentinel system that monitors site health, tracks Google algorithm updates, and alerts when performance regressions correlate with known rollouts. This skill documents how to interpret Sentinel data, diagnose algorithm impacts, and execute recovery strategies.

Sentinel Architecture

Alert Types and Thresholds

From DEFAULT_THRESHOLDS in lib/sentinel/types.ts:

Update Type Reference Guide

Core Updates

• What it targets: broad quality, relevance, and authority reassessment
• Symptoms: gradual ranking shifts across multiple pages; can be positive or negative
• Diagnosis: compare GSC analytics 28 days before vs after rollout start; check page-level changes
• Recovery: improve content quality, demonstrate E-E-A-T, add original research; recovery typically occurs at next core update (3-6 months)
• Timeline: rollout 2-4 weeks; recovery at next core update cycle
• Known updates in fallback: March 2024, August 2024, November 2024, December 2024, March 2025, June 2025, December 2025, March 2026, May 2026

Spam Updates

• What it targets: scaled content abuse, expired domain abuse, parasite SEO, link manipulation
• Symptoms: sudden deindexing of affected pages; manual action in Search Console
• Diagnosis: check for manual actions in GSC; review content for auto-generated or scraped patterns
• Recovery: remove or substantially rewrite flagged content; submit reconsideration request
• Timeline: rollout 1-2 weeks; recovery after reconsideration (weeks to months)
• Known updates in fallback: March 2024, June 2024, December 2024, August 2025, March 2026, June 2026, August 2026, September 2026
• Policy: site reputation abuse policy updated 28/08/2026; scaled content abuse covers mass generation without added value, AI-assisted or not [VERIFIED]

Helpful content (no longer a separate update)

• Status: since the March 2024 core update Google assesses helpfulness inside its core ranking systems: "There's no longer one signal or system used to do this" [VERIFIED: developers.google.com/search/blog/2024/03/core-update-spam-policies]. There are no standalone helpful content updates to wait for.
• Diagnosis: treat helpfulness drops as core-update impact; audit for pages that add no unique value, are thin, or target keywords without genuine expertise
• Recovery: improve or remove unhelpful pages; show firsthand experience; E-E-A-T is guidance, not a specific ranking factor [VERIFIED]
• Retracted: an earlier fallback list carried an invented "February 2026 helpful content update"; it never existed and is hidden by RETRACTED_SEED_IDS

Link Spam Updates

• What it targets: unnatural link patterns, private blog networks (PBNs), guest post networks, paid links
• Symptoms: links nullified (no manual action, but rankings drop as link equity disappears)
• Diagnosis: audit backlink profile for unnatural patterns; check for sudden loss of referring domains
• Recovery: disavow toxic backlinks; focus on earning editorial links through quality content
• Timeline: rollout 1-2 weeks; recovery gradual as new editorial links are earned
• Known updates in fallback: none as a separate named update since March 2024 (link spam is handled by spam updates; an invented "December 2025 link spam update" was retracted)

Other dashboard incidents

• February 2026 Discover update is in the fallback as other; the live feed maps non-ranking incidents (crawling, indexing, serving) to gssd-<id> rows with low impact

Product Reviews / Local Updates

• What it targets: review quality, firsthand experience signals, local result accuracy
• Symptoms: product review pages or local listings shift in rankings
• Diagnosis: check review content for genuine firsthand experience; verify local listing accuracy
• Recovery: demonstrate firsthand experience with products; ensure GBP listing is complete and accurate

Diagnosis Workflow Decision Tree

Sentinel alert triggered
  |
  +--> What alert type?
         |
         +--> algorithm-update (correlated with active rollout)
         |      |
         |      +--> Identify the update type from AlgorithmUpdate.updateType
         |      |      core / spam / helpful-content / link-spam / other
         |      |
         |      +--> Follow type-specific recovery strategy (see above)
         |      |
         |      +--> Cross-reference: which pages lost the most?
         |             fetchAnalytics with dimensions: ['page']
         |             Compare 28 days before vs during rollout
         |
         +--> traffic-drop (no active rollout)
         |      |
         |      +--> Technical issue investigation:
         |      |      1. Check server errors in coverage (GSCSnapshot)
         |      |      2. Check robots.txt for accidental blocks
         |      |      3. Check for recent deploys that may have broken pages
         |      |      4. Check CWV regression (SiteHealthSnapshot.coreWebVitals)
         |      |
         |      +--> Competitive investigation:
         |             Invoke competitive-local-strategy skill
         |             Check if competitors published superior content
         |
         +--> ranking-drop
         |      |
         |      +--> Single page or site-wide?
         |      |      Use fetchAnalytics dimensions: ['page'] to identify
         |      |
         |      +--> Single page: content or technical issue with that URL
         |      +--> Site-wide: algorithm impact or domain-level issue
         |
         +--> crawl-error-spike
         |      |
         |      +--> Check coverage report details
         |      +--> Server errors: check hosting, SSL, CDN
         |      +--> Redirect errors: audit redirect chains
         |      +--> Soft 404s: fix page content or return proper 404
         |
         +--> cwv-regression
                |
                +--> Which metric failed?
                |      LCP >3.0s: check image sizes, server response, render-blocking resources
                |      INP >300ms: check JavaScript execution, main thread blocking
                |      CLS >0.25: check layout shifts from ads, images without dimensions, dynamic content
                |
                +--> Run PageSpeed Insights for detailed recommendations
                       SiteHealthReport.coreWebVitals from health-checker.ts

Proactive Hardening Checklist

Protect against future algorithm impacts:

• E-E-A-T on every page — author bios, credentials, firsthand experience, citations
• Core Web Vitals passing — LCP <2.5s, INP <200ms, CLS <0.1 (good thresholds, not just passing)
• No link schemes — no purchased links, no PBN links, no excessive reciprocal links
• AI content human-reviewed — all AI-generated content edited and verified by a human
• No fake or gated reviews — genuine reviews only, never incentivised with discounts
• Accurate schema markup — LocalBusiness, FAQ, HowTo as appropriate; no misleading markup
• Mobile parity — mobile version has same content as desktop
• Sentinel + GSC monitor running daily — automated detection of regressions
• Regular content audit — remove or improve thin, outdated, or duplicate pages quarterly

Health Score Interpretation

Computed by computeHealthScore() in lib/sentinel/health-checker.ts:

Score Ranges

Monitoring Cadence

Adding New Algorithm Updates

The live feed picks up new Google Search Status Dashboard incidents automatically. Only update the offline fallback, and only from the dashboard itself (never from a blog or memory):

1. Add entry to KNOWN_ALGORITHM_UPDATES in lib/sentinel/algorithm-feed.ts with the dashboard's UTC begin/end and status.search.google.com/incidents/<id> URL
2. Include: name, updateType, announcedAt, rolloutStart, rolloutEnd (null if ongoing), impactLevel, description, sourceUrl
3. The seedAlgorithmUpdates() function will upsert it on next cron run
4. Set rolloutEnd once Google confirms the rollout is complete

Reference skill: This is a read-only architecture guide — it documents existing systems and does not generate creative or code output. No capability uplift block is needed.

Foundation & Gate Wiring (SYN-1049)

Adopted from the senior-skill standard so every artefact this connector produces is checked against the locked foundation before it lands.

Reads at every invocation (never cached — re-read each run):

• .claude/memory/ceo-foundation.md — AI-search realism, verification gates for impact claims, universal taboos.
• .claude/memory/verification-gates.md — gate state for any claim referenced.

Output gate: every client-facing artefact this connector produces routes through brand-voice-enforce before the CEO batched-review queue. A REJECT blocks the artefact until the quoted offending string is fixed.

Evidence standard: every quantitative or factual claim carries exactly one tag — [VERIFIED] / [INFERENCE] / [UNCONFIRMED]. Untagged = defect (.claude/rules/fabel-evidence-standard.md). Never state a projected result as fact.

Spec: see spec.md in this skill directory.