click-recovery

name	click-recovery
version	1.1
description	Find pages Google trusts but users ignore — high impressions, low CTR. Get a prioritized list of titles and meta descriptions to rewrite, then publish directly to Webflow. Triggers: click recovery, improve CTR, low CTR, fix CTR, wasted impressions, title optimization, meta description optimization, SERP optimization. Requires: Google Search Console MCP server, Webflow MCP server. Optional: Keywords Everywhere API for volume/intent validation. Workflow: Analyze → Recommend → Approve → Publish to Webflow. Modes: /click-recovery:analyze (report only), /click-recovery (full workflow with publishing).

Click Recovery Skill

Find pages where Google already ranks you but users scroll past. These are your fastest wins — no need to build authority, just fix the pitch.

Prerequisites

• Required: Google Search Console MCP server
• Required: Webflow MCP server (for publishing)
• Optional: Keywords Everywhere MCP server (needs API key — for search volume and intent data)

Skill Modes

This skill supports two modes for flexibility:

Mode	Command	Use Case
Analyze	/click-recovery:analyze	Report only — no changes made. Returns CTR analysis and recommendations.
Full	/click-recovery	Complete workflow: analyze → recommend → approve → publish.

Workflow Overview

[MODE SELECT] → FETCH → FILTER → VALIDATE → PRIORITIZE → RECOMMEND → [APPROVE → PUBLISH]

Phases in brackets only run in full mode. After presenting recommendations, ask user which pages to update. On approval, publish directly to Webflow CMS.

Threshold Rationale

These thresholds are based on industry benchmarks and can be adjusted:

Threshold	Value	Rationale
Min impressions (page)	500	Enough data for statistical significance
Min impressions (query)	100	Lower bar for query-level insights
Low CTR	< 2%	Below typical position 1-10 averages
Position cutoff	≤ 20	Realistic click potential; beyond page 2 clicks are rare
Ranking opportunity	21-50	Visible to Google but not users
Min search volume	50/month	Below this, effort may not be worth it

---

Conditional Guards (Global)

These guards apply throughout the skill execution.

⚡ GUARD — Load SEO Copilot config: At the start of execution, check if .claude/seo-copilot-config.json exists:

• If yes: Load and apply settings:
  • brandVoice.tone → Match tone in title/description rewrites
  • brandVoice.avoid → Filter out forbidden words from suggestions
  • seo.metaTitleFormat → Apply brand suffix/prefix to titles
  • audience.primary → Frame messaging for target audience
• If no: Proceed with defaults, note: "No SEO Copilot config found. Run /getting-started for personalized recommendations."

⚡ GUARD — User requests abort: If user says "stop", "cancel", "abort", or "nevermind" at any phase:

• Confirm: "Stop the workflow? Progress will be lost."
• If confirmed: Exit cleanly with summary of what was completed
• Output any partial results (e.g., analysis done but not published)

⚡ GUARD — GSC MCP unavailable: This skill requires GSC data. If unavailable:

• Inform user: "Click Recovery requires Google Search Console access. Please connect GSC MCP and try again."
• Stop execution

⚡ GUARD — Webflow MCP unavailable: If Webflow MCP is not connected:

• Run analysis and recommendations (Phases 0-4)
• Skip execution phase
• Inform user: "Webflow MCP not connected. Recommendations generated but cannot publish. Connect Webflow MCP and run again, or update manually."

⚡ GUARD — Keywords Everywhere unavailable: If KE API is not available:

• Proceed without volume/intent data
• Note in report: "Search volume and intent data unavailable. Recommendations based on GSC impressions only."
• Suggest user add KE API for richer insights

---

Phase 0: Mode Selection & Prerequisites

0.1 Choose Mode

At the start, determine which mode to run:

How would you like to run Click Recovery?

1. **Full workflow** (recommended) — Analyze, recommend, and publish approved changes
2. **Analyze only** — Generate report with recommendations, no publishing

• If Full workflow or no mode specified: Run Phases 0-5
• If Analyze only: Run Phases 0-4, output report, stop

0.2 Prerequisites Check

Search for these tools BEFORE starting analysis:

GSC MCP (required):

• Search: +gsc search analytics
• If missing: Stop and inform user: "Click Recovery requires GSC MCP. Install from: https://github.com/sofianbettayeb/gsc-mcp-server"

Webflow MCP (required for publishing):

• Search: +webflow data cms
• If missing in full mode: Switch to analyze-only mode automatically. Note: "Webflow MCP not connected — running in analyze-only mode."

Keywords Everywhere (optional but check):

• Search: +keywords everywhere volume
• If missing: Continue, but ADD to report header:

  ⚠️ Keywords Everywhere not connected.
  Recommendations based on GSC impressions only.
  Connect KE for search volume and intent data:
  https://github.com/hithereiamaliff/mcp-keywords-everywhere
  

---

Phase 1: Fetch GSC Data

1.1 Identify Target Site & Set {domain}

Ask the user which property to analyze if multiple GSC properties are available. Extract the domain from the selected property URL (e.g., https://www.checklist-seo.com/ → checklist-seo.com).

1.2 Review Activity Log

Check .claude/reports/{domain}/activity-log.md:

• If it exists: read the last 10 entries and surface a brief summary:

  Recent activity on {domain}:
  - 2026-02-10 /weekly-report — Week W06 report. Health: STABLE.
  - 2026-02-05 /click-recovery — Updated meta titles on 5 pages.
  

• Redundancy check: if /click-recovery was run in the last 7 days → warn: "You ran /click-recovery on [date]. CTR changes typically take 2-3 weeks to show in GSC. Run again anyway?"
• Content refresh check: scan all entries for /refresh-content runs in the last 14 days. Extract any page URLs mentioned. Store as recently_refreshed_pages — used in Phase 5.4 to warn before overwriting titles that were just set by a full content refresh.
• If the log doesn't exist: proceed silently

Use recent activity as context (e.g., if /weekly-report flagged specific CTR issues, prioritize those pages).

1.3 Pull Performance Data

Use GSC MCP to fetch the last 90 days of data:

Query-level data:

• All queries with impressions > 100
• Include: query, page, impressions, clicks, CTR, position

Page-level data:

• All pages with impressions > 500
• Include: page, impressions, clicks, CTR, average position

1.4 Calculate Benchmarks

Compute site-wide averages:

• Average CTR by position bracket (1-3, 4-10, 11-20)
• Overall average CTR
• Median impressions per page

These benchmarks help identify underperformers relative to the site's own performance.

---

Phase 2: Filter Opportunities

2.1 CTR Opportunity Filter

Flag pages/queries meeting these criteria:

High-impression, low-CTR pages:

• Impressions > 500
• CTR < site average for that position bracket
• Position ≤ 20 (realistic click potential)
• OR Position 21-30 with impressions > 1000 (high-volume ranking opportunities)

Wasted impression queries:

• Impressions > 100
• CTR < 2% (or < half the position-bracket average)
• Position ≤ 10 (should be getting clicks)

Ranking opportunities (separate category):

• Position 21-50 with impressions > 1000
• These pages rank but need content improvement to reach page 1
• Flag for /refresh-content rather than just meta tag fixes

2.2 Quick Win Filter

Prioritize opportunities where small changes yield big impact:

• Position 1-3 with CTR < 5% → title/description mismatch with intent
• Position 4-10 with CTR < 2% → being outcompeted by better snippets
• High impression queries not reflected in current title → keyword gap

2.3 Group by Page

Consolidate query-level findings to page level:

• List all underperforming queries per page
• Sum total wasted impressions per page
• Identify the primary keyword (highest impressions) for each page

2.4 Fetch Current Meta Tags from Webflow

Do this BEFORE making recommendations. For each flagged page URL:

1. Parse the URL to extract collection slug and item slug
2. Use Webflow MCP get_collection_list to find collections
3. Match URL path to collection + item
4. Fetch the CMS item using list_collection_items with slug filter
5. Use get_collection_details to identify meta title and meta description fields
6. Store current values:
   • Current meta title (or name field if no dedicated meta title)
   • Current meta description (or post-summary if no dedicated field)

For static pages (not in CMS):

1. Use Webflow MCP list_pages to find the page by path
2. Use get_page_metadata to fetch current SEO title and description
3. Mark as "static page" for different update method later

⚡ GUARD — Meta description field missing in collection: If the collection has no meta description field:

• Use post-summary or similar if available
• If no suitable field exists: note in report "No meta description field. Consider adding meta-description field to this collection."
• Still generate recommended meta descriptions for manual use

Store all current values — these are needed for the before/after comparison in recommendations.

⚡ GUARD — No opportunities found: If filtering returns zero pages:

• Inform user: "No significant CTR opportunities found. Your titles and meta descriptions are performing well relative to your positions."
• Suggest: "Consider running /refresh-content on older articles to improve rankings first, then re-check CTR."

⚡ GUARD — Low traffic site: If total impressions < 1000 in 90 days:

• Warn user: "Limited data available. Results may not be statistically significant."
• Proceed but note confidence is lower

---

Phase 3: Validate with Keywords Everywhere

IMPORTANT: Do not silently skip this phase. Either:

1. Use KE data to enrich recommendations, OR
2. Explicitly note in the report that KE was unavailable (see Prerequisites Check)

If Keywords Everywhere MCP is available:

3.1 Enrich Query Data

For each flagged query, fetch:

• Monthly search volume
• CPC (indicates commercial intent)
• Competition score
• Trend data (rising/falling)

3.2 Intent Classification

Classify each query by intent:

• Informational: how, what, why, guide, tutorial
• Commercial: best, top, review, vs, compare
• Transactional: buy, price, discount, deal, near me
• Navigational: brand names, specific product names

3.3 Validate Demand

Filter out queries with:

• Search volume < 50/month (unless highly commercial)
• Falling trend > 30% YoY (dying queries)

Flag high-value queries:

• Volume > 1000/month
• CPC > $2 (commercial intent)
• Rising trend

---

Phase 4: Prioritize & Recommend

4.1 Ranking Model

Rank pages by CTR opportunity using these signals (higher = do first):

1. Wasted impressions — pages with the most impressions getting no clicks
2. CTR gap — how far below the expected CTR for their position bracket
3. Position — positions 1–10 have higher recovery potential than 11+
4. Keyword value — volume × CPC as a tiebreaker when KE is connected
5. Quick fix — title missing the primary keyword = bump up

4.2 Rank Opportunities

Sort pages by rank, group into buckets using CLAUDE.md criteria:

• Must do — page CTR is less than half the expected rate for its position (confirmed in GSC) AND impressions > site average
• High value — meaningful CTR gap with 50+ weekly impressions, or primary keyword missing from title on a high-impression page
• Nice to have — small CTR gap, low impressions, or marginal improvement potential

4.3 Generate Recommendations

For each page, generate title and meta description suggestions following these guidelines:

Titles (≤60 characters):

• Lead with primary keyword
• Include a compelling hook (number, year, power word)
• Match search intent (informational = "Guide", commercial = "Best", etc.)
• Differentiate from competitors in SERP
• Apply brand suffix from config if set (e.g., "| Brand Name")

Meta descriptions (≤155 characters):

• Include primary keyword naturally
• Add a clear value proposition
• Include a call-to-action or curiosity hook
• Mention specifics (numbers, timeframes, outcomes)

Intent-based framing:

• Informational: "Learn how to...", "Complete guide to...", "X steps to..."
• Commercial: "Best X for [use case]", "X vs Y: Which is...", "Top X in [year]"
• Transactional: "Get X today", "Free X", "X% off", "Starting at $X"

4.4 Avoid AI Writing Tell-Tales

Phrases to NEVER USE in titles/descriptions:

• "Ultimate Guide" (overused)
• "Everything You Need to Know"
• "game-changer" / "revolutionary"
• "In [Year]" at the start (put year elsewhere)
• "Discover how..."
• "Unlock the secrets..."

Patterns to AVOID:

• Identical structure across all suggestions ("Best X for Y" for every title)
• Generic value props: "save time and money"
• Exclamation marks in descriptions
• Starting every description with the brand name

Human signals to ADD:

• Specific numbers: "47 templates" not "dozens of templates"
• Honest qualifiers: "for small teams" instead of "for everyone"
• Casual confidence: "The only X you'll need" (sparingly)
• Outcome focus: "Rank higher in 30 days" not "Improve your SEO"

4.4 Output Report

Present findings as a structured report:

# Click Recovery Report

**Site**: [property URL]
**Period**: Last 90 days
**Total wasted impressions**: [sum across all flagged pages]
**Estimated recoverable clicks**: [impressions × expected CTR improvement]

**Data sources**:
- ✅ GSC: Connected
- ✅ Webflow: Connected [or ⚠️ Not connected — analysis only]
- ✅ Keywords Everywhere: Connected [or ⚠️ Not connected — no volume/intent data]
- ✅ SEO Copilot Config: Loaded [or ℹ️ Not found — using defaults]

[If KE not connected, add:]
> ⚠️ **Keywords Everywhere not connected.** Recommendations based on GSC impressions only.
> For search volume and intent data, connect KE: https://github.com/hithereiamaliff/mcp-keywords-everywhere

---

## Must Do

### 1. [Page Title]
**URL**: [page URL]
**Current performance**:
- Impressions: [X] | Clicks: [X] | CTR: [X]% | Avg Position: [X]
- Expected CTR for position: [X]% | CTR gap: [X]%

**Top underperforming queries**:
| Query | Impressions | CTR | Position | Volume | Intent |
|-------|-------------|-----|----------|--------|--------|
| ...   | ...         | ... | ...      | ...    | ...    |

**Diagnosis**: [Why is CTR low? Title mismatch? Weak description? Competitor snippets?]

**Recommended changes**:

| Field | Current | Suggested | Chars |
|-------|---------|-----------|-------|
| Meta title | [current title] | [suggested title] | [N]/60 |
| Meta description | [current desc or "❌ Missing"] | [suggested desc] | [N]/155 |

**Keyword framing**: [advice based on intent]

**Estimated impact**: +[X] clicks/month

**Page type**: [CMS item / Static page]

---

[Repeat for each Must Do page]

## High Value
[Summary table with key metrics, current titles, suggested titles]

## Nice to Have
[Summary table with key metrics]

## Ranking Opportunities (Position 21-50)

These pages have high impressions but need more than meta tag fixes to reach page 1:

| Page | Impressions | Position | Recommendation |
|------|-------------|----------|----------------|
| [title] | [X] | [X] | Run `/refresh-content` for full optimization |

---

## Next Steps

1. Review and approve Must Do recommendations
2. Publish approved changes directly (this skill handles it)
3. For ranking opportunities: run `/refresh-content [URL]`
4. Monitor CTR in GSC after 2-4 weeks
5. Re-run `/click-recovery` monthly

4.5 Competitor SERP Analysis (Optional)

For Must Do pages, offer to analyze the SERP:

1. Web search the primary keyword
2. Extract titles and descriptions of top 5 results
3. Identify patterns (what hooks are they using?)
4. Find differentiation opportunities

⚡ GUARD — Analyze-only mode: If running in analyze-only mode (/click-recovery:analyze):

• Output the report
• End with: "Analysis complete. Run /click-recovery (full mode) to publish approved changes."
• Stop here — do not proceed to Phase 5

---

Phase 5: Execute

5.1 User Approval

After presenting the report, ask the user:

Which pages would you like to update?
1. All Must Do pages ([N] pages)
2. Select specific pages
3. Review only — don't publish yet

If user selects option 3, stop here. They can run the skill again or use /refresh-content manually.

5.2 Map Pages to CMS Items

For each approved page:

1. Extract the URL path from the GSC data
2. Use Webflow MCP get_collection_list to find all collections
3. Match the URL path to a collection slug and item slug
4. Use list_collection_items to fetch the CMS item by slug
5. Store the item ID and current field values

⚡ GUARD — Page not found in CMS (static page): If a URL doesn't match any CMS item, check if it's a static page:

1. Use Webflow MCP list_pages to search for the page by path
2. If found as a static page:
   • Use update_page_settings to update SEO title and meta description
   • Include in the publish workflow
3. If not found anywhere (external URL or deleted page):
   • Inform user: "[URL] not found in Webflow. It may be external or deleted."
   • Skip this page and continue with others

For static pages like homepage / or /contact:

• These are updated via update_page_settings, not CMS
• The page must be republished after updating settings

5.3 Map CMS Fields

For each CMS item, identify the correct fields using get_collection_details:

• Meta title field: may be meta-title, seo-title, titile, or the main name field
• Meta description field: may be meta-description, seo-description, post-summary, or similar

Store the field mapping for updates.

5.4 Present Changes for Final Approval

Before publishing, show the exact changes with warnings:

## Ready to Publish

| Page | Field | Current | New | Status |
|------|-------|---------|-----|--------|
| [title] | Meta title | [old] | [new] | ✅ 52 chars |
| [title] | Meta title | [old] | [new] | ⚠️ 68 chars — will truncate |
| [title] | Meta title | [old] | [new] | ⚠️ 25 chars — too short |
| [title] | Meta description | [old] | [new] | ✅ 142 chars |
| ... | ... | ... | ... | ... |

Publish these changes to Webflow? (yes/no)

Title length warnings:

• ⚠️ > 60 chars: "Title will be truncated in search results. Consider shortening."
• ⚠️ < 30 chars: "Title is short — missing opportunity to include keywords or hooks."
• ✅ 30-60 chars: Optimal length

Meta description warnings:

• ⚠️ > 155 chars: "Description may be truncated."
• ⚠️ < 70 chars: "Description is short — add more value proposition."
• ✅ 70-155 chars: Optimal length

If any warnings exist, ask user: "Some titles/descriptions have length issues. Proceed anyway, or revise first?"

⚡ GUARD — Page recently refreshed by /refresh-content: For each page in the approval list, check if its URL appears in recently_refreshed_pages (built in Phase 1.2):

• If yes: add a row-level warning in the table: "⚠️ /refresh-content ran on [date] — meta title was updated as part of a full content refresh"
• Before publishing that specific page, ask: "This page was recently refreshed — overwriting its title may undo that change. Proceed with this page anyway? (yes/no/skip)"
• If user says skip: remove the page from the publish batch and continue with others

Wait for explicit user confirmation before publishing.

5.5 Update CMS Items

On confirmation, for each approved page:

1. Call Webflow MCP update_collection_item with the new meta title and description
2. Log success or failure for each item

5.6 Publish Changes

After all items are updated:

1. Ask user: "Publish changes now or save as drafts?"
2. If publish: call publish_collection_items for all updated items
3. If draft: inform user changes are saved but not live

5.7 Summary

Present a final summary:

## Click Recovery Complete

**Updated**: [N] pages ([N] CMS items, [N] static pages)
**Published**: [Yes/No — saved as drafts]

| Page | Type | Status | Title Updated | Description Updated |
|------|------|--------|---------------|---------------------|
| [title] | CMS | ✅ Published | ✅ | ✅ |
| [title] | Static | ✅ Published | ✅ | ✅ |
| [title] | CMS | ⚠️ No desc field | ✅ | ❌ Add field manually |
| [title] | — | ❌ Not found | — | — |

## Re-check Schedule

| When | What to check |
|------|---------------|
| **2-3 weeks** | Verify changes are indexed (`site:URL` in Google). Check GSC for crawl errors. |
| **4-6 weeks** | Measure CTR impact. Compare clicks before/after. Note any position changes. |
| **Monthly** | Re-run `/click-recovery` to find new opportunities as rankings shift. |

## Next Steps
- For ranking opportunities (position 21+), use `/refresh-content [URL]`
- For pages missing meta description field, add field in Webflow Designer

---

Error Handling

Error	Action
GSC MCP not connected	Stop and instruct user to connect GSC
GSC property not found	List available properties, ask user to select
No data for date range	Try shorter range (28 days), warn about limited data
Keywords Everywhere API error	Proceed without KE data, note in report
Rate limit hit	Pause, retry with exponential backoff
Webflow MCP not connected	Run analysis only, skip publishing
Page URL not found in CMS	Check if static page via list_pages, use update_page_settings
Page not found anywhere	Skip page, may be external or deleted
CMS field not found	Try common field name variations, ask user if still not found
No meta description field in collection	Note in report, suggest adding field, still provide recommendations
Static page update fails	Check page permissions, may need to publish site
Publish fails	Show error, check if item is in draft-only state
Meta title > 60 chars	Warn user, suggest shortening before publish
Meta description > 155 chars	Warn user, suggest shortening before publish

---

Success Metrics

After implementing recommendations, track at each milestone:

2-3 weeks (indexing check):

• Are updated pages indexed? (site:URL in Google)
• Any crawl errors in GSC?

4-6 weeks (impact measurement):

• CTR change per page (compare to pre-update baseline)
• Click change per page
• Position stability (ensure rankings held or improved)

Monthly (ongoing):

• Total organic traffic change
• New CTR opportunities from ranking shifts
• Re-run /click-recovery to catch new issues

---

Integration with Other Skills

Click Recovery focuses on meta tags (title + description) for CTR optimization. Refresh Content does full content refreshes (body, keywords, FAQs, schema, internal links).

When to use each:

Scenario	Use
Low CTR, content is fine	/click-recovery — fix the pitch, not the content
Outdated content, rankings dropping	/refresh-content — full content overhaul
Both CTR and content issues	Run /click-recovery first for quick wins, then /refresh-content for deeper fixes

Workflow:

1. Run /click-recovery monthly to catch CTR opportunities
2. Approve and publish meta tag updates directly
3. For pages needing more than meta fixes, run /refresh-content [URL]

---

Activity Log

After every execution, append a row to .claude/reports/{domain}/activity-log.md.

Determining {domain}: Extract from the GSC property URL used for analysis (e.g., https://www.checklist-seo.com/ → checklist-seo.com).

If the file doesn't exist, create it with the header:

# Activity Log — {domain}

| Date | Skill | Summary |
|------|-------|---------|

Then append a row:

| YYYY-MM-DD | /click-recovery | [one-line summary: e.g., "Analyzed 45 pages. Updated meta titles on 5 pages, descriptions on 3. Published to Webflow."] |

Log even if the skill exits early due to a guard — note why (e.g., "Aborted: GSC MCP unavailable"). Log in analyze-only mode too (e.g., "Analyze only — 8 CTR opportunities found, no changes published.").