Author: Vicky Lalwani
Profile: https://www.linkedin.com/in/vicky-lalwani/
I built a Claude skill that tells me if my blog will rank BEFORE I write it.
No more "let's publish and see what happens."
No more wasted 3,000-word posts that never rank.
No more guessing which topic to prioritize.
It scores every blog across 8 dimensions.
And it's connected to everything through MCP.
Here's what it does:
1/ Paste a blog title, brief, or full draft
2/ The skill auto-detects which stage it's at
3/ It pulls live data through MCP:
→ Semrush MCP (volume, KD, CPC, top 10 DA)
→ Windsor.ai GSC (current ranking, impressions)
→ GA4 (traffic patterns, conversions)
→ Client knowledge files (brand voice, cannibalization map, existing blogs)
Then it reasons through 8 dimensions using Opus 4.7 adaptive thinking:
✦ SEO score /10
✦ AEO-GEO score /10 (how AI engines will cite it)
✦ Competition score /10 (inverted, so lower competition = higher score)
✦ Time-to-Rank in specific days (not ranges)
✦ Internal Link score /10
✦ Cannibalization risk (LOW / MEDIUM / HIGH)
✦ Conversion potential /10
✦ Effort-to-Reward bonus
Output: one of three verdicts.
→ SHIP — publish as-is, overall score above 8
→ REVISE — fixable issues, specific fixes generated
→ KILL — kill it before you waste the 8 hours
If the verdict is REVISE, the skill doesn't just say "improve SEO."
It says: "Re-target to [specific lower-KD keyword]. Add FAQ schema with these 6 questions pulled from your GSC data. Add internal link to [exact URL] with anchor text [exact phrase]. Expand to 3,000 words to match top-10 depth."
Named pages. Named keywords. Named word counts.
No fluff. No generic advice.
And if I run it on a full draft, it goes further.
ENHANCE mode generates ADD and REMOVE suggestions with paste-ready copy.
Every ADD includes exact location, paste-ready text, and which score dimension it lifts.
Every REMOVE quotes the exact phrase so my developer can find it in the file.
Cross-vertical contamination? Flagged MANDATORY.
Brand voice violations? Flagged MANDATORY.
All inside one Claude window.
No switching to Semrush. No exporting from GSC. No spreadsheet math.
Most marketers argue about "will this rank?" in meetings.
I get the answer in 90 seconds.
Here's what came out when I scored my client's 12-blog calendar last week:
→ 1 SHIP verdict (quick win, 45 days to rank)
→ 9 REVISE verdicts (each with specific fixes)
→ 2 KILL verdicts (cannibalization with existing pages)
The KILL verdicts alone saved 16 hours of wasted content production.
Want the complete skill?
1️⃣ Comment "FORECAST" below
2️⃣ Connect with me (add "FORECAST" in your connection note so I can send it faster)
3️⃣ Like this post
♻️ Repost if someone in your network is still publishing blogs and praying they rank in 2026.
#learninpublic #buildinpublic #SEO #ContentStrategy #ClaudeAI #Opus47
No more "let's publish and see what happens."
No more wasted 3,000-word posts that never rank.
No more guessing which topic to prioritize.
It scores every blog across 8 dimensions.
And it's connected to everything through MCP.
Here's what it does:
1/ Paste a blog title, brief, or full draft
2/ The skill auto-detects which stage it's at
3/ It pulls live data through MCP:
→ Semrush MCP (volume, KD, CPC, top 10 DA)
→ Windsor.ai GSC (current ranking, impressions)
→ GA4 (traffic patterns, conversions)
→ Client knowledge files (brand voice, cannibalization map, existing blogs)
Then it reasons through 8 dimensions using Opus 4.7 adaptive thinking:
✦ SEO score /10
✦ AEO-GEO score /10 (how AI engines will cite it)
✦ Competition score /10 (inverted, so lower competition = higher score)
✦ Time-to-Rank in specific days (not ranges)
✦ Internal Link score /10
✦ Cannibalization risk (LOW / MEDIUM / HIGH)
✦ Conversion potential /10
✦ Effort-to-Reward bonus
Output: one of three verdicts.
→ SHIP — publish as-is, overall score above 8
→ REVISE — fixable issues, specific fixes generated
→ KILL — kill it before you waste the 8 hours
If the verdict is REVISE, the skill doesn't just say "improve SEO."
It says: "Re-target to [specific lower-KD keyword]. Add FAQ schema with these 6 questions pulled from your GSC data. Add internal link to [exact URL] with anchor text [exact phrase]. Expand to 3,000 words to match top-10 depth."
Named pages. Named keywords. Named word counts.
No fluff. No generic advice.
And if I run it on a full draft, it goes further.
ENHANCE mode generates ADD and REMOVE suggestions with paste-ready copy.
Every ADD includes exact location, paste-ready text, and which score dimension it lifts.
Every REMOVE quotes the exact phrase so my developer can find it in the file.
Cross-vertical contamination? Flagged MANDATORY.
Brand voice violations? Flagged MANDATORY.
All inside one Claude window.
No switching to Semrush. No exporting from GSC. No spreadsheet math.
Most marketers argue about "will this rank?" in meetings.
I get the answer in 90 seconds.
Here's what came out when I scored my client's 12-blog calendar last week:
→ 1 SHIP verdict (quick win, 45 days to rank)
→ 9 REVISE verdicts (each with specific fixes)
→ 2 KILL verdicts (cannibalization with existing pages)
The KILL verdicts alone saved 16 hours of wasted content production.
Want the complete skill?
1️⃣ Comment "FORECAST" below
2️⃣ Connect with me (add "FORECAST" in your connection note so I can send it faster)
3️⃣ Like this post
♻️ Repost if someone in your network is still publishing blogs and praying they rank in 2026.
#learninpublic #buildinpublic #SEO #ContentStrategy #ClaudeAI #Opus47
Images:
