I Built a Claude Skill to Show Me Why My LinkedIn Reach Collapsed

It scrapes your LinkedIn analytics, shows in-network vs. out-of-network reach, and reveals which posts actually break out of your bubble.

Anyone building a personal audience through content has, at some point, been whacked by algorithmic changes. This is happening on LinkedIn all the time as they attempt to deal with the onslaught of AI-generated content.

The Pangram Labs study (published July 9, 2026) scanned over 1 million posts across five platforms. LinkedIn was the most AI-saturated by a wide margin: over 40% of long-form posts (250+ words) were fully AI-generated, and LinkedIn accounted for 62% of all AI-flagged content across the five platforms studied.

It’s not just LinkedIn; I stopped really investing in X when the algo update no longer rewarded the kind of content I enjoyed posting. We’ve recently seen our YouTube channel take a huge hit as the algorithm has changed to account for AI-generated videos.

And so, this month, I’ve seen a huge drop in the performance of my posts on LinkedIn, and I thought, huh, I wonder why. LinkedIn’s content analytics suck. So, I got busy and built my own dashboard featuring a new metric that I believe is the most important to track if you want to grow an audience on that platform.

In this post, I’ll walk you through it, and you can download a templatized version of the dashboard and a Claude Skill to build your own version.

See the live demo dashboard →

What You Get

A single interactive dashboard that lives inside Claude as an Artifact you can update anytime, populated by a single Claude skill that uses Claude computer.

The interactive dashboard features:

• Monthly trend charts for impressions, reach, followers, gained, and engagements.

• Click any month to see an overview of performance for that month, broken out at the post level.

• Top 10 leaderboards broken out by impressions, engagements, and virality.

• In-network vs. out-of-network reach so you can see which posts break out of your bubble. THIS is the most important metric for LinkedIn that no one tracks. The dashboard makes it easy.

• A 🔥 viral badge on posts where more than half the reach came from outside your network

• Full post content you can click to expand and re-read

A living dashboard inside Claude that updates every time you run the skill.

---

What You Need

• The Claude desktop app (download here)

• Chrome with Claude connected (Claude desktop app → Settings → Connectors → Chrome)

• Node.js installed on your machine (download here if you don’t have it)

• A LinkedIn account you’re logged into in Chrome

Step 1: Install the Skill

The skill is a single file that teaches Claude how to scrape your LinkedIn analytics. It handles everything: creating files, scraping data, building the dashboard.

1. Download SKILL.md from the GitHub repo: github.com/searchbrat/linkedin-dashboard (it’s in the skill/ folder)

2. Open the Claude desktop app

3. Go to Settings → Skills → Add Skill

4. Select the SKILL.md file you downloaded

One file. One install. The skill is called linkedin-dashboard-updater, and it shows up in your Skills list immediately.

Step 2: Populate Your Dashboard

Open a new Cowork session in Claude. Make sure Chrome is open, and you’re logged in to LinkedIn.

Then say:

“populate July 2026”

Claude will:

1. Create all the necessary data files automatically

2. Open your LinkedIn analytics page in Chrome

3. Find every post you published that month

4. Visit each post’s analytics and scrape the numbers: impressions, engagements, followers gained, members reached, in/out network percentages

5. Build your dashboard and save it as a live Artifact you can access anytime

The whole process takes a while depending on how many posts you published that month.

Step 3: Keep It Updated

Want to add another month? Open a new Cowork session and say:

“populate August 2026”

Want to refresh the numbers on a previous month (metrics keep changing as posts get more views)?

“update July 2026”

You can do both at once:

“populate August 2026 and update July 2026”

The skill figures out the right thing to do. “Populate” discovers new posts and scrapes them. “Update” refreshes the metrics on posts it already knows about. Each time, your Artifact dashboard gets rebuilt with the latest data.

The reason you’d want to update previous months is that content on LinkedIn has a long tail, so you’ll find that old posts continue to gain impressions and engagement. I’d suggest always updating the previous month when populating a new month, e.g., when I populate for August 2026, I’ll also update July’s numbers.

The Chrome Extension Quirk

One thing you’ll run into. While Claude is scraping your LinkedIn analytics, the Chrome extension can sometimes lose its connection. You’ll notice Claude pausing or saying it can’t read the page.

The fix is simple: click the Claude extension icon in Chrome to re-activate it, then tell Claude to continue. You might need to do this a couple of times during a scrape, especially if you have a lot of posts. It’s annoying, but Claude will pick up from where it left off each time.

Also, unfortunately, you can’t simply run this as a cron job to update content every month. You need to have Claude desktop open, be logged into LinkedIn on Chrome.

How does it work?

The skill uses Claude’s Chrome automation to navigate LinkedIn like you would. It goes to your analytics page, scrolls through posts, clicks into each one. It reads the numbers from the page, structures them into JSON, and saves them to a data file.

A build script then reads that JSON and generates pure inline SVG charts. Everything is baked into a single HTML file that becomes your Artifact.

You own all the data. It’s JSON files on your machine. Import them into spreadsheets, feed them back to Claude for deeper analysis, or do whatever you want with them.

---

How this helps you do better at LinkedIn

There is so much great analysis you can do with this dashboard, let me go through 3 great ones:

1. In-Network vs Out-of-Network Reach: I remember feeling like there was a similar drop in my posts performance around February. But, I wasn’t sure, until I saw this chart:

How much of your posts engagement are coming from within your followers (In) vs. outside (Out). The more coming from outside, the better you’ll perform. You’ll see July-26 is trending to be a poor month, similar to Feb-26. This tells me LinkedIn, for some reason, is showing my posts to less people from outside of my followers.

2. Chat with Live Artifact: Chat to your live artifact and gather insights about the data it’s showing you. For example, I want to know why July is performing so poorly on ‘In-Network vs. Out-of-Network” vs. June. After going back and forth it told me:

• My last 3 posts were to ‘inside baseball’ targeting my existing followers vs. people who’d sit outside my network. I loosely agree. I do believe LinkedIn changed something 7 days ago.

• It pulled out a couple of posts in June’26 that would appeal to a much broader audience and gave me ideas of experiments to run.

• It said to cut the self-promotion lol. I don’t believe this is accurate, it’s because in July I referenced things I built. It marks them as self-promotion.

It’s incredibly helpful to chat about the data and gather insights. The fact all post text is added to the interactive dashboard allows you to derive real insights from writing patterns that are working vs. not working.

3. Top Leaderboards: I love this tab, it allows you to see your top 10 posts across impressions, engagements, and virality (posts with > 50% out-of-network engagements). Add the chat feature and you can gather insights into what is working, not working and what’s changed.

---

The key to getting better at anything, including growing your audience on LinkedIn, starts with looking at the right data & insights.

» Get the skill and preview the dashboard

Until Next Time,

Happy AI’fying

Kieran