Almost Timely News: 🗞️ Bringing the LinkedIn Algorithm Guide to Life With AI (2025-05-25) :: View in Browser

The Big Plug

👉 Grab your copy of the Unofficial LinkedIn Algorithm Guide for Marketers, newly refreshed!

Content Authenticity Statement

100% of this week’s newsletter was generated by me, the human. You will see bountiful AI outputs in the video. Learn why this kind of disclosure is a good idea and might be required for anyone doing business in any capacity with the EU in the near future.

Watch This Newsletter On YouTube 📺

Click here for the video 📺 version of this newsletter on YouTube »

Click here for an MP3 audio 🎧 only version »

What’s On My Mind: Bringing the LinkedIn Algorithm Guide to Life With AI

I recommend you watch the video version of this newsletter to see the AI outputs and process.

This past week, I let you know about the freshly revised Unofficial LinkedIn Algorithm Guide for Marketers. It’s a compilation of all the data LinkedIn releases about how its systems work.

Lots of people downloaded it and commented on it, expressing their appreciation. That’s fine, but I still wonder what people are DOING with it. It’s not intended to be shelfware, put on a shelf or on a to-read list that never gets read. It’s meant to be used.

The guide is highly prescriptive, with explanations about why things matter and what you should do about it, but… we all know that times are busy. Time itself is in short supply.

Which got me thinking, what would it look like to actually use this thing, to walk through some practical use cases for it? So let’s do that today. Let’s put the guide’s contents into practice in ways that are meaningful and tangible.

Part 1: Mise en Place

Here’s a poorly hidden secret about the guide. Unsurprisingly, it’s written as much for AI as it is for humans. Yes, the technical explanations are there so that enterprising or skeptical folks can check out the bona fides of the guide, but let’s be honest, almost no one checks sources any more. We can barely get people to read headlines, much less dig deep into the guts of an explanation.

No, the technical language in there is more for the machines than it is for the humans.

So with that, let’s tackle a very specific scenario. Let’s say you’re someone who’s looking for work. You know the kinds of companies you want to work for, and maybe there are even specific people that you’re thinking about trying to influence, trying to attract the attention of.

Who are those people? What do they talk about?

Our first step in our mise en place is to gather that information. Let’s pretend I didn’t know my CEO and partner, Katie Robbert. Let’s say I was an intrepid job seeker and I wanted to get her attention, hopefully get my posts into her feed, get LinkedIn to recommend me as someone to check out.

I’d first want to know – from the guide – what language Katie is all about. What’s in her profile, what’s in her posts, what’s in her comments. Then I’d want to know who she interacts with, who she sees currently in her feeds, and what they’re about.

Now, there are any number of legit and less-legit tools that can do this sort of data extraction, but I’ll give you the absolute simplest:

1. Open up LinkedIn on your mobile device.


2. Turn on screen recording.


3. Scroll through Katie’s profile at a moderate pace.


4. Scroll through the posts, comments, and connections that interact with Katie and vice versa.


5. Scroll through the obvious first degree connections of hers she interacts with.


6. Turn off screen recording.


7. Upload the video to any generative AI tool that can see video.


8. Have generative AI transcribe the video.

Here’s a simple transcription prompt for this.

I’ve attached a screenshare of me browsing the profile and activities of Katie Robbert. Transcribe the LinkedIn profile of Katie Robbert. Ensure you have the complete profile transcribed as displayed in the video. Then transcribe the text of Katie Robbert’s posts and comments in the order displayed in the video. Then transcribe the profiles of the people shown in the video, Brooke Sellas and Danielle Blackwell. Then transcribe their posts and comments, organized by person.

In just a few steps, you’ve extracted all the relevant information you need to do this analysis.

You’ll want to do the same to yourself. Scroll through your profile. Scroll through who you interact with, what you see in your feed, what comments you leave. Perform the same process.

Now you’ve got two corpuses of data: yours, and your subject of interest.

Part 2: Extraction and Analysis

Open up the generative AI tool of your choice and use the best reasoning model you have access to (Gemini 2.5, o3, Claude 4, DeepSeek R1, etc.). Put in the Trust Insights LinkedIn guide.

Start with this prompt and your transcribed data from Part 1.

Let’s perform a semantic analysis of my LinkedIn profile, activities, and connections. Using the transcript I’ve included plus the knowledge from the Unofficial LinkedIn Algorithm Guide for Marketers, assess how the LinkedIn algorithm sees me, mimicking to the best of your ability the systems described in the guide. Explain the language I use, the topics I engage with, and how LinkedIn’s systems perceive me based on the guide. Rank the topics in descending order by prevalence, with the associated language I use for each, and your explanation of how you did your analysis.

Then perform the exact same process on the transcribed data about Katie from Part 1.

What you should have are detailed analyses of these pools of data, arranged in the way that the LinkedIn systems see it, as semantic information and embeddings.

Part 3: Comparison

Fundamentally, what happens under the hood at LinkedIn is an analysis of our semantic space – all the things we say and do – compared to the semantic space of the rest of the people in our network and their network. Part of the decision systems behind the LinkedIn feed are to try matching up people whose spaces are similar, on the premise that like attracts like. Topics that I post about, if your activities are similar to mine, are probably topics you’d engage with.

What we’re trying to do is effectively the same thing. Part of LinkedIn’s new systems use LLMs, language models like LiRank and LiGNN to perform this task at massive scale. We’re replicating it in foundation LLMs like ChatGPT’s o3, Gemini 2.5, etc.

Our next step is to compare the two semantic analyses of my profile and Katie’s profile.

Here’s a sample prompt:

Using my semantic profile and Katie’s semantic profile, compare and contrast the two. Where do Katie Robbert and I ove