How Does The LinkedIn Algorithm Work?

What you will learn

In this article, I am going to give you a big-picture overview of the different elements of the LinkedIn algorithm.

This will help you to create the right, high-quality content signals that are essential to generate a ton of views, engagement and ultimately business on LinkedIn.

How does the LinkedIn algorithm work?

In order to understand how to go viral on LinkedIn, it is important to understand how the LinkedIn algorithm works.

LinkedIn has a massive user base of 500+ million users.

To keep them happy it is essential to create a great user experience on a daily basis.

Besides connecting and communicating with other people, one of the most popular features on LinkedIn is the newsfeed with user-generated content.

Unfortunately, not all content is great, and not all topics are relevant to everyone.

To create a pleasant user experience, LinkedIn has to crunch the numbers and quickly identify:

  • which content is of high quality
  • which content is relevant for who

If you want to learn more about the technology and software behind it, check out these three articles on LinkedIn engineering, here, here and here.

Important note: before you begin creating viral LinkedIn posts, make sure your LinkedIn profile is up to date and optimized so you can make the perfect first impression with your new LinkedIn connections. You can check out my how to write the perfect LinkedIn headlines + 39 examples article if you need some guidance.

LinkedIn Activity Graph

One of the core building blocks of the LinkedIn algorithm is the Activity Graph.

LinkedIn looks at user-generated content in terms of activities.

What types of activities contribute to the algorithm

Each piece of content has a subset of attached activities, whenever someone is interacting with a piece of content in any way:

  • liking
  • commenting
  • sharing

Activities can become content themselves.

  • Tim liked a text-only post.
  • Jane commented on an article.

Different priorities for different content types

At the root of each activity, we have content.

  • User-generated content by people and pages you follow and indirectly via #hashtags and predicted interests
    • LinkedIn Pulse article
    • Posts with attached
      • text-only
      • photo
      • video
      • document
      • link
        • internal or preferred domains such as
        • external
  • User activities and events
    • liking
    • commenting
    • sharing
    • connecting with others
    • new job
    • birthday
    • completing a course
  • Promoted content
    • internal
      • LinkedIn Learning courses
      • LinkedIn Premium
    • external
      • paid ad campaigns by companies
  • Curated content
    • hand-picked articles by the LinkedIn editors team
  • Dynamic content
    • recommend jobs for you

Each content type has a different internal rating in terms of feed priority.

Hand-picked articles by the LinkedIn editors team and paid ad campaigns have a much higher probability of showing up in other peoples’ newsfeeds, than user-generated content.

Within the space of user-generated content, different content types have different priorities as well. In late 2017, text-only posts were the only content type that would allow you to be seen by as many people as possible.

In early 2018, LinkedIn switched priorities and started to prioritize video content over text-only posts in the newsfeed to push and promote the launch of their new, native video platform.

As of today, there is a stronger balance between the different content types, with a focus on video and text-only posts.

How does LinkedIn rank content?

A great online community such as LinkedIn is always attractive to spammers.

Spam always creates a bad user experience. As such, it is one of LinkedIn number one priorities to quickly identify and eliminate spam and low-quality content from their newsfeed.

Let’s say someone has created an article on LinkedIn. Some users are starting to engage with it.

  • Article

Each of these elements could contain or promote spam in specific circumstances.

If an article was identified as spam, LinkedIn would no longer display it in everyone’s newsfeed. It also has to make sure that attached activities such as “James liked a spam article” are suppressed as well.

On the other hand, if the article is ok, we wouldn’t want to suppress it, just because some of the comments contained spam.

LinkedIn is using recursive and graph-based algorithms to identify spam within this tree-like structure. Each node receives a quality score and is resolved. If a top node is identified as spam, all subsequent subnodes are suppressed as well.

Spam and low-quality content to avoid

Now, what exactly is low-quality spammy content to avoid?

Let’s start with the obvious, anything that is harmful or dangerous to other users:

  • External links that lead to
    • viruses and malicious code
    • phishing sites
    • scams
    • get-rich-quick schemes
    • pyramid schemes
    • MLM
    • ICOs

Then we have content that is generally considered to be Not Safe For WORK / NSFW

  • porn
  • violence
  • gambling
  • etc.

Next, we have low-quality SEO articles.

  • very short, 300 to 500 words long
  • Often stolen from other content creators and respun with software.
  • The better ones are written by people in sweatshops in third world countries
  • frequently contain a ton of links to other sites, they’re trying to promote

What is respinning? A sentence such as “Dogs like to eat sausages.” could be restructured by breaking sentences down into their building blocks: “{Dogs|Many dogs|Some dogs} {like|enjoy|prefer|love} {to eat|to consume|to feed on} {sausages|meat|…}.

Many SEO experts use this experts to quickly and cheaply generate a lot of new looking content. They steal other peope’s blog post, run them through a spinning software and then publish it on one of their websites to generate backlinks for other sites.

These texts usually sound stupid. The idea is to fool search engines such as Google into thinking that many relevant websites and pages are linking to a target page, such ranking the target page higher on Google. Fortunely, with the help of AI and machine learning, this technique becomes less and less effective.

Then we have content that isn’t really spam but still of little or no value to readers. This could be

  • a press release
  • an article was written with the best intentions by someone who doesn’t know how to
    • research the topic
    • write engaging copy
    • structure the text with paragraphs and sections
    • use images and visuals to break up text
    • create a pleasant reading experience

Bad content signals

Without going into too much detail, here are some of the techniques

  • keyword and phrase matching
  • Artificial Intelligence and machine learning to understand texts and recognize patterns
  • external databases of domains that have been flagged and blacklisted
  • anti-virus software to detect malicious code

Identifying low-quality content can be challenging if none of the above techniques create a match.

For these cases, LinkedIn is using user behavior to train their AI system.

  • If someone clicks on a link, do they instantly click the back button
  • Do people like, comment and share the post?
  • Do people leave original, meaningful comments?
  • How do new people react to the content? Sometimes, spammers use their own engagement pods to give low-quality content the appearance of quality. By showing the content in question to people who are not connected to the content creator, it becomes much more likely to get objective feedback. If they don’t like it, they won’t react.
  • How much effort do people put into their activities? For example, do they spend the time to write a unique summary when they share a post with their network?
  • How does the engagement deviate from the baseline engagement rate? Let’s say, that the average LinkedIn post, received a like to view rate of 1%. If 100 people see a post, 1 person would like it. If the post had been shown to 300 people and less than 300 people liked it, the post might be considered low-quality, because of the below average like rate.

Content creators who consistently receive low-quality scores on their content receive a negative quality score. That means that whatever new content they publish, the system will assume that it is probably of low-quality and will show it to fewer people slower.

Produce high-quality content

In the same way that LinkedIn identifies spam and low-quality content, it also identifies high-quality content.

Content creators who consistently receive good individual content ratings will receive a positive quality score. In other words, LinkedIn is giving these creators the benefit of the doubt, if a post doesn’t receive a high number of engagements in a short period of time and will show it to a larger group of people first, before coming to a definite conclusion.

Here is a list of activities that will help you to get a high-quality content status.

  • Content consistently
    • not marked as spam
    • low quality
  • A high amount of engagement above baseline
    • likes
    • comments
    • shares
  • High-quality engagement
    • long comments
    • long comments with sub-comments
    • shares with a text summary
  • Follow-up engagement
    • A “share” of a post receives a high amount of quality engagement as well
  • Readers or viewers decide to follow or connect with a content creator

Content Relevancy

LinkedIn tries to predict what kind of content users are interested in.

It uses a combination of self-classification and indirect learning via machine learning technology.

LinkedIn is classifying each piece of content into different themes and topics. You can imagine this as invisible hashtags that are attached to each post, once the content has been analyzed by the LinkedIn algorithm.

The system will then show the post to different users. If a user is engaging with the content, it will then record the topic and activity and store it in the user profile.

Example: I read a post about “content marketing” and liked it. LinkedIn would then add “content marketing” to my interests and record my engagement. The more I engage with other “content marketing” content, the more certainty that I am actually interested in this.

The next time LinkedIn has a choice to display “content marketing” or “cat pictures” to me, it will probably pick the first one.

Relationship Relevancy

Similar to Content Relevancy, the relationship between the content creator and the content consumer is very important.

Every time someone interacts with someone else’s content, this is recorded in the system with three primary data points.

  • Type of interaction and intensity?
    • like
    • comment
    • share
    • connected
    • followed
    • exchanged messages
    • wrote recommendations
    • endorsed
    • work in the same company
  • With who or whose content?
  • When?

You can think of this as a decay function. If I interacted with someone yesterday, the relationship score would be high. A higher score makes it more likely, that the LinkedIn algorithm would show me their next-days content first.

Let’s say I responded to someone’s content six months ago, but then never again. The system would then decay the relationship score, with the result that the person will show up less and less in my newsfeed. Kind of a one-hit wonder band. 🙂

Make the LinkedIn algorithm work for you

The LinkedIn algorithm and all of its quality indicators are complex, BUT if you have the right strategy and plan in place, it’s possible to get into LinkedIn’s good books!

Here is what you have to focus on if you want to generate high-quality leads on LinkedIn with content marketing.

  • Understand how to avoid low-quality content.
  • Integrate as many high-quality content signals into every piece of content as possible.
  • Create an effective content strategy that guarantees a high degree of topic and relationship relevancy for your target audience.
  • Have an effective sales funnel in place that guides your audience towards your website.
  • Build an email list.

I have achieved amazing results for myself by using systematic planning and a well thought-out content strategy.

If you need a hand to gain traction on LinkedIn and to understand how to always hit the right buttons, get in touch to learn how I can help you.