The LinkedIn Posts You Remember Are Lying to You
Why scrolling successful creators gives you the wrong idea about what works, and what to analyse instead.
Most people study successful LinkedIn creators by scrolling through their profiles and reading a few recent posts.
That feels like research.
It usually isn’t.
You remember the post with 2,000 likes. You forget the seven similar posts that went nowhere. You notice the clever opening, because it’s easy to see, but you miss the subject, posting frequency, length, timing, format, and months of audience conditioning behind it.
Then you copy the opening.
The result sounds familiar, performs badly, and makes you wonder whether LinkedIn has changed again.
I’ve done versions of this myself. One evening, while I was meant to be finishing a recruiting presentation, I spent almost an hour reading posts from three creators. I had cold coffee next to my keyboard, seventeen browser tabs open, and a half-eaten protein bar stuck to its wrapper.
By the end, I believed I had found several patterns.
I hadn’t written any of them down.
The next morning, I could remember one hook and a post about bad managers. That was the full output of my research.
This is one person’s experience, and some people are far better at close reading than I am. Still, memory is a poor place to store content analysis.
LinkedIn’s Feed Is a Terrible Spreadsheet
LinkedIn shows you content in a format designed for consumption, not comparison.
One post fills most of the screen. The metrics sit underneath it. Some text is hidden behind “see more.” Dates may be shown as relative times. You move through the profile one post at a time, interrupted by reposts, comments, featured items, and whatever LinkedIn decides to load next.
Your brain tries to build a pattern from that.
It gives extra weight to what is recent, emotionally strong, or already popular. Psychologists have studied versions of this memory bias for decades. Information that is easier to recall often feels more common or more important than it really is, even when recall is being driven by vividness rather than frequency.
A viral post is very vivid.
A run of 14 ordinary posts isn’t.
That matters because the ordinary posts often contain better information. They show which ideas the person repeatedly tests, what stops working, and whether one successful post was part of a pattern or a lucky exception.
Jonah Berger and Katherine Milkman examined nearly 7,000 New York Times articles published over three months. They found that content associated with high-arousal emotions, including awe, anger, and anxiety, was more likely to appear on the most-emailed list. Practical usefulness also mattered. Sadness, a lower-arousal emotion, was linked with less sharing.
That study wasn’t about LinkedIn, and emailing a newspaper story isn’t the same action as liking a recruiter’s post. I wouldn’t treat its findings as a recipe for LinkedIn success.
It does explain why one emotional post can dominate your memory of someone’s work.
Recruiters are especially vulnerable to this because we work with subjects that carry emotion already: rejection, salary, layoffs, bad interviews, hiring bias, recruiter ghosting. A sharp post about any of those can outperform months of calm advice.
You may conclude that anger is the creator’s content strategy.
The export might show that 70 percent of their posts are practical tips and only two angry posts performed unusually well.
That’s a different conclusion.
Export the Posts First
The best place to begin isn’t another LinkedIn profile.
Start with your own.
You already have opinions about your content. You probably believe certain topics work, that short posts perform better, or that your audience prefers advice over personal stories.
Get the data before defending those beliefs.
PostPackr, the LinkedIn Post Exporter, lets you collect posts from a LinkedIn profile or company page and export them into a clean file. Rather than keeping twenty profile tabs open, you can compare the posts in rows and columns.
Export your own content first.
Look at the post text, date, reactions, comments, reposts, and any other fields available in your export. Add a few columns yourself:
Main subject
Post format
Opening type
Approximate length
Personal story, yes or no
Clear opinion, yes or no
Don’t build a giant classification system.
I’ve made that mistake too. You start with five useful columns, then add emotional tone, audience intent, sentence rhythm, post objective, call to action, degree of controversy, and eleven other labels. After categorizing twelve posts, the file feels like unpaid data entry and you stop opening it.
Pick a few things you genuinely want to understand. Start with the small group of posts, ideally 30 or 100 for better analysis.
For my own recruiter content, I’d start with subject, opening, length, and whether the post gives the reader something practical. That’s enough to challenge several assumptions without turning the exercise into academic research.
Then sort.
Which subjects repeatedly beat your normal results? Which posts attracted comments rather than passive likes? Did long posts fail, or did several weak long posts fail? Are personal stories working because they’re personal, or because they contain a clear conflict that your other posts lack?
The numbers won’t answer every question. They’ll remove a few bad explanations.
That’s useful.
Once you have examined your own content, export posts from two or three recruiters whose work reaches the audience you want to reach. I wouldn’t analyze thirty creators. At that point you’re mixing different audiences, career levels, countries, and content goals into one file.
A corporate recruiting page isn’t comparable to an independent recruiter writing about candidate experience. A recruiter with 200,000 followers isn’t a clean comparison for someone who began posting six weeks ago.
Even when two creators discuss the same subject, the audience history beneath those posts may be completely different.
I don’t know how much of LinkedIn performance can be separated from existing audience trust. I haven’t seen convincing public data that assigns a reliable percentage to content quality versus distribution, network size, past engagement, or account history. Anyone giving you a neat formula is making assumptions.
Still, repeated patterns inside one account can tell you plenty.
A Popular Post Can Still Teach You Nothing
When people export competitors’ posts, they usually sort by reactions and begin studying the top ten.
I understand the temptation.
It’s also where the analysis can become misleading.
The highest-performing post might concern a layoff, a new job, a personal loss, a public controversy, or an announcement the creator will never repeat. The post performed because of the event, not because the first line had seven words.
Sometimes the event and writing are impossible to separate.
A recruiter once showed me a competitor’s post that had received several thousand reactions. He wanted to reproduce its structure for an employer-branding post. The original was about the creator losing a family member.
There wasn’t a structure worth borrowing.
The attention came from human concern, prior relationships, and the situation itself. Turning that into a corporate writing pattern would have been strange, though LinkedIn contains enough strange content that someone has probably tried it.
Instead of examining only the winners, compare groups.
Take five posts on the same subject that performed well and five that didn’t. Look for differences. Maybe the successful posts made a sharper claim. Maybe they included a recognisable recruiting situation. Perhaps the weaker posts were generic and could have been written by anyone.
One popular post tells you what happened once.
A repeated difference gives you a hypothesis.
Notice the word hypothesis. You aren’t proving why a post worked. LinkedIn doesn’t give you controlled test conditions, and public engagement metrics don’t show impressions, dwell time, profile visits, saves, or private messages.
Visible reactions can be a poor measure of commercial value too.
A hiring manager might read your post, remember your name, and contact you three months later. That post may look average in your spreadsheet. Meanwhile, a joke about interview clichés could receive 1,500 likes and produce no useful conversation.
I still pay too much attention to visible reactions. Most people do. They’re immediate, public, and reassuring, while reputation builds in ways that don’t fit neatly in a column.
I don’t have a satisfying fix for that.
Competitors Give You Patterns, Not Permission
Studying another creator should change what you notice, not erase your own judgment.
Suppose you export 150 posts from a recruiter whose content performs consistently. You find that their strongest posts often begin with a firm opinion, use a real hiring situation, and stay under 250 words.
That doesn’t mean your next post should imitate their sentences.
It gives you questions to test:
Are your openings too cautious?
Do your posts discuss recruiting in general terms rather than showing what happened?
Are you adding another 400 words after the useful part has ended?
The distinction matters because writing style is more than sentence length and punctuation. Researchers studying linguistic accommodation have found that people naturally adjust subtle features of their language during interaction, including function words, syntax, and utterance length. In a large Twitter dataset, Cristian Danescu-Niculescu-Mizil, Michael Gamon, and Susan Dumais found evidence that people shifted elements of their style in response to one another. bsorb parts of other people’s writing without trying.
Read one creator every day for six months and their rhythms may begin appearing in your drafts. That isn’t always conscious copying. Language is social.
Direct imitation creates a different issue. You copy the visible habits while missing the thinking that produced them.
A creator writes short sentences because they remove every unnecessary explanation. You copy the line breaks but keep the unnecessary explanations. Now the post is twice as long vertically and no clearer.
Or you borrow their confrontational tone without having the experience to support it. The post sounds bold until someone challenges you in the comments.
Recruiters know what happens next.
The author either disappears or replies with a sentence that reveals the original claim was built on very little.
There’s a whole separate discussion about the ethics of training writing tools on someone’s public posts. I’m not settling it here. Publicly visible text can be studied, but publishing near-copies under your own name is still copying, regardless of which tool rearranged the words.
The Awkward Work After the Export
The export is easy.
Reading it properly is where most people quit.
By the way, you can and should use AI to spot things you might otherwise miss!
You have a spreadsheet with hundreds of posts. The first fifteen rows feel interesting. By row sixty, every recruiter seems to be writing about candidate experience, salary transparency, AI, and why hiring processes are broken.
Your categories get inconsistent. You mark one post as “career advice” and a similar post as “job search.” You begin wondering whether reposts should count. Then an email arrives and the analysis remains unfinished for three weeks.
I’d rather complete a rough review of fifty posts than design a perfect method for five hundred.
Choose one question.
For example: Why do this recruiter’s posts about candidate rejection outperform mine?
Filter their posts to that subject. Compare the strong and weak ones. Read the comments, because the comment section often shows which sentence people reacted to. Then write down four or five observations in plain language.
Not commandments.
Observations.
You might notice that the creator rarely discusses rejection as a policy issue. They describe a specific moment: a candidate waiting after a final interview, a hiring manager delaying feedback, or a recruiter having to deliver news they disagree with.
You might also find that their best post breaks every rule you believed. It’s long, starts slowly, and has no call to action.
Keep that contradiction.
Content analysis should make your assumptions less tidy. When every finding confirms what you believed before opening the spreadsheet, you’re probably sorting the evidence around your opinion.
There’s a cost. Reviewing posts takes time, and the work becomes stale as the creator and platform change. A pattern from last year may no longer hold. A popular format attracts copies until readers become tired of seeing it.
Berger has made a similar point about online sharing: once everyone applies the same headline pattern, it can lose part of its effect. What once produced arousal becomes familiar. That’s another reason to study the thinking underneath a post rather than cloning its surface.
I’ll keep exporting my own content because my memory keeps giving me a flattering version of what worked. I’ll study a few other recruiters too, though probably fewer than I initially planned, because the categorizing part still gets boring around row sixty.
Turn Exported LinkedIn Posts Into Writing Instructions
You can identify patterns manually, but the spreadsheet eventually becomes too large to hold in your head. The section below contains the exact Claude process I’d use to study one creator’s exported posts, produce a style report, and draft new material without asking Claude to rewrite or closely copy any original post.