Table of Contents
Do not index
Is everyone on LinkedIn writing their posts with AI now? More people than on any other platform, and they are paying for it in reach. A Pangram Labs study of more than 1 million posts found LinkedIn generates nearly twice the AI-written content of any platform, with more than 40% of long-form posts fully machine-written. Tech Times, covering the study, reported that LinkedIn posts made up roughly one third of scanned content but accounted for 62% of all AI-flagged material. A separate Originality.ai study found AI-generated LinkedIn posts received an average of 45% less engagement than human-authored posts. So yes, the feed really is flooded, and the people doing the flooding are getting less back for it every month.
I run a content team that writes for founders and operators every day, so let me be specific about who should read this. This is for founders running personal-brand content on LinkedIn, for ghostwriters charging $5k to $30k per month who need to defend what clients pay for, and for agency owners between $200k and $2M in revenue whose pipeline depends on the feed still rewarding them. This is not for people who post twice a year, and it is not for teams whose LinkedIn presence is legal-approved product announcements. If you are still deciding whether LinkedIn matters to your business at all, this article will not change your model.
The strangest finding in the study is not the volume. It is the pattern. LinkedIn usage is all-or-nothing. Only 4.3% of long-form content was AI-assisted or mixed, the lowest share of any platform Pangram measured. People are not using AI to polish drafts. They are handing the entire post to a machine and publishing it under their own name. Sit with that for a second. The one platform where your face, your job title, and your employer appear next to every word is the platform where people most completely outsource their voice.
What the LinkedIn AI content study actually found
Two numbers from the coverage deserve your attention. The first is 62%. LinkedIn contributed about a third of the posts Pangram scanned but nearly two thirds of everything flagged as AI-written. That is not a platform with an AI problem. That is the AI problem concentrated on one platform. The second is 45%. That is the average engagement gap Originality.ai found between AI-generated and human-authored LinkedIn posts. If your content pipeline is machine-written, you are running the same race as everyone else while carrying a 45% weight.
My favorite detail is smaller. Tech Times noted that LinkedIn VP Laura Lorenzetti published an announcement post about fighting AI slop, and Pangram's model flagged that very post as AI-generated. When the platform's own anti-slop messaging reads as machine-written, you learn something important. The bar for sounding human is not high. Most people have simply stopped trying to clear it.
Here is why this is good news if you sell writing or do your own. The 40% who fully outsourced their voice are not your competition anymore. They are your tailwind. Every machine-written post in the feed makes a specific, opinionated, recognizably human post easier to spot. Scarcity is doing the positioning work for you.
How to use AI for LinkedIn posts without losing reach
Here is what I would actually do, and what my team does daily. I call it the Byline Test. If your name and face sit next to the post, the thinking in it has to be yours. Everything else is negotiable. AI can pull research, structure an outline, tighten a bloated paragraph, and catch the sentence that does not land. It cannot supply the opinion, the client story, or the number from your own P&L, because those are the only parts of a post a reader cannot get from anyone else.
In practice the Byline Test works in a fixed order. You write the core claim and the story in your own words first, even if the draft is flat. The machine compresses and arranges. Then you make the final pass yourself and put back the phrasing that sounds like you, because that phrasing is the fingerprint both readers and detection models notice. Human first, machine middle, human last. The study says almost nobody works this way. That 4.3% assisted share is the emptiest lane on the platform, and it happens to be the one that keeps your voice and your speed at the same time.
This also changes what you should measure. Impressions on machine-written content are hollow because the engagement discount arrives whether you see it in your dashboard or not. I wrote a longer breakdown of how to measure LinkedIn success beyond the analytics dashboard if you want the full argument.
A word for the ghostwriters reading this. Your clients saw this study too, or they will soon. The ones paying $5k a month and up are going to start asking which part of the work is human. If your answer is a prompt and a paste, that retainer has a shelf life measured in quarters. If your answer is that you own the thinking and use machines only for mechanics, this study is the best sales asset you have had in years.
The question worth sitting with is about trajectory. Reach on LinkedIn is being redistributed right now, away from the 40% who outsourced their voice and toward the people who still sound like themselves. Redistribution phases are short. The founders and agencies who hold their voice through this one will spend the next two years compounding an audience their competitors handed away one generated post at a time.
