Does AI-generated content hurt your reach on LinkedIn? If it reads generic, yes, badly. Originality.AI studied 3,368 posts from 99 influential profiles in 2025 and classified 53.7% of long posts as likely AI-generated. Those likely-AI posts underperformed human-written ones in most professional sectors, by up to roughly 80% in strategy and innovation topics. But the mechanism is not an AI detector. It is behavioral: posts without a clear perspective fail the engagement signals, and the feed buries anything with weak signals. That distinction changes everything about how you should use AI.
Generic AI posts get buried, and they should. The penalty is behavioral (weak engagement signals), not a detector. The line that matters is AI-generated (topic in, generic out, auto-post) versus AI-assisted (your real ideas in, your voice out, human gate before publishing). One loses. The other is how I got 4M impressions in under a year.
Yes: the data is ugly, and I am not going to spin it
I build an AI content tool. It would be convenient for me to tell you the penalty is a myth. It is not.
Originality.AI's numbers say more than half of long LinkedIn posts are now likely AI-written, and that they lose head-to-head against human writing in most professional sectors. The one category where AI won was Leadership & Inspiration, which tells you something about how much original thought that category ever contained.
Zoom out and the pool itself is shrinking. AuthoredUp analyzed over 3 million LinkedIn posts and found median impressions fell 47% year over year: 1,211 in June 2024 down to 636 in May 2025. More posts chasing less attention, and a growing share of those posts written by the same handful of models saying the same things.
So when a founder tells me "I heard AI content kills your reach," I do not argue. Most of it does. Most of it deserves to.
No: it is not an AI detector. It is a perspective detector.
Here is the part the scary headlines miss. LinkedIn does not need to detect AI. Your readers do it for free.
The feed runs on early behavioral signals: do people stop scrolling, do they read to the end, do they comment, do they share. A post that opens with "In today's fast-paced business landscape" fails every one of those tests within seconds. Not because a classifier flagged it. Because nobody has ever stopped scrolling for a sentence like that.
Strip the AI question away and the rule underneath is old: content with no point of view gets no engagement, and content with no engagement gets no distribution. AI just made it possible to produce pointless content at industrial scale. The tool did not create the penalty. It industrialized the thing that was always penalized.
This is why the same study that shows an 80% gap in strategy topics does not mean "AI text is radioactive." It means generic is radioactive, and pure AI generation is a generic factory.
We have watched this movie before
There is now more AI-generated content online than human-written content. And head-to-head, it performs worse. Ethan Smith laid this out on Lenny's podcast: his team ran a controlled test of AI-generated content against original content in search results, and the AI-generated pages came back inferior. Not slightly worse at the margins. Inferior.
I lived through the last version of this. Circa 2012, SEO link farms were winning. People duplicated millions of pages, stuffed keywords, bought backlinks, and gamed the algorithm profitably, right up until Google cut the middlemen out in what felt like one day. Everyone who built on gaming lost everything. Everyone who built on substance kept compounding.
The identical dynamic is playing out with AI content. The platforms' incentive is to surface what humans actually want to read. Every algorithm update moves in that direction. Betting on volume of generic output is betting against the house, and the house has a perfect record.
The distinction that actually matters
The question "should I use AI for content" is badly framed. There are two workflows hiding inside it, and they produce opposite outcomes.
| AI-generated | AI-assisted | |
|---|---|---|
| Input | A topic typed into a box | Your real ideas and events: voice notes, calls, commits |
| Voice | The model's default voice, same as everyone else's | Your extracted voice, learned from your own writing |
| Perspective | Averaged consensus, no position taken | Your take, your numbers, your scars |
| Publishing | Auto-blast on a schedule | Human gate: you approve every post before it ships |
| Typical outcome | Weak signals, buried by the feed | Performs like strong human writing, because the substance is human |
AI-generated content is losing and will keep losing. AI-assisted content will win. Not because the words are magic, but because AI removes the production bottleneck while the human supplies the one thing the feed actually rewards: authenticity, originality, a way of thinking that belongs to a person.
My own math
I dogfood this daily, so here is my math, shown in full. I have captured 180 voice-note seeds: raw thoughts recorded while walking or driving, each one a real event or a real opinion from my week. My AI pipeline drafts posts from those seeds in my extracted voice, and nothing publishes without my explicit approval. That workflow produced 4 million LinkedIn impressions in under a year.
Every one of those posts was AI-assisted. Zero were AI-generated in the topic-box sense. The ideas, the numbers, the failures in them are mine. The AI did the drafting labor. When a draft came back generic, I killed it at the gate. The gate is the product, not a formality.
And LinkedIn itself is worth this effort more than ever: Profound's 2026 citation reports show LinkedIn is the #1 cited domain for professional questions in AI search. The posts you write with an actual perspective are not just feed content anymore. They are becoming source material for the answers AI engines give about your space.
The honest close from a vendor
If your plan is to type topics into a tool and auto-blast whatever comes out, do not use AI for content. And do not use my product either. You would be paying to join the 53.7% that readers skip, the feed buries, and the next algorithm update erases. I would rather lose the sale than watch you become the data point.
If you have real ideas and no time to write them, that is the problem AI actually solves. It is the problem Liftli exists for: it runs inside the AI you already use (Claude today, ChatGPT and Cursor next), learns your voice from your writing, drafts from your voice notes, calls, and commits, and holds everything behind a one-tap approval gate. Free tier, no card. Paid starts at $29. But the workflow matters more than the tool. Get the input human and the gate human, and AI stops being a reach risk and starts being leverage.