Scroll through LinkedIn for five minutes and you'll start to see it. The same cadence. The same sentence fragments used for emphasis. The same breathless enthusiasm about lessons learned from a morning run, a delayed flight, or a difficult client conversation that turned into a "growth opportunity."
AI-generated LinkedIn content has a signature, and by now, most people can feel it - even if they can't name it. (This is also why most AI LinkedIn post generators miss the mark.)
The problem isn't that people are using AI to help them write. The problem is that they're asking AI to do all the thinking, and then posting the result verbatim. The output is technically correct and completely hollow.
If you want your LinkedIn posts to actually sound like you, you need to understand why AI slop happens - and how to stop it before it starts.
Why AI-generated posts all sound the same
When you ask an AI to write a LinkedIn post from scratch, it does what it's trained to do: it produces the most statistically likely version of a LinkedIn post. And the most statistically likely LinkedIn post is a blend of the millions of LinkedIn posts it has already seen.
That's why you get the same structures over and over:
- The one-word opener. "Uncomfortable." "Unpopular." "Honest."
- The three-line story followed by the turn. "Then something changed."
- The numbered list of lessons that ends with a call to engage. "Which one resonates most with you?"
- And of course, the classics: "I'm thrilled to announce", "Grateful and humbled", "This one's for anyone who..."
These patterns exist because they worked - at some point, for some people. AI learned them. Now it reproduces them for everyone, flattening thousands of distinct voices into one beige paste.
The result is a feed where every post sounds like it was written by the same slightly over-caffeinated LinkedIn influencer who just got back from a Tony Robbins event.
But the real damage isn't aesthetic. It's strategic. When your posts sound like everyone else's, your audience can't form a relationship with you as a thinker. They're following a persona-shaped blur, not a person. Over time, that erodes the very thing that makes LinkedIn worth investing in: trust, recognition, and a reputation for having a genuine point of view.
Consider the difference between two founders posting about the same topic - say, why their last product launch underperformed. The AI-generated version talks about "pivoting mindset," "leaning into learnings," and "embracing the journey." The human-sounding version says: "We shipped too early because I was scared of a competitor, and that fear cost us three months of credibility with our best customers." One of those posts will get saved and shared. You already know which one.
The real issue: AI doesn't know you
A language model has no idea how you actually talk. It doesn't know that you always use rhetorical questions when you're making a point, or that you swear occasionally in casual contexts, or that your humor is dry and understated. It doesn't know the specific industry shorthand you use, the references that land with your audience, or the opinions you hold that aren't exactly mainstream.
When you ask ChatGPT to "write a LinkedIn post about my experience at a sales conference," it fills in everything it doesn't know about you with generic LinkedIn energy. And generic LinkedIn energy is exactly what you're trying to avoid.
Your voice is made of specifics. The way you phrase things. The asides you make. The things you choose to leave out. The rhythm of your sentences when you're making a real point versus when you're warming up. None of that gets transmitted when you type a topic into a text box.
Here's a concrete example. Imagine you want to post about how you've changed the way you run team meetings. You type that into an AI tool and you'll probably get something that lists "three key principles for more productive standups" - practical, inoffensive, instantly forgettable.
But what you actually wanted to say was: "I used to run meetings like a status update theater. Fourteen people on a Zoom call watching each other's shoulders. Then I just... stopped. And nothing broke. In fact, three people told me they finally had time to think." That's a post. That's something only you could write, because it came from something that actually happened to you, in a way you actually experienced it.
The AI can't invent that specificity. You already have it. The challenge is extracting it in a form that becomes content.
Why starting from speech changes everything
Here's something worth sitting with: you already know how to sound like yourself. You do it every day when you talk to people.
When you explain a problem to a colleague over Slack, you sound like you. When you debrief after a meeting, you sound like you. When you vent about a bad client call or get excited about a new strategy that's working - that's your voice, right there, completely unfiltered.
The issue isn't that you don't have a voice. It's that the act of writing makes people second-guess it. You open a blank document and suddenly you're performing professionalism instead of just talking.
Speech bypasses that block. When you talk through an idea - even just to yourself - you naturally produce the things that make your writing feel like you: the specific details, the real opinions, the way you actually frame things when you're not overthinking it.
This is why voice memos are underrated as a content creation tool. Next time you find yourself telling a story about something that happened at work, notice how naturally the key points fall into place. The setup, the complication, the thing you realized - you structure it intuitively because that's how human storytelling works. That natural structure, captured in your actual words, is infinitely better raw material for a post than a blank prompt.
This is the insight behind how Edgar works. Instead of prompting users to describe what they want to write, Edgar runs a short weekly voice conversation with an AI agent - think of it like a quick debrief call with someone who knows your work. You talk through what's on your mind: what happened this week, what you're thinking about, what surprised you, what you're frustrated by. From that raw material, Edgar generates LinkedIn posts. Not from a blank slate, but from what you actually said, in the way you actually said it.
Before generating anything, Edgar also analyzes your existing LinkedIn posts to understand how you write: your sentence length, your structural habits, the words you overuse, the tone you default to. So the output isn't just a post about your topic - it's a post that sounds like the specific person who said the thing. That's a meaningfully different starting point.
How to recognize when AI is overwriting your voice
Learning to spot AI-voice creep in your drafts is a skill worth developing. Some signals to watch for:
You'd never say that sentence out loud. Read your draft aloud. If any sentence makes you stumble, or sounds like something a press release would say, it didn't come from you. Flag it immediately.
It's using your topic as decoration. Generic AI posts use your specific experience as a thin wrapper around a universal lesson anyone could have written. "After ten years in SaaS, I've learned that customer success is really about relationships." Technically true. Completely yours? No.
The emotions are named but not felt. AI tends to tell you what the writer felt ("I was humbled," "I felt inspired") rather than showing the situation that produced that feeling. Real writing shows the moment; the emotion arrives in the reader, not the text.
It's trying to appeal to everyone. Your voice has a point of view that some people won't like. If a post has been smoothed down to something universally palatable, AI has probably flattened the edges that would have made it yours.
The length is padded, not earned. If you can cut the first two sentences and the post gets better, the intro was throat-clearing - something AI does constantly.
Practical tips for keeping your voice when using AI tools
Whether you're using Edgar or any other AI writing tool, these principles apply.
Start with your own words, not a topic. Instead of typing "write a post about leadership," say something out loud first - to a voice memo, a friend, a rubber duck - and then work from that transcript. The raw material matters enormously. A two-minute voice memo will produce better input than a ten-word prompt every time.
Feed the AI your real opinions, not just your subject. If you think the conventional wisdom on something is wrong, say that explicitly. "Everyone says you should respond to every comment - I think that's mostly a waste of time and here's why." AI is much better at helping you express a real point of view than it is at inventing one for you. Give it the opinion; let it help you shape the argument.
Watch for phrases you'd never actually say. Read the output out loud. If you hit a sentence and think "I would never say it like that," cut it or rewrite it in your own words. This is the fastest editing filter you have, and it takes about ninety seconds per draft.
Keep your quirks. If you have a habit of using em dashes - maybe a bit too much - that's yours. If you tend to write very short sentences when you're being direct. Keep it. If you use parenthetical asides the way most people use footnotes (because you always have three things going on at once), leave them in. AI tools will often sand these down into something smoother and more generic. Push back. Your quirks are the fingerprints that make readers recognize your work.
Give the AI examples of your actual writing. The more context it has on how you write, the better the output. Paste in three or four of your best previous posts and say "write in this style." This is laborious to do manually every time, which is why Edgar handles it automatically - but even doing it by hand is worth the effort.
Edit for removal, not addition. AI tends to over-explain. It adds context you don't need, restates points you already made, and wraps up with a bow when the post was already done two paragraphs earlier. Your job in editing is usually to cut - to remove the scaffolding and leave just the structure. A post that's 20% shorter is almost always better. If the insight can stand on its own, let it.
Protect the opening. The first one or two lines of a LinkedIn post are what determine whether anyone reads the rest. AI openers are almost always generic because they're trying to appeal to the broadest possible audience. Write your own opener, even if the AI writes everything else. Your opener should sound like you talking - not like a headline.
What great LinkedIn content actually does
It's worth stepping back and asking: what are you actually trying to accomplish?
Most founders and operators on LinkedIn aren't there to go viral. They're there to build a specific kind of reputation - as someone who thinks clearly, has real experience, and is worth paying attention to over time. That kind of reputation is built through consistent, recognizable voice over many posts, not a single banger that the algorithm happened to push.
This means your content strategy isn't really a content strategy. It's a thinking-out-loud strategy. The goal is to capture what you're actually processing - the decisions you're making, the things you're noticing, the opinions you're forming - in a form that other people can follow. When someone reads ten of your posts over six months, they should feel like they know how your mind works.
That's why authenticity isn't just a nice-to-have. It's the mechanism by which LinkedIn actually creates business value. People hire, refer, and buy from people they feel like they know. A feed full of AI-generated posts that could have come from anyone will never create that feeling. A feed that sounds unmistakably like you will.
The bottom line
The problem with AI-generated LinkedIn content isn't AI - it's the workflow. When people use AI as a ghostwriter who knows nothing about them, they get posts that sound like nobody in particular. When they use AI to help shape and refine something that started with their actual thinking, in their actual voice, the result is something worth reading.
Making your LinkedIn posts sound like you isn't about avoiding AI tools. It's about not outsourcing the part that only you can do.
Your perspective. Your experiences. Your specific way of seeing things. The detail that only you observed because you were the one in the room. That's what makes content worth following someone for. AI can help you package it, structure it, and sharpen it. But it cannot generate it from nothing - and every attempt to do so produces the same beige paste that's already clogging everyone's feed.
The shortcut isn't a better prompt. The shortcut is starting with something real: a conversation you can repurpose, a voice memo, a story you've already told someone this week. AI that works with your voice, rather than replacing it, is genuinely useful. AI that works instead of your voice is just noise with your name on it.
Start with your voice. Build from there.
