How to Write a LinkedIn Post That Doesn't Sound Like AI
The tells that give away an AI-written LinkedIn post, why they cost you credibility, and a six-pass editing method to get rid of them.
· Content creation
A LinkedIn post sounds like AI when anyone could have written it. To avoid that, start from something only you know (a number from your business, a scene, a mistake), strip out the tells readers spot right away (the suspense hook, lists of three, "it's not X, it's Y," the closing moral), and read it out loud before you publish. AI can help you structure; the raw material has to come from you.
This article is part of our series on automating social media: automating production only makes sense if the result still reads like someone actually speaking. Below are the most common tells, why they hurt on LinkedIn in particular, and an editing routine that makes them disappear.
Why does a post that "sounds like AI" cost you credibility?
People read LinkedIn posts to find out what a specific person or company thinks. Generic copy signals the opposite: nobody actually took a position. Readers don't stop to ask whether AI was involved; they just sense the post tells them nothing about you.
LinkedIn has said as much in writing. In its feed update announced on March 12, 2026, the company says it wants to show more posts offering genuine insight, actionable ideas and thoughtful perspectives, and fewer repetitive, low-substance posts and engagement bait. It adds that members come to hear from real professionals rather than automated scripts. LinkedIn does not claim to detect or penalize AI as such: what it targets is thin content, whether typed by hand or not.
On accountability, LinkedIn's User Agreement (section 3.6, version effective November 3, 2025) warns that generated content may be inaccurate, incomplete or misleading, and asks users to review and edit it before sharing. In other words, what you post is on you, whatever tool drafted it.
Which tells give away generated copy?
There is no official list. These are patterns readers pick up on, because a language model falls back on them whenever it isn't given specific material. Any one of them is fine on its own. Stacked together, they give the post away.
| Tell | Typical example | Why it rings false |
|---|---|---|
| The cliffhanger hook | "Three years ago, I made a mistake that changed everything." | The promise is vague and the payoff rarely lives up to it. |
| "It's not X, it's Y" | "Leadership isn't a title. It's a mindset." | A false contrast: nobody said otherwise. |
| Lists of three | "Clarity, consistency, courage." | A constant three-beat rhythm sounds rehearsed. |
| One sentence per line | Ten one-line paragraphs separated by blank lines | The post gets longer without saying more. |
| Inflated words | "crucial," "game-changer," "unlock," "in today's fast-paced world" | An adjective stands in for a fact. |
| The closing moral | "What do you think? Let me know in the comments." | A generic question that asks nothing specific. |
| No verifiable detail | No number, city, date or client (even anonymized) | The post could apply to any company. |
The last row matters most. The other tells are symptoms; missing detail is the cause. A model can't make up your week: if all it gets is a topic, it fills the gap with stock phrases.
How do you start from material that's yours alone?
Before you open any tool, jot down three lines only you could write. Three questions are enough:
- What happened? A dated fact: a client call, a botched delivery, a question you got three times this week.
- What did you take from it? One idea, phrased the way you'd say it to a coworker.
- What detail proves it? A number from your business, a line someone said, a concrete before and after.
Take a fictional firm, Northpeak Advisory, which helps small companies with hiring. The topic "why onboarding matters" produces an interchangeable post. The material "on Tuesday, a hire we placed in September told us he almost quit his first week because nobody had set up his email" produces a post only that firm could write.
If you use AI, that material is what you hand it, along with your goal and your constraints. We walk through how in how to write a good brief for an AI content tool.
How do you edit a post to remove these tells?
Good editing happens in several short passes, each with a single goal. Plan on about ten minutes for an average-length post.
- The first-line pass. In the feed, LinkedIn only shows the opening before "see more," so the first line should state the idea or the fact, not tease it. Swap "Here's what I learned" for what you learned.
- The detail pass. Does every paragraph contain at least one verifiable element? If not, add one or cut the paragraph.
- The adjective pass. Underline intensity words (crucial, essential, incredible). Replace each with the fact that justifies it, or delete it.
- The rhythm pass. Break up lists of three when the third item is just there for cadence. Merge stranded one-liners into real two- or three-sentence paragraphs.
- The ending pass. Cut the moral and the generic question. End on the concrete takeaway, or on a specific question readers can answer from their own experience.
- The read-aloud pass. Read the post as if you were saying it in a meeting. Any sentence you would never say out loud gets rewritten.
The same goes for hooks: our piece on hooks by platform, with 30 examples shows openers built on a fact rather than on suspense.
Before and after: what does the rewrite look like?
Generated with no material:
"In today's fast-paced world of work, onboarding new team members has become crucial. It's not a formality. It's an investment. Three keys: anticipate, support, listen. How do you welcome your new hires?"
Rewritten with the firm's own material:
"On Tuesday, a hire we placed in September told us he almost quit his first week. Not because of the job: by Thursday he still had no email login. Since then, we send every client a list of five accesses to set up before day one. It fits on one page. If you're hiring this month, which access do you forget most often?"
The second version is shorter, includes a dated fact, a decision and a concrete tool, and ends on a question people can actually answer. No intensity words required.
What can AI do without the post losing your voice?
AI is useful for anything that doesn't require an opinion. Split the work like this:
| Task | Hand to AI | Keep for yourself |
|---|---|---|
| Finding the idea | Suggest angles from your notes | Pick the one you'd defend in public |
| Structuring | Order fact, takeaway, proof | Decide what matters most |
| Drafting | A first draft from your material | The key sentences, especially the first and last |
| Tightening | Cut repetition, fit the length limit | Check nothing important got lost |
| Checking | Flag the tells from the table above | Verify every number and every name |
For more on what AI can take off your plate and what stays your job, see can AI run my social media?
Should you post more often to "feed" LinkedIn?
No. One more thin post doesn't make up for one fewer solid post, and the 2026 feed update targets repetitive posting specifically. Pick a cadence you can sustain with real material every time; we cover that in how often to post on LinkedIn. For formats and technical limits (3,000 characters per post, documents, video), our LinkedIn formats guide pulls together what LinkedIn's own help pages say.
Frequently asked questions
Does LinkedIn penalize posts written with AI?
LinkedIn does not say it penalizes AI as such. Its feed update announced in March 2026 targets repetitive, low-substance posts and engagement bait, whether written by hand or generated. A useful, specific post has nothing to worry about.
Do I have to disclose that a post was written with AI?
LinkedIn does not require it for a text post. Its User Agreement does ask you to review and edit generated content before sharing it: you remain responsible for what you publish.
What is the biggest giveaway of generated copy?
No verifiable detail. Without a number, date, place or concrete example, the post could apply to anyone. The other tells, like lists of three or a closing moral, follow from that.
How long does it take to edit a generated post?
About ten minutes for an average-length post, in six short passes: first line, details, adjectives, rhythm, ending, then reading it out loud.
A brand voice, not a machine voice, with Oposto
Oposto learns your brand from your website (voice, products, audience) to build your content DNA, then drafts posts adapted to each platform, LinkedIn included. Every post is reviewed before publishing, with a score and recommendations; you read it, add your own material and approve it, and it goes out at the scheduled time. You can try it on the free plan, no credit card required, at oposto.app.