How to Write a LinkedIn Post That Does Not Read as AI

A LinkedIn post reads as AI when it could have been posted by anyone. That is not a style judgment, it is LinkedIn's own definition: content that is "polished on the surface but lacks a clear point of view, unique perspective, or substance." So the fix is not a list of banned words. It is putting one thing only you could know in the first two lines, taking a position someone could argue with, and cutting every sentence that would survive unchanged in a stranger's post.

Since 30 July 2026 readers can flag a post as "seems like AI slop", and LinkedIn says content it classifies that way now gets 40 percent fewer views. Below is what LinkedIn actually counts as slop, a seven question test derived from that definition, a worked rewrite, and the data on which opening lines earn the read.

What does LinkedIn actually count as AI slop?

LinkedIn published its definition on a help page titled best practices for content created with the help of AI. The wording matters, so here it is close to verbatim.

• Slop is "low-effort, likely AI-generated content that may sound polished on the surface but lacks a clear point of view, unique perspective, or substance"
• It "goes beyond style to include content that feels generic, repetitive, recycled, or designed primarily to game attention"
• "LinkedIn's focus is not on how content is created, but whether it adds value." AI-assisted posts are welcome when they carry a real person's perspective, experience or expertise
• Using AI "in refining language, proofreading or being more concise" is named as a supported use
• If enough members flag a post, the author sees a tip in Post analytics. The page says this is not a takedown and not a policy decision

The mechanics, per TechCrunch on 30 July 2026 and Social Media Today on 20 August 2026: a "seems like AI slop" option in the three-dot menu on posts and comments, classifiers that keep flagged content out of suggested posts, more than a million people using the option in two weeks, and LinkedIn's Chief Product Officer Hari Srinivasan saying content the platform defines as slop was getting 40 percent fewer views.

One detail most guides skip: LinkedIn pulled its own "enhance your post" rewriting feature and replaced it with a proofreader. The platform that built an AI rewriter decided rewriting was the problem.

Why do lists of AI-sounding words miss the point?

Every guide on this topic ships a list: delete "delve", delete the buzzwords, delete the triplets, delete the closing question. Do that and a slop post is still a slop post with better vocabulary. LinkedIn's definition is about what the post contains, not how it sounds, and a human can write substance-free content without any help.

The tells still cost you, because a reader who spots them stops reading, and the ranking paper LinkedIn published in February 2026 names exactly two objectives: long dwell and a like, comment or share. A post abandoned at line three loses on both. Clean the tells, but clean them last.

The seven question slop test

Each question maps to a phrase in LinkedIn's definition. A post that fails two or more is the kind readers flag, whether or not a model wrote it.

Run this before you post

• 1. Could anyone else have posted this? If you swap your name for a competitor's and nothing breaks, it is generic. (LinkedIn: "generic")
• 2. Is there a specific moment in it? A date, a place, a number you measured, a sentence someone actually said to you. (LinkedIn: "experience")
• 3. Does it take a position someone could disagree with? If every reader nods, there is no perspective. (LinkedIn: "clear point of view")
• 4. Does the first line make a claim rather than ask a question? Story and contrarian openers earn the most engagement in the largest study available; questions earn the least. (LinkedIn: "designed to game attention")
• 5. Would it lose anything if you deleted the last line? Summaries, morals and "agree?" closers are filler. (LinkedIn: "repetitive")
• 6. Is every number sourced or measured by you? An unsourced statistic is recycled content by definition. (LinkedIn: "recycled")
• 7. Did you write the first draft, or did the model? If the model did, you have a proofreading job ahead of you, not a proofreading job behind you. (LinkedIn: "whether it adds value")

How do you rewrite a post that fails the test?

An illustrative example. Neither version is a real post, and the numbers in the rewrite are invented for the illustration.

Before: fails questions 1, 2, 3, 5 and 7

"Consistency is the key to building a personal brand on LinkedIn. Many founders struggle to post regularly because they feel they have nothing to say. But the truth is, your journey is valuable. Share your wins, share your lessons, and show up every day. Over time, the results will speak for themselves. What is stopping you from posting today?"

After: the same idea from one founder's week

"I posted every weekday for 40 days and got two inbound calls. Both came from the same post, the one where I published our churn number. The other 39 posts were advice. Nobody books a call off advice. They book a call when you show them something you would normally hide."

What changed: a specific run (40 days, two calls, one post) replaced "consistency is key", an arguable claim replaced a truism, the first line states an outcome, and the closing question is gone. Nothing in the rewrite could be posted by a different founder.

The rewrite sequence, in order

• Start from an event, not a topic. Something that happened this month with a number or a quote attached
• Write the first line as a claim under 40 characters if you can, under 80 at most
• State the position in one sentence, then give the evidence from your own experience in two to four short paragraphs
• Delete the summary line, the moral and the closing question
• Source any number that is not yours, with the name of the source in the post
• Only now run a proofreader, human or AI, and accept fixes to grammar and length, not to voice

What does the data say about opening lines?

AuthoredUp classified 309,614 personal profile posts from December 2025 to May 2026 by first-line style. It is a vendor study, not peer reviewed, but it is the largest published sample and the method is stated.

Median engagement rate by hook style (AuthoredUp, updated 11 August 2026)

• Story opener: 2.60 percent (22,745 posts)
• Contrarian opener: 2.31 percent (19,603 posts)
• Statement opener: 2.27 percent (171,391 posts)
• Results opener: 2.19 percent (64,311 posts)
• Question opener: 2.16 percent (31,564 posts)

Median engagement rate by hook length, same study

• 0 to 40 characters: 2.61 percent
• 41 to 80: 2.39 percent
• 81 to 120: 2.23 percent
• 121 to 200: 2.15 percent
• Over 200: 2.08 percent

The same study checked the closing question: 2.33 percent median with one, 2.26 without. The most repeated piece of LinkedIn advice is worth 0.07 points, and it is the most recognisable AI tell on the platform.

One independent point on substance: the controlled experiment on what AI answer engines pick up, published at KDD 2024, found statistics, quotations and citations lifted visibility roughly 30 to 40 percent while keyword stuffing did nothing. Specific, sourced detail is what both a reader and a model treat as substance.

Can you still use AI to write LinkedIn posts?

Yes, and LinkedIn says so in writing. The help page names refining language, proofreading and being more concise as fine, and asks that you disclose it if you relied on AI heavily and it is not obvious from context. The workflow that stays on the right side of that line:

• Record yourself talking about the event for two minutes. Voice memo is fine
• Transcribe it. This is where a tool earns its keep
• Write the post yourself from the transcript. Your phrasing, your order, your claim
• Paste it into a proofreader with the instruction to fix errors and cut length only, and reject any change to a sentence you would say out loud
• Read it back against the seven questions before posting

The reverse order, prompt, generate, tweak two words, post, produces a post that could belong to anyone, which is what the button is for.

What should you do if you got the "seems like AI" tip?

Nothing dramatic. The help page says the tip is not a takedown and no policy decision was made. Run your last ten posts through the seven questions, be honest about question seven, and post the next one from a real event. If your posts are already first-hand and still get flagged, the likely cause is formatting that looks generated: line-break staircases, emoji bullets, a question at the end. Fix the shape, keep the substance.

Two cases this does not cover. On a company Page, text posts are the weakest format regardless of quality, and the profile versus company page comparison explains why. And if you have no first-hand events because you are not yet doing the work you post about, no rewrite fixes that.

Frequently asked questions

Does LinkedIn penalise AI-written posts?

Not for being AI-written. LinkedIn's help page says its focus is whether content adds value, not how it was made. Posts classified as slop get reduced distribution, 40 percent fewer views per Srinivasan in August 2026. A first-hand post drafted with AI help is not the target.

What is the best first line for a LinkedIn post?

A short claim or the opening beat of a story. In AuthoredUp's 309,614 post sample, story openers earned a 2.60 percent median engagement rate and question openers 2.16 percent. Hooks of 40 characters or fewer outperformed every longer band. State something, keep it short, stop.

How long should a LinkedIn post be?

The limit is 3,000 characters, per LinkedIn's post and share updates page. There is no published ideal length and the large vendor studies contradict each other. Write until the point is made with evidence, then delete the last line.

Should I disclose that I used AI?

LinkedIn recommends it if you relied on AI heavily and it is not obvious from context. For proofreading or trimming your own draft, no disclosure is expected. For a post the model wrote, the disclosure is less of a problem than the post.

Key takeaways

• LinkedIn defines AI slop as polished content lacking a clear point of view, unique perspective or substance. The test is what the post contains, not how it was made
• Since 30 July 2026 readers can flag posts, more than a million did so in two weeks, and flagged content gets about 40 percent fewer views per LinkedIn's CPO
• Word lists fix vocabulary, not substance. A post that could be posted by anyone fails regardless of phrasing
• Run the seven questions: anyone could post it, no specific moment, no arguable position, question opener, deletable last line, unsourced numbers, model wrote the draft
• Story and contrarian openers under 40 characters earn the highest median engagement across 309,614 posts. Question openers earn the lowest
• The closing question adds 0.07 points of engagement and is the most recognisable AI tell. Delete it
• AI for transcription and proofreading is explicitly fine. AI for the first draft is what the button was built for

Related reading

The ghostwriting cost breakdown lists published 2026 rates for having this done by someone else. The specs reference covers every image, video and document limit from LinkedIn's own help pages, and the thought leadership post covers building the backlog of first-hand events this method depends on.

If the events are there and the writing time is not, EchoPulse ghostwrites LinkedIn posts from a recorded voice interview so the first-hand detail stays yours. It starts with a 14 day, $299 pilot and no retainer. Details on the LinkedIn ghostwriting page and the page for founders