Does LinkedIn detect AI posts? What it does, what it doesn't, and what readers notice

Short answer: partly, and LinkedIn hasn't said how. Reports from May 2026 say LinkedIn is cutting the reach of generic posts that look AI-written. It has not published a public AI label for text, and it doesn't ban AI-assisted writing. The label you can see applies to images and video that carry provenance metadata. The detector that matters more is your reader, who spots filler in about two lines.

What LinkedIn does

It reduces distribution of generic, AI-looking posts. According to coverage of the announcement, LinkedIn's Global Editorial VP, Laura Lorenzetti, described changes to recommendations aimed at what she called "AI slop." The Decoder reports that flagged content that is thin on opinion gets less reach and mostly stays inside the author's own network. It also reports LinkedIn's claim that early tests flagged generic posts correctly 94 percent of the time. That figure is LinkedIn's own, as relayed by a news site. I couldn't check it against a LinkedIn primary document, so treat it as a claim, not a measurement.

Note the target. The reporting describes content that is both AI-looking and generic. It is not "anything a model touched."

It labels some AI images and video. LinkedIn announced in 2024 that it would label AI-generated content using the C2PA standard. The "CR" Content Credentials icon appears on media whose files carry that metadata. One explainer says it currently applies to visuals, not text. Another notes the rollout is gradual and that LinkedIn can't identify or label all AI-generated content.

What LinkedIn doesn't appear to do

• Put an "AI-written" badge on your text post. I found no sign of one.
• Ban you for using AI. The reporting describes lower reach, not removal. One write-up puts it as suppressing, not removing.
• Explain its method. One analysis of LinkedIn's published policies found no statement that LinkedIn detects AI-written posts, and said the only AI detector LinkedIn has publicly described looks at profile photos. That sits awkwardly beside the May reporting. Both can be true if the ranking changes were described in interviews and posts rather than in policy documents.

So the honest position is: some reach penalty for generic posts looks likely, the mechanism is unknown, and anyone selling you a precise "LinkedIn AI score" is guessing.

What readers notice

Whatever LinkedIn's systems do, your audience has been reading these posts for two years. Pangram, a company that sells an AI detector, scanned over a million social posts and reported that more than 40 percent of long-form LinkedIn posts were flagged as AI-generated. That's a vendor's classifier, so the exact number is soft. The feeling behind it isn't. People are tired of the same shape.

The patterns coverage of that study lists are the ones readers describe too: a rigid structure such as three lessons and a call for engagement, a recognisable set of favourite words, and opening lines in formats models like. Our guide to words that make writing sound like AI covers the vocabulary side.

From editing a lot of professional copy, these are the tells that matter most:

• No specifics. No client size, no date, no number, no name, no thing that went wrong.
• Every sentence the same length. Even rhythm, neat triplets, a punchy one-liner after each paragraph.
• A lesson with no story. "Here's what I learned" followed by advice that could sit under any post.
• A closing question that asks for nothing real. "Thoughts?" tacked on to everything.
• Opinions that nobody could disagree with. Thin on opinion is the exact quality the reach reports describe.

How to check a post before you publish: 6 steps

1. Find the one claim. State what you actually believe in a sentence. If you can't, the draft isn't ready, whoever wrote it.
2. Add one thing only you know. A real number, a real client situation (anonymised if needed), a mistake, a date.
3. Cut the template. Remove the hook line, the numbered lessons and the engagement bait if they were there for structure alone.
4. Vary the sentences. Let one run long. Let another be four words.
5. Read it aloud. Anything you'd never say to a colleague gets rewritten.
6. Run a check. A tool can catch repeated phrasing and stock openers you've stopped seeing. It can't tell you what LinkedIn will do. More on that in why AI detectors flag human writing.

Example: the same post, two ways

This is an invented example for illustration.

Generic draft: "Leadership isn't about having all the answers. It's about asking the right questions. Here are 3 lessons I've learned about building great teams. 1. Listen first. 2. Empower others. 3. Celebrate wins. What would you add?"

Revised: "Last spring I ran our weekly sales review and talked for 40 of the 45 minutes. Nobody pushed back, which I read as agreement. Two reps told me later they'd stopped bringing deals they were unsure about. Now I ask for the worst deal in the pipeline first, and I go last."

The second one is shorter on advice and longer on evidence. It could still have started life in a model. What matters is that a person supplied the facts.

Common mistakes

• Swapping synonyms. Rewording a generic post gives you a different-sounding generic post.
• Hiding AI use instead of fixing the content. If the post says nothing, nobody cares how it was made.
• Trusting a free detector score. Those scores measure a model's guess about text. They don't measure LinkedIn's ranking. Humans get flagged too.
• Ignoring image metadata. If you add AI-generated graphics, the label may appear regardless of your text.
• Posting at volume. Daily AI-drafted posts in the same format teach your own audience to scroll past you.

Where Plainspoke fits

Plainspoke is a writing checker and humanizer for professional writing, including LinkedIn posts, cold email and resumes. It's useful for the editing pass: spotting stock phrasing and flat rhythm, then rewriting toward how you actually talk. It can't add your stories. And it can't promise reach, because nobody outside LinkedIn knows how the ranking works. If you want to compare approaches, see our comparison pages or the guide to making AI writing sound human.

Sources

• the-decoder.com/linkedins-war-on-ai-slop-is-not-just-a-policy-update-it-is-an-admission-that-the-platform-lost-control-of-its-feed/
• www.techjuice.pk/linkedin-limit-ai-generated-content-reach/
• www.faryalrazabhatti.co.uk/post/linkedin-ai-content-policy-2026
• originality.ai/blog/ai-content-published-linkedin
• www.growtempo.com/blog/linkedin-ai-content-detection
• www.mediapost.com/publications/article/396225/
• www.lilachbullock.com/linkedin-content-credentials-label-meaning/
• gdprlocal.com/linkedin-content-credentials/
• www.pangram.com/blog/ai-in-your-feed
• www.techtimes.com/articles/320534/20260715/linkedin-leads-all-platforms-ai-written-posts-study-million-feeds-confirms.htm

Check a draft free