AI and authenticity
LinkedIn's AI slop button: what it means for founders who post
On July 30, 2026, LinkedIn gave every member a way to report posts that read like machine output. Here is what the button does, what LinkedIn has said since, and how to keep AI in your workflow without it showing up in your words.
By Dr. Banpreet KaurUpdated 6 min read
- Long-form posts fully AI-generated, Apr to Jun 2026 (Pangram)
- 41%
- Times the slop button was used in its first weeks
- 1M+
- Fewer views on content LinkedIn classifies as slop
- 40%
The short answer
LinkedIn's "Seems like AI slop" option, launched July 30, 2026, lets members flag posts from the three-dot menu. Authors see flags in post analytics, and LinkedIn says slop it classifies now gets about 40% fewer views. Use AI for research and analysis if you like, but write and approve the actual words yourself.
What the AI slop button is
From the three-dot menu on a post, members can now choose "Seems like AI slop". Gizmodo's launch report quotes LinkedIn chief product officer Hari Srinivasan explaining the rationale: slop is hard to define and keeps changing, so member reports help LinkedIn tune its models. The person who posted is told privately, in their post analytics.
At the same time LinkedIn retired its "enhance your post" AI writer and replaced it with a proofreading tool that, in LinkedIn's words, keeps your voice intact. That is a telling reversal. The platform that once nudged members to let AI rewrite their posts now treats AI-sounding posts as a quality problem.
LinkedIn has not banned AI. It has said it will expand classifiers that detect AI-generated and low-quality content, and that it has already stopped billions of AI-generated posts and comments.
What LinkedIn has said since launch
By late August, Engadget reported that the button had been used more than a million times. Srinivasan wrote that members were seeing 40% fewer views on what LinkedIn classifies as AI slop compared with a few weeks earlier. He also said LinkedIn uses many signals to decide reach and has safeguards so individual reports cannot be used to unfairly target other members.
What LinkedIn has not said is how many flags it takes to reduce a post's reach, or whether flags affect the author's future posts. Assume it is one signal among several, combined with its own classifiers.
How much of LinkedIn is AI-written
The scale explains why LinkedIn acted:
| Data point | Source |
|---|---|
| 41% of long-form LinkedIn posts from April to June 2026 were fully AI-generated | Pangram, via Fortune |
| LinkedIn made up about 62% of flagged AI content while being roughly a third of items Pangram scanned | Pangram, via Gizmodo |
| The slop button was used more than a million times in its first weeks | LinkedIn, via Engadget |
| Content classified as slop now gets about 40% fewer views | LinkedIn, via Engadget |
When four in ten long posts are machine-written, sounding human becomes a competitive advantage rather than a baseline.
How readers spot AI posts: the common tells
Readers are not running detectors. They are reacting to patterns they have seen a thousand times. These are the ones we remove most often when editing drafts:
| Tell | What it looks like | What a person writes instead |
|---|---|---|
| The contrast frame | "It's not about the product. It's about the people." | Say the actual point once, with an example |
| The rhetorical opener | "Ever wondered why most startups fail?" | Open with the specific thing that happened |
| Lists of three, every time | "Clarity, consistency and courage." | Two items, or five, whatever is true |
| One-line dramatic closers | "Read that again." | End with a question or a concrete next step |
| Long dashes everywhere | Clauses joined with dashes in every sentence | Commas, full stops, parentheses |
| No names, no numbers | "A client once told me..." | "Our second customer, a 40-person logistics firm, told me..." |
| Symmetrical paragraphs | Every paragraph two lines, every line the same rhythm | Uneven length, the way people talk |
| Generic lesson at the end | "The lesson? Never give up." | The non-obvious thing you would do differently |
| Emoji bullets and hashtag blocks | Rockets, checkmarks, eight hashtags | Plain text, one or two inline tags at most |
The deepest tell is not stylistic. It is the absence of anything only you could know: the deal you lost, the number on your dashboard, the patient question that changed how you explain a diagnosis.
One post, before and after
Here is a hypothetical draft of the kind we see every week, from a founder of a mid-sized logistics software company:
"Leadership isn't about having all the answers. It's about asking the right questions. This year taught me three things: resilience, adaptability and trust. What's your biggest lesson of the year? Let me know below."
Nothing in it is false, and nothing in it could only have come from this founder. It has the contrast frame, the list of three, the generic closing question. A reader has seen it hundreds of times and scrolls on, or reports it.
Here is the same idea after a 15-minute interview:
"In March our biggest customer threatened to leave over late deliveries in Pune. I wanted to promise a fix by Friday. Our ops lead asked one question instead: which three routes cause 80% of the delays? We did not know. It took us two weeks to find out, and the routes turned out to be fine. The problem was a single warehouse shift change."
The second version is longer and less tidy. It is also impossible to fake, which is exactly why it holds attention, earns comments from other operators and gives nobody a reason to flag it.
What happens if your post is flagged
You will see a notice in that post's analytics. LinkedIn has not published thresholds, so one flag from an annoyed reader is unlikely to sink an account. A pattern of flags on posts that also trip LinkedIn's classifiers is a different matter. If you get flagged, reread the post against the table above. Usually the fix is adding specifics and cutting the formula.
How to use AI responsibly: research yes, words no
AI is useful in a LinkedIn workflow. The problem is using it for the one part readers judge you on: the voice.
| Good uses of AI | Risky uses of AI |
|---|---|
| Summarizing a long report before you form your view | Generating the post from a one-line prompt |
| Transcribing a voice note or interview | "Make this sound more professional" rewrites |
| Pulling performance patterns from your analytics | Auto-generated comments on other people's posts |
| Checking a draft for typos and factual slips | AI-written replies to DMs |
| Brainstorming angles you then reject or rewrite | Posting a draft without reading it aloud |
A simple test: if you deleted every AI tool tomorrow, would your posts still contain the same facts, stories and opinions? If yes, AI is a research assistant. If no, it is the author.
What ghostwriting means now
The slop button has made people ask whether ghostwriting is the same problem. It is not, when done properly. A ghostwriter interviews you, captures your stories, opinions and phrasing, drafts in your voice, and you approve every post. The ideas and experience are yours; the craft of shaping them is the ghostwriter's. That is how speeches, op-eds and executive letters have long been written.
What has changed is that "ghostwriting" now covers two very different products. One is a person who knows your business. The other is a prompt wrapped in a subscription. Our comparison of an agency versus AI tools goes into the trade-offs, and our explainer on what LinkedIn ghostwriting is covers the ethics, including when disclosure makes sense.
A five-minute check before you publish
- Read the post out loud. Anything you would never say in a meeting goes.
- Find at least one name, number or specific situation. If there is none, add one.
- Delete any "it's not X, it's Y" sentence and say Y directly.
- Count your lists. If every list has three items, change one.
- Ask: could a competitor post this unchanged? If yes, it is not finished.
How Postlit handles AI
Our rule has been the same since before the button existed: human-written, AI-assisted. We use AI for research, transcription and analytics, never for the words. Every post starts from a monthly voice interview with you, is written by a person on our team, and goes live only after you approve it. As Muskan Agrawal, founder of Venture Voices, put it: "Every post sounds like me, not a template."
If you are a doctor, CXO or founder who wants to post consistently without sounding like everyone else, start with our LinkedIn ghostwriting service or read how we work with CEOs and executives.
Want this running for you?
Get a free LinkedIn teardown. A founder reviews your profile and tells you what is costing you reach and what we would change first, whether or not you hire us.
Led by a Toptal Verified Expert in Marketing and a doctor turned content strategist.
Frequently asked questions
What is LinkedIn's AI slop button?
It is an option in a post's three-dot menu, labeled "Seems like AI slop", that LinkedIn began rolling out on July 30, 2026. Members use it to report posts they think are low-quality AI output. LinkedIn says the reports help tune its models, and the author is notified privately in post analytics.
Does getting flagged as AI slop reduce my reach?
LinkedIn says reach depends on many signals and that it has safeguards against individual reports unfairly targeting members. It has not published thresholds. It has said content it classifies as slop gets about 40% fewer views, so repeated flags on posts that also look AI-generated to its classifiers are likely to hurt.
Is it allowed to use AI to write LinkedIn posts?
Yes. LinkedIn has not banned AI and still offers an AI proofreading tool. The risk is not a rule violation but reader reaction and ranking. Posts that read like generic AI output get reported and shown less. Using AI for research, transcription or proofreading, while writing the words yourself, avoids that.
How can I tell if a LinkedIn post was written by AI?
Common signs include "it's not X, it's Y" framing, rhetorical question openers, lists of exactly three, one-line dramatic closers, heavy dash use, symmetrical paragraphs and no specific names or numbers. None proves AI on its own, but several together, with no detail only the author could know, usually do.
Is LinkedIn ghostwriting the same as AI slop?
No. Good ghostwriting starts with interviews, uses your real stories and opinions, is written by a person in your voice, and you approve every post. AI slop is generic text generated from a prompt with no lived detail. The first is how executive communication has long worked; the second is what LinkedIn is demoting.