In 2026 LinkedIn started treating generic AI writing as a feed-quality problem. It now shows less of it outside the author's network, added a "Seems like AI slop" report option and privately tells authors when members flagged a post. At the same time it says AI help is fine. This guide reads LinkedIn's own announcements, User Agreement and community policies for a founder who drafts with AI: a short answer, what changed and when, a table of what you and your tools may do, what the button does and does not do, a check to run before posting, an illustrative rewrite and a policy you can copy. It is useful without a Soloop account and is not legal advice.
The short answer
LinkedIn does not ban AI-assisted posts. Its stated line is that AI may help you write, but the post should carry your own voice and perspective. Content its systems classify as low-effort AI writing without a clear point of view is not removed; it is shown less to people outside your network. Members can now report a post as "Seems like AI slop", and if enough of them do, LinkedIn tells you privately in Post Analytics.
Automation is a different and stricter question. LinkedIn's User Agreement bans bots and other unauthorized automated methods, including tools that add contacts, send messages or drive engagement for you, and its systems block automated comments at scale. Drafting with AI and publishing an approved post yourself, or through LinkedIn's documented posting API, stays on the right side of both.
What changed in 2026
On May 20, 2026, LinkedIn VP of Product Laura Lorenzetti announced that generic, low-effort AI content would be less likely to travel beyond the author's immediate network, and that automated comments posted at scale and replies that only restate the post would be acted on. LinkedIn said its system identified generic content correctly 94% of the time in initial testing; it did not publish how that was measured or how often human writing is misjudged.
Around July 30, 2026, Chief Product Officer Hari Srinivasan added a "Seems like AI slop" option to the post menu, said new classifiers would reduce slop in suggested and out-of-network content, and replaced the "Enhance your post" writing feature with a proofreader that keeps the author's wording. About three weeks later he reported more than a million clicks on the option, a private Post Analytics message when a post receives enough of that feedback, and 40% fewer views on content LinkedIn classifies as AI slop. All three figures are LinkedIn's own.
Low-effort AI content without a clear perspective is less likely to spread beyond the poster's immediate network; "It's ok to use AI to help you write" if the post reflects your voice; automated comments at scale and restatement replies are acted on; 94% accuracy in initial testing.
Laura Lorenzetti, LinkedIn, May 20, 2026New classifiers for suggested and out-of-network content; members can report posts and comments as AI slop; a private analytics flag is tested; "Enhance your post" replaced by a proofreader; hundreds of thousands of automated comment attempts blocked daily.
Hari Srinivasan, LinkedIn, July 2026Over a million clicks; enough feedback prompts a message in Post Analytics; no single piece of feedback determines distribution; 40% fewer views on content classified as AI slop.
Hari Srinivasan, LinkedIn, August 2026What you and your tools may do
The feed changes decide how far a post travels. The User Agreement and community policies decide what can get an account restricted. Keep the two apart when you choose a tool.
| Dimension | What LinkedIn says | Founder setting |
|---|---|---|
| Draft a post with AI | Allowed if the post reflects your own voice and perspective. | Use AI for a first draft, then rewrite it around something only you know. |
| Proofread with AI | LinkedIn's own proofreader corrects wording without changing your voice. | Fine. |
| Publish an approved post through an app | LinkedIn documents a posting API (Share on LinkedIn) that lets an app post for a member who signs in and grants permission. | Publish only posts you approved, through a tool that uses the official API. |
| Publish generic AI text | Shown less beyond your network; members may report it. | Do not publish it. |
| Comment with an AI tool | Automated comments at scale and replies that only restate the post are acted on. | Write comments yourself; AI may help you understand a post, not speak for you. |
| Send connection requests or messages with a tool | The User Agreement bans bots and unauthorized automated methods, including to add contacts or send messages. | No. Send them yourself. |
| Engagement pods and pre-arranged likes | Do not artificially increase engagement, including agreeing in advance to like or re-share each other's posts. | No. |
| Scrape profiles or run a browser extension that automates the site | Not permitted; accounts may be restricted or shut down. | No. |
| Run an AI persona account | A member profile may only be created for yourself, a real person. | No. |
Section 8.2: do not use bots or other unauthorized automated methods to access the Services, add or download contacts, send or redirect messages or create inauthentic engagement; do not use software to scrape; create a profile only for yourself.
LinkedIn User AgreementThird-party crawlers, bots, browser plug-ins or extensions that scrape, modify the site or automate activity are not permitted; accounts risk restriction or shutdown.
LinkedIn Help: Prohibited software and extensionsDon't artificially increase engagement, including agreeing in advance to like or re-share; don't spam; disclose synthetic media of a person.
LinkedIn Professional Community PoliciesThe w_member_social permission is required to create a post on behalf of an authenticated member; 150 requests per member per day.
Microsoft Learn: Share on LinkedInWhat the "Seems like AI slop" button does
The option sits in the three-dot menu on a post. Reporting a post hides it from the reporter, tells LinkedIn's models what that member considers slop and counts toward the private notice the author may see. LinkedIn says it uses many signals to set a post's reach, that no single report determines distribution and that it has safeguards against members targeting someone unfairly.
What LinkedIn has not published: how many reports trigger the notice, how much reports weigh against other signals, whether a flagged post can be appealed and how often human writing is misclassified. Treat any tool or post that claims to know the threshold as guessing.
Option in the post's three-dot menu; the reporter sees "Thanks for letting us know"; the post is hidden for them and the feedback trains LinkedIn's models.
TNW, July 30, 2026A badge in post analytics reading "some members told us this post seems like AI"; LinkedIn uses many signals and safeguards so individual feedback cannot unfairly target members.
Engadget, August 21, 2026What LinkedIn says it looks for
LinkedIn has described the target rather than the signals: low-effort AI content that lacks a real perspective or substance, engagement bait, recycled thought leadership and posts with obvious AI construction patterns. Engadget reported that LinkedIn had already reduced the reach of common AI phrasing such as "it's not X, it's Y". In comments, the target is automation at scale and replies that only repeat what the post said.
The scale explains the urgency. The AI-detection company Pangram estimated in July 2026, from scans shared by users of its browser extension, that more than 40% of long-form LinkedIn posts were fully AI-generated. That is one vendor's detector on an opt-in sample, not a LinkedIn figure.
Engagement bait, recycled thought leadership, generic content lacking authenticity and originality, and obvious AI patterns such as "it's not X, it's Y".
Engadget, May 20, 2026More than 40% of long-form LinkedIn posts flagged as fully AI-generated in opt-in extension scans of posts over 50 words.
Pangram, July 9, 2026Check each post before you publish it
LinkedIn does not publish its classifier, so this list is built from what it says it rewards and reduces. Run it on any post an assistant drafted, and on your own drafts too.
- The first two lines contain something specific to you: a number you observed, a customer's words (with permission), a decision you made or a mistake you fixed.
- You could defend every claim if a reader asked how you know.
- It states one opinion you actually hold, even if it is narrow.
- It does not open with a contrast formula, a one-word line or a question you answer yourself.
- It does not end by asking for a keyword in the comments, likes or shares.
- You read it aloud and it sounds like how you talk to customers.
- A tool did not choose who to tag, and every person tagged is part of the story.
- You will be around for the first hour to answer comments yourself.
An illustrative rewrite
Illustrative example, not a real customer. A typical AI first draft: "Most founders think distribution is about posting more. It isn't. It's about showing up consistently with value. Here are 5 lessons I learned about growth. Comment GROWTH and I'll send you my playbook." It has a contrast formula, no specific fact and an engagement-bait ending.
The same idea rewritten around one observation: "Last week I emailed 40 trial users who never connected their calendar. 11 replied, and 7 of them said the same thing: they didn't know which calendar we needed. We changed one sentence on the setup screen. If you run a trial, the people who stall are often the fastest research you can get." The numbers are invented for this example; in a real post they must be yours. 11 of 40 is about a quarter (27.5%), and 7 of 11 is most of those who replied, which is the honest way to describe them.
If a post gets the analytics notice
The notice tells you some members felt the post seemed like AI; LinkedIn has not said it is a penalty on its own or that it carries over to later posts. Treat it as reader feedback.
- Compare the post with your recent ones on the same topic: impressions from outside your network are the number most likely to move.
- Run the pre-post check on it and note which items it failed.
- Do not delete and repost the same text; LinkedIn's spam rules discourage repetitive content.
- Write the next post yourself from a specific observation, and use AI only to proofread or tighten it.
- If you use a tool that drafts posts, change its instructions so drafts start from your notes rather than a topic.
Copy the AI policy
Paste this into your assistant's instructions or your own notes and replace the bracketed fields. It turns the table above into rules a person and a tool can both check.
Where Soloop fits
Soloop does not connect to LinkedIn and does not post, comment or message there for you. According to the Soloop manual, it can research with sources, post on X after you connect it with OAuth, prepare Reddit posts for you to publish yourself and send cold email through a connected Gmail account, with actions waiting for your approval by default. For LinkedIn, that means research and drafts you rewrite and publish yourself.
A bounded first request: I sell [product] to [role]. Find, with sources, three recent public discussions where [role] describes [problem], and list the specific facts and quotes in them. From my notes below, draft one LinkedIn post that starts with one of my own observations. Do not post anything.
Sources, limitations and refresh
LinkedIn's May 20, 2026 announcement, Hari Srinivasan's July and August 2026 posts, the User Agreement, the Prohibited software help page, the Professional Community Policies and the Share on LinkedIn documentation were read on November 7, 2026, with Engadget and TNW reporting for dates and details the posts show only in relative time. The 94%, one-million and 40% figures are LinkedIn's; the prevalence estimate is Pangram's. The table's founder settings, the check, the rewrite and the policy are editorial guidance, not legal advice or a statement from LinkedIn.
This guide will be reviewed by February 7, 2027, or sooner if LinkedIn documents its signals, changes the report option or updates the User Agreement.
The Soloop manual describes research with sources, X publishing with OAuth, Reddit preparation with manual publication and cold email through a connected Gmail account; approval comes first by default. It does not connect to LinkedIn.
Soloop product manual