Contents
Comments are the replies members write under a LinkedIn post or article, which let people join a conversation and which LinkedIn says can extend a post's reach1. LinkedIn's engineering team described ordering the comments under a post separately for each viewer, and a LinkedIn patent describes using programs trained on examples to filter out spam and offensive ones23. Authors can choose who may comment on their posts, including no one, and readers can report a comment45. LinkedIn's Pressroom described “Seems Like AI Slop” in September 2026 as a way for members to report posts and comments they believe are low-quality, generic, inauthentic content6.
Why comments matter for reach and the feed
Comments help personalize the LinkedIn feed because its systems track social actions such as commenting on another member's posts or articles and use that data to provide relevant content7. A 2017 LinkedIn blog post gave an example: a member saw a post because one of his connections had commented on it, and that led to a fellowship1. LinkedIn's Help Center tells creators to judge whether content resonates by looking first at the conversations it starts rather than only views, and recommends replying to article comments to build a community and reading them for ideas for the next article8.
Comment settings and reporting
To change who can comment on a shared post, go to the post, click the More icon on its upper right, click Who can comment on your post?, choose Anyone, Connections only (members you are directly connected with) or No one, and click Save4. LinkedIn's Help Center says you can change the setting before or after you share a post4. When you turn off comments, existing comments are not deleted, and the setting you choose is saved as the default for your future posts4. To report a comment on desktop, click the More icon on the right of the comment, choose the report option, then pick the reason that best fits in the “Why are you reporting this?” pop-up5. On mobile, tap the Comments below the post in the feed, tap More on the comment, tap Report and pick a reason, and LinkedIn says it removes reported items only if its team finds they break the User Agreement5. In August 2025, LinkedIn's DSA Systemic Risk Assessment Report said it would restrict members who repeatedly violate its Professional Community Policies from features including commenting and editing past comments14.
Automated, AI-written and inauthentic comments
LinkedIn's Pressroom and Tim Jurka treat automated and inauthentic comments as a problem that shapes which comments readers see and who may keep commenting1516. Engagement pods, as Oscar Rodriguez described them, are coordinated groups who like, comment and share posts to boost virality17. In September 2025, Oscar Rodriguez wrote that LinkedIn was cracking down on third-party tools and reducing automated comments, which can affect how a poster's content is distributed17. LinkedIn's Pressroom defines automated comments as comments posted with a browser extension, script or third-party tool, without a human clicking the comment button, and says they are not allowed15. In March 2026, LinkedIn's Pressroom said detected automated comments may be left out of the “Most Relevant” comments and their authors restricted15. Tim Jurka wrote in March 2026 that LinkedIn is working to make engagement pods ineffective and curb comment automation, and that this is not about limiting reach or discouraging participation16. In June 2026, LinkedIn's Pressroom said it would also take action on comments posted at scale with automation tools and little or no human involvement, and on responses that only restate the post18. Laura Lorenzetti wrote in May 2026 that content that appears generated by AI and lacks clear perspective is less likely to be widely distributed beyond a person's immediate network, and that members can now filter comments and conversations in the feed to LinkedIn's 100M+ verified members19.
How it has changed
- 2017
September
LinkedIn's engineering team described a system that personalizes comment rankings for each viewer2.
- 2025
August
LinkedIn's DSA Systemic Risk Assessment Report said it would restrict members who repeatedly violate its Professional Community Policies from features including commenting and editing past comments14.
September
Oscar Rodriguez wrote that LinkedIn was cracking down on third-party tools and reducing automated comments, which can affect how a poster's content is distributed17.
- 2026
March
LinkedIn's Pressroom said detected automated comments may be left out of the “Most Relevant” comments and their authors restricted15.
June
LinkedIn's Pressroom said it would also take action on comments posted at scale with automation tools and little or no human involvement, and on responses that only restate the post18.
September
LinkedIn's Pressroom described “Seems Like AI Slop”, a way for members to report posts and comments they believe are low-quality, generic, inauthentic content6.
How comments are ranked for each viewer
From LinkedIn's research and patents, which may not describe how LinkedIn works today
In September 2017, LinkedIn's engineering team described information used to rank comments; a Microsoft patent also describes such information29. Its system used a machine learning model, a program that learns patterns from examples, to predict whether a member would engage with a comment, using close to 100 pieces of information about the comment, the commenter and the viewer2. These covered how close the viewer and commenter are (connection and follow relationships, profile similarity, past interactions), the comment's own text (language, length, grammar, hashtags) and its freshness (when it was created and last replied to or liked)2. A 2020 Microsoft patent likewise describes scoring each comment using the commenter's employment history, skills and network relationship to the viewer, and the number of likes and replies the comment gets9. Another 2020 Microsoft patent says a comment from someone identified as a leader on a topic may be ranked higher10. Our read a comment from someone you are connected to or have interacted with probably appears higher for you, because both the engineering post and the 2020 patent use the closeness between viewer and commenter29.
Filtering low-quality and irrelevant comments
From LinkedIn's research and patents, which may not describe how LinkedIn works today
In an August 2026 LinkedIn post, Hari Srinivasan said members can flag posts and comments as “seems like AI slop”11. He said feedback is considered alongside other information and that safeguards help prevent individual feedback from unfairly targeting members11. LinkedIn and Microsoft patents provide historical examples of ways to filter inappropriate or irrelevant comments31213. A 2017 LinkedIn patent says some comments are spam, offensive, or have no supporting text, such as a comment that is only an email address or a link, and describes training a classifier, which is a program taught on labeled examples to sort comments, to filter them out3. The same patent describes sending subsets of comments to crowd reviewers for labeling3. A 2018 Microsoft patent describes identifying irrelevant comments on an article13. It says the system compares an indication of a comment's irrelevance against preset cutoff values: the comment can be removed from the article if the indication exceeds a cutoff, or identified for review if it exceeds a first cutoff but not a second13. Another 2018 Microsoft patent describes hiding selected comments on a social networking service12.
See also
References
- 1New Features to Get the Most From Posting on LinkedIn
- 4Turn off and limit comments on posts
- 5Report inappropriate or offensive updates in your LinkedIn Feed
- 6How LinkedIn is Continuing to Tackle AI Slop
- 7Visibility and impact of your social activity on the LinkedIn feed
- 8Increasing distribution of your articles
- 14DSA Systemic Risk Assessment Report, 2024–2025
- 15What We’re Doing to Support Authentic Content and Conversations on LinkedIn
- 18Keeping conversations real on LinkedIn
LinkedIn's leaders
LinkedIn research and patents
- 2Serving Top Comments in Professional Social Networks
- 3Leveraging annotation bias to improve annotations
- 9Usability and resource efficiency using comment relevance
- 10Identification of a trigger-type leader in a social network
- 12Transfer learning for bilingual content classification
- 13Context-aware approach to detection of short irrelevant texts