LinkedIn is starting to push empty AI comments out of sight. That is a relief if your feed has filled up with replies that could sit under any post. It is also a worry if you work in India, write in English as a second or third language, and use AI to make your writing clearer.
My take is simple. The algorithm is not punishing AI. It is punishing empty writing. Use AI to sharpen what you actually think, not to fake engagement. The fix is a clear personal point of view, not avoiding the tools.
What LinkedIn has actually changed
Hari Srinivasan, LinkedIn's chief product officer, announced the comment changes in a November 2026 update. I read the wording as it was displayed in a LinkedIn embed of his post, and it matches the quotation published by Social Media Today on 15 November 2026. Srinivasan wrote: "We are not slowing down on reducing AI Slop. This week's update is on comments, an area that pretty clearly needed attention." He said LinkedIn had launched "new classifiers that help reduce the amount of slop comments that show up in the 'most relevant' section," and that, for people creating content, the company had "started to create smarter notification systems, so we only notify you on high-quality comments you're likely to care about." Then: "In testing we're already seeing a >20% reduction in views of slop comments, and they will be rolled out to everyone soon."
That is a ranking change, not a delete button. Suspected slop comments can still exist. They are less likely to appear in the default most relevant view, and the author of the post may not be notified about them.
This sits on top of earlier, official measures. In a LinkedIn newsroom post dated 15 September 2026, the company said it had rolled out a "Seems Like AI Slop" button in August, and that "more than one million members have used it" within the first two weeks. The same post said recent efforts had "helped reduce views of content classified as AI slop by 40%." That 40% figure is about classified slop content, not the later comment test.
Importantly, LinkedIn has also said what it thinks slop is. The newsroom post defines AI slop as "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 also includes content that feels "generic, repetitive, recycled, or designed primarily to game attention rather than add to a meaningful conversation."
In a 30 July 2026 post I read on his profile, Srinivasan was even more direct: "First, AI and slop are not the same thing; many people refine thoughts with AI, and we believe they want to know when they sound inauthentic." In the same post he said LinkedIn was removing the "enhance your post" feature and replacing it with one that "proofreads your words, but does not change your voice." He also offered a reason people turn to AI in the first place: "LinkedIn isn't a one-word kind of place and they feel more confident running their posts through AI."
That last line should sound familiar if you write in English at work in India.
What the X file actually says
The same week, attention turned to X's open-sourced recommendation code. I read the post enforcement file in the x-algorithm repository at commit a389166f. I am describing only what that file says.
A header line states that the file is "mirrored from GrowthBook dynamic config; last sync 2026-08-06T16:20:00Z." Among the rules is one named act_add_llm_slop_post_label. It runs when "llm_slop_post" in score.labels. If that condition is met, the rule's action is act_add_post_labels_v2, the label it adds is RiskyHighVizReply, and ttl_msec is set to 2592000000. That figure is 30 days in milliseconds. The file does not define how a post receives the llm_slop_post score label, and it does not say what RiskyHighVizReply does to distribution or reach. I have not filled in that gap.
The pushback, and the India problem
On LinkedIn, the debate is not only about classifiers. In a post I read on 18 November 2026, Siddarth Vijay Anand wrote: "everyone linkedin stop calling people out when they use AI. It's okay!" He added: "there is nothing more cringe when people post screenshots of people sending AI written msgs. As long as it delivers what it has to, I don't think it should matter." And: "No need to publicly humiliate someone for just using some technology."
I agree that calling people out in public is a poor substitute for judgement. I do not agree that delivery is the only test. A comment that "delivers" a generic compliment still adds nothing, and it is exactly the empty writing platforms are now trying to bury.
The sharper risk, for Indian professionals, is different. Many people here use AI because English is not their first language. They want to sound clear, not fake. If a classifier or a colleague treats polished English as a tell, the people who get hurt will not be the ones mass-producing empty replies. They will be the ones who were trying to be understood.
LinkedIn has not said it is targeting second-language writers. Its official definition of slop is about missing substance and a missing point of view. That is the right test. The work, then, is to give the algorithm, and the reader, something only you could have said.
I made a related point in How to use AI without sounding like AI. Use the tool to think and to check. Write the sentences in your own words, even if the English is a little imperfect.
Five checks before you post or comment
- Start with a point only you could make.
A view from your desk, your team, or a mistake you actually made will not sound like a template. If you cannot state that point in one spoken sentence, you are not ready to publish, with or without AI.
- Use AI to sharpen the thought, not to invent the engagement.
Ask the tool what a sceptical colleague would question, or to flag unclear sentences. Do not ask it to "write a thoughtful comment" on a post you have barely read. Srinivasan's own line is the right one: AI can help you refine thoughts. It should not replace them. I set out that habit in more detail in how to use ChatGPT without switching off your brain.
- Avoid the comment that could sit under any post.
"Great insights!" and "This is so important" are empty whether a person or a tool typed them. If your reply does not add a fact, a question, or a disagreement, it is slop by LinkedIn's official definition, because it has no point of view.
- If English is not your first language, keep your phrasing.
Fix grammar if you need to. Do not let the tool sand off the way you actually think. A clear, slightly uneven sentence from you will serve you better than smooth copy that could have come from anyone. That is also the safer path if a classifier is looking for generic machine style.
- Before you publish, ask whether a reader would learn something.
If the answer is no, do not post it. The same test will help in an AI fluency review: judgement, not volume.
Don't avoid AI. Avoid empty writing
Platforms are getting stricter about low-effort text, and they should. Your feed is not improved by a hundred replies that say nothing. Indian professionals still need the tools, especially when English is the language of the job but not the language of home.
Use AI to sharpen what you actually think. Then say it in your own voice. That is the writing LinkedIn says it wants, and it is the writing a hiring manager can trust.
At The Arthi AI Collective, I help working professionals use AI as a thinking partner, and leave with work they can explain in their own words. If that is what you are looking for, explore the programs.