LinkedIn Now Lets Members Report Your Post as AI.

On 30 July 2026, LinkedIn added a “Seems like AI slop” option to the three-dot menu on posts and comments.

Chief product officer Hari Srinivasan announced it in a post on the platform, alongside something that matters far more to anyone publishing there. LinkedIn says it will test a way to privately flag, in your own analytics dashboard, when members feel your post came off as ingenuine or leaned heavily on AI.

The button is the headline. The dashboard flag is the story.

For the first time, whether your content reads as machine written stops being a question of taste inside your team and becomes a number someone can point at.

What LinkedIn announced, and what stage each piece is at

Four things moved, and they are not at the same stage. The difference decides whether you act now or watch.

• Members can report a post or comment as “Seems like AI slop” from the three-dot menu. LinkedIn describes this as ramping up, and has said the option applies to posts and comments.
• New classifiers are being rolled out to identify AI slop and generally low-quality posts. LinkedIn says this will reduce that content in suggested posts and in content from outside your network.
• The private analytics flag for people who share content is a stated test, in future tense. Srinivasan’s wording was “we will test a way to privately flag”. Nobody should assume it is sitting in their dashboard today.
• The “enhance your post” AI writing button is being retired and replaced with a tool that proofreads your words rather than rewriting them.

That last one is the tell. The platform that put a rewrite button on every composer has decided the rewrite button was part of the problem.

The dashboard flag is a feedback loop, not a punishment

Reporting tools are moderation, and moderation is LinkedIn’s problem to run.

A private count in the poster’s own analytics is a different thing entirely. It is a platform offering to tell a brand, inside the brand’s own reporting, that readers thought a machine wrote the copy.

Srinivasan’s framing was that AI and slop are not the same thing, that plenty of people refine their thinking with AI, and that those people want to know when they sound ingenuine. He was explicit that the feedback should come from real humans reading the post rather than from an AI detector that gets it wrong.

Hold on to that distinction, because it governs everything else. The signal is not “you used AI”. The signal is “it read that way to a person”. That is the sharper version of the question underneath every sensible argument about how AI should be used in a content team.

It settles an argument most teams cannot settle

Ask three people whether a caption sounds like the founder and you get three answers. The most senior one wins, and nobody learns anything.

Voice has always been the least evidenced thing in social. Reach has numbers. Retention has numbers. “Does this sound like us” has a meeting.

A count of members who read a post and thought a machine wrote it is not a perfect measure of voice. It is still a measure, and it belongs to the audience rather than the reviewer.

That changes what a content review is for. Instead of arguing about whether a draft sounds right, a team can start comparing which drafting methods produce flags and which do not.

What LinkedIn has not said

Three gaps, and they are the ones that decide how much weight to put on this.

• Availability. The analytics flag is a test. LinkedIn has not said who gets it, when, or in which markets. Some coverage has described it as already live; the company’s own wording is future tense.
• Consequence. LinkedIn has said its classifiers will cut slop in suggested posts and out-of-network content, and that member reports help tune those models. It has not published what a single report does to a single post’s distribution.
• Definition. Srinivasan said slop is hard to define and the definition changes, which is exactly why LinkedIn is asking members instead of trusting a detector.

So treat the flag as a quality signal you will eventually be able to read, not a penalty you can model. Anyone quoting you a figure for what one flag costs in reach is guessing.

The answer is not to stop using AI, it is to stop it writing in your voice

The useful split is between scaffolding and opinion. AI is genuinely good at the scaffolding and genuinely dangerous at the opinion.

• Research: gathering what has already been said on a topic before anyone writes.
• Structure: turning a messy voice note or a call transcript into an ordered set of points.
• Recall: checking a draft against everything the account has already published so you stop repeating yourself.
• Options: producing twenty hooks so a human can reject nineteen.
• Proofing: catching the typo, the wrong product name, the claim that does not hold.

What it should not do is write the sentences that carry the view. The moment a model supplies the final wording of the thing you actually believe, you have handed over the only part of the post a reader can tell apart from everyone else’s.

That split is a workflow decision before it is a tooling one, and it usually means rebuilding the drafting and approval process rather than bolting another tool onto the front of it.

What good looks like

A team that can answer one question about any published post: who wrote this sentence.

• The opinion starts as a human artefact. A voice note, a recorded call, a message in a thread. Something with a person’s actual words in it.
• AI shapes that raw material. It never supplies the view.
• One named reviewer owns voice, and now has an external signal to check their instinct against.
• Anything generic gets killed at draft stage rather than defended because it took effort to produce.

None of that is new advice. What is new is that the platform is about to grade you on it.

Founder-led accounts carry the most exposure

A great deal of brand content on LinkedIn is published under a person’s name rather than a company’s. That is the format the platform rewards, and it is why founder-led accounts do so much of the work in business-to-business social.

It is also where this lands hardest. A company page reading as machine written is a poor look. A named founder reading as machine written is a credibility problem, because the entire proposition was that a real person was thinking out loud.

If that is your model, the drafting process has to protect the voice rather than accelerate past it. That is the case for a LinkedIn content system built around capturing what the founder genuinely says, not around filling a calendar.

How NBK thinks about AI in a content system

NBK treats social as an operating system. AI sits inside that system as a tool for speed and recall, never as a substitute for the thinking the system exists to move.

The reason this announcement is worth your attention is not the moderation. It is that a platform is about to hand publishers an outside opinion on one of the few things they have never been able to measure. It would be easy to read that as a threat. The better read is that it is the first honest feedback loop most brands have ever had on whether their copy sounds like a person.

Next step

If your social output feels busy but not effective, and nobody can say for certain whether it still sounds like you, start with an audit of how the content actually gets made. NBK can help find the constraint in the system before the platform finds it for you.