"Does LinkedIn's 'AI Slop' Button Actually Do Anything?"

Dear Rich,

What is your take on LinkedIn’s new ‘seems like AI slop’ button? I am over the moon that the end is nigh for the endless tidal wave of mundanity.

Will, London

Dear Will,

I feel really bad to burst your bubble mate, especially when I enjoyed the fantastic way you framed your question!

Unfortunately, there are a lot of incorrect assumptions being made across LinkedIn about what this ‘seems like AI slop’ button actually does. But, no, it will not do what we all hoped it would.

Hari Srinivasan, their Chief Product Officer, announced the change directly in a post on the platform. When you click ‘Seems like AI slop’ on a post, it disappears from your feed, much like the 'not interested' button and you get a confirmation message thanking you for the feedback.

That feedback helps train the reach algorithm. It is not a content police force. It does not get a post deleted. It does not queue it for human moderation. The poster does not get told. It is only a signal that feeds LinkedIn’s classifiers, which then decides how far that post travels outside of the poster’s own network. Report enough slop, and in theory, LinkedIn gets better at recognising the shape of slop before you have to.

Hari also confirmed that this is part of a series of classifiers and that this ‘user flagged button’ is only one input into that system. If these models think your writing sounds inauthentic or heavily AI assisted, users will get some sort of ‘private flag’ on your own analytics. There will be no ‘public shaming’. In line with this direction, LinkedIn have killed their own AI writing assistant, called ‘enhance your post’, that used to offer to rewrite your drafts for you.

So flagged posts don’t vanish from the network of an offender. It can just trim distribution beyond their immediate connections. And that is not guaranteed to even happen.

And a private flag is a warning shot, not a punishment. If your analytics start telling you your content reads as inauthentic, that's its model giving you a chance to adjust before your reach quietly shrinks. But do we even know where one can view those analytics? I’d put money on the majority saying they would struggle to guess where that data would be shown.

It is good to see LinkedIn take some positive action. And we all want more. But this new button will likely not return LinkedIn to its former state just a few years ago, when all we needed to worry about was those self-appointed leadership gurus suffering from ‘Social Influenza’. Who on this earth would ever have thought that we'd eventually look back on that time as the good ol’ days?

LinkedIn has a hell of a job in front of it. It already blocks hundreds of thousands of automated slop comments every single day, whilst also stopping billions of other automation attempts like ‘scaled posting’.

Everyone with common sense can see the problem getting out of hand. The amount of AI comments on other people’s posts is frustratingly common. Who on earth gains from that? Why are real humans thinking it’s ok to do that? What do bot farms have to gain from it?

In LinkedIn’s defence, AI Slop is extremely hard to define, especially when, apparently, a scary amount of the human population is very poor at spotting it. And an even scarier amount of people think their AI generated posts are a good idea.

Structure seems to be the tell, not vocabulary. AI tools trained on the same data produce the same shapes, even when the words differ. A hundred posts with different phrasing but an identical three-part “here's what I learned” skeleton are exactly what both human readers and LinkedIn’s classifiers have learned to spot.

If you use AI to tighten, or check the grammar of, a real point you actually hold, that is a different category to someone posting generic faux leadership pitches that all involve an ‘uncomfortable truth’ and numerous em-dashes but that is extremely hard to police automatically and accurately. With clear commercial repercussions for 'false convictions'.

AI detection company Pangram’s data found that around 40% of long-form LinkedIn posts and roughly 30% of short-form posts were flagged as fully AI-generated. They also found LinkedIn accounted for nearly two-thirds of all AI content it flagged across the platforms it scans, despite LinkedIn posts making up only a third of the items scanned.

According to Pangram, LinkedIn is the worst offender among major platforms hosting long-form content they analysed.

Substack’s CEO, Chris Best, made the comparison explicit when he announced his own platform’s AI detection integration with Pangram, saying bluntly that Substack didn’t want to “turn into LinkedIn.”

That's the reputational backdrop that LinkedIn is finally starting to respond to.

The platform that pushed us to use their tool to generate AI posts is now asking users to report it and help it stem the tide.

To complicate matters, LinkedIn’s parent company, Microsoft, has a direct commercial stake in AI’s success across its product suite. LinkedIn has not publicly addressed the apparent tension between encouraging AI adoption across Microsoft's products while simultaneously trying to reduce low quality AI generated content on its own platform. And they likely never will.

What should LinkedIn do next? What can they do?

Some have pointed out that maybe LinkedIn should consider a genuinely useful counterpart button, one that lets users flag a post as clearly human and well-argued. Reddit style. But we all know that can be manipulated by these dark forces too.

The system currently has a stick and no visible carrot.

So, Will, to answer your actual question: no, this will not change anything in the way you’re picturing.

For all intents and purposes, this new button is simply a training signal. And a little bit of positive PR for the platform itself.

Is this the end of LinkedIn? It could be. If they don’t find a solution, someone else might.

The AI era is proving one thing. Real human relationships, real interaction, in real life is coming back in vogue.

Onwards!