91探花

There鈥檚 only one thing more annoying than the horrendous flood of AI slop on our LinkedIn feeds these days, and that鈥檚 people constantly complaining about the horrendous flood of AI slop on LinkedIn.

Not because it鈥檚 not warranted, but because, for one, it鈥檚 exhausting to hear people constantly complain about the same thing day in and day out (and this is true for most things, regardless of the merit of the concern). And secondly. it鈥檚 actually creating a secondary problem, a trap that many people are falling into without even realising 鈥� that is, being so nervous that things have been AI-generated that they assume everything is AI-generated. Of course, that鈥檚 not true, and going down this road could have some pretty dire consequences.

It feels like a frustration and panic-fuelled frenzy that鈥檚 leading to intense mistrust and, to be perfectly frank, some pretty weird behaviour.

And this has been happening for a while already. I鈥檝e noticed not only oodles of users posting about how sick and tired they are of seeing AI slop all over their feeds, but many more actually accusing people of doing it鈥� accusations that are bold and brazen.

But most importantly? They鈥檙e unproven.

And now, LinkedIn has introduced a new feature that gives users the option and ability to report a post or content as 鈥淎I slop鈥�.听 The feature supports this narrative, and that鈥檚 not necessarily inherently a bad idea at all, but the questions we should be asking are:

• What is classified as 鈥淎I slop鈥�?
• How does LinkedIn determine that something鈥檚 been written/generated by AI?
• What action will they take if they deem it to be 鈥淎I slop鈥�?

But perhaps more importantly, what happens if they get it wrong?

听

Reporting 鈥淎I Slop鈥�

听

According to recent reports, and my own basic sleuthing, LinkedIn has begun rolling out a new option that allows users to report posts that 鈥渟eem like AI slop鈥�. The feature appears in the three-dot menu on posts and is part of a wider effort by LinkedIn to reduce the amount of low-quality, AI-generated content appearing in users鈥� feeds. LinkedIn has also said it鈥檚 introducing new classifiers designed to identify content that appears generic, repetitive or lacking in original insight.

On the surface, this may sound entirely reasonable.听After all, there is clearly a problem. A recent study by AI detection company Pangram suggested that over 40% of long-form LinkedIn posts may be AI-generated. LinkedIn itself has acknowledged a rise in what it describes as 鈥淎I slop鈥� 鈥� content that sounds polished but lacks genuine perspective or expertise.

But the question I haven鈥檛 yet seen anyone ask amid this LinkedIn AI slop fiasco is: but how exactly do you know?

Not how do you suspect or how do you feel. I鈥檓 asking, how do you know?

Because there is currently no foolproof way to determine whether a piece of writing was produced by a human, by AI or by a human using AI as an assistant, and that is the very simple truth. So how can it be that while we know this is true, we鈥檙e still being convinced that we can and should just report things as AI.

Personally, while I also find generic AI-generated content annoying as all hell when it constantly pops up on your feed, the fact that nobody is questioning this is actually quite concerning. It kind of feels like we鈥檙e policing writing and opinion without having any way of accurately discerning what actually should or shouldn鈥檛 be classified as bad.

听

听

The Rise Of The AI 鈥淲itch Hunt鈥�

听

The strange thing about the AI era is that we鈥檝e become almost as obsessed with spotting AI as we have with using it.

A slightly formal tone? AI.听A well-structured argument? Definitely AI. A sentence that uses an em dash? Straight to AI jail.

The problem is that 鈥淎I slop鈥� isn鈥檛 exactly a technical category (because it can鈥檛 be, we haven鈥檛 found a foolproof way to detect it yet!). It鈥檚 more of a 鈥渧ibe鈥� that鈥檚 absolutely subjective.

Sure, sometimes it鈥檚 obvious, and this certainly was even more true when AI writing was a little newer and less complex. We鈥檝e all seen those posts that read like they were generated by asking ChatGPT to write a motivational speech after three espressos and an inspiring video on Instagram.

But the thing is, sometimes, bad writing is just bad writing. And sometimes good writing is simply good writing.

It really can be as simple as that. And conversely, merely having a feeling that something is AI-generated is not good enough to classify it as AI-generated 鈥� it鈥檚听not听as simple as that.

The internet is full of examples of AI detectors getting things wrong. Students have been accused of using AI when they hadn鈥檛 and writers have been flagged incorrectly when writing a piece and putting it directly into detection systems. Human-created content has repeatedly triggered false positives, while genuinely AI-generated content has slipped through undetected.

That鈥檚 because AI detection is ultimately a probability game, not a certainty game. And the problem is, the concept of AI detection is built on a principle that I, for one, believe is fundamentally contradictory and, thus, means it can never be successful or completely trustworthy. Indeed, the whole point of AI is for it to sound as human as possible, as indistinguishable from human writing as can be, so why are we surprised that it鈥檚 started picking up human habits and style?

听

What Happens If LinkedIn Gets It Wrong?

听

According to LinkedIn, reports from users will help train its systems and improve the quality of feeds over time. The company has also said it uses both automated systems and human review in parts of its moderation process. The latter is, of course, at least a slightly reassuring (although the extent to which this is happening is a mystery).

But even if humans are involved, they鈥檙e still being asked to answer an impossible question:听Was this written by AI?

And even in merely being asked that question, their objectivity has been immediately removed and they鈥檙e already looking at it through a lens of suspicion and doubt.

Unless the author admits it, the answer is usually somewhere between 鈥渕aybe鈥� and 鈥渨ho knows?鈥� Well, it should be, because that鈥檚 the honest truth.

Imagine spending an hour writing a thoughtful post only to have it reported because it sounded too polished?听Or because somebody didn鈥檛 like your opinion? Or because your writing style didn鈥檛 feel sufficiently 鈥渉uman鈥�. I鈥檝e heard stories of many people, many experienced professionals and writers, who have been accused of using AI to write content, posts, CVs and more. And how on earth do you respond to that kind of accusation? Once you鈥檝e lifted your jaw back up off the floor and calmed yourself down from fury to frustration, what do you even say to someone who confidently tells you, like a detective who鈥檚 just solved a big case, that they 鈥渃an tell AI was used鈥� to write something you painstakingly crafted, word for word?

Indeed, I believe that we need to all open our eyes a little. The danger isn鈥檛 just that LinkedIn鈥檚 systems might get things wrong, but rather the danger is that users might. And if enough people begin treating 鈥淎I-generated鈥� as shorthand for 鈥淚 don鈥檛 like this鈥�, the entire reporting system becomes a popularity contest rather than a moderation tool.

听

Are We Fighting AI Slop Or Just Slop?

听

The thing is, LinkedIn was full of terrible content long before ChatGPT existed. There were hustle gurus, there were engagement farmers, there were people posting 17-paragraph stories about how a trip to Tesco taught them leadership.

AI didn鈥檛 invent any of that; LinkedIn has always been kind of a weird corner of the internet, boasting both great opportunities for professionals to network and learn while also being a cesspool of egos.

What it did do, however, was make it faster. In many ways, that鈥檚 why the debate feels so strange. We keep talking about AI slop as though the AI is the problem, but perhaps the real issue is something much older: low-quality content dressed up as expertise. And not just that, but this witch hunt kind of game we seem to play when we decide, as a collective, that we don鈥檛 like something.

If that鈥檚 the case, then the challenge isn鈥檛 merly detecting AI. Rather, it鈥檚 deciding what actually adds value, and that鈥檚 a far more subjective question than LinkedIn鈥檚 new button might suggest.

But it鈥檚 a lot more difficult of a question to answer, so we get a button instead.

Because spotting AI slop is easy when it鈥檚 obvious, but the hard part, now that AI is writing increasingly well, is proving it when it isn鈥檛.