AI slops are flooding social media.

On LinkedIn, users have expressed concerns about the prevalence of time-consuming AI-generated posts and comments. Reddit, which is often regarded as a platform for authentic human content, is increasingly dominated by AI-generated material. A study found that 15% of Reddit posts are likely to be AI-generated. YouTube is also experiencing a surge in AI-generated ‘faceless’ channels producing hundreds of historical documentaries, movie summaries, horror story compilations and celebrity gossip videos. On Instagram, AI-generated content is heavily customised to appear as high-end lifestyle, travel or fitness material. On TikTok, split-screen videos are emerging, with the top half featuring an AI voice reading a fictional Reddit thread or a fabricated true-crime story, while the bottom half displays a stolen loop from a mobile game.

The galling fact is that some fake AI influencers on social media are racking up more followers than real human influencers!

AI-slop purge is arriving

Reddit is taking some limited measures against AI-slop by tightening moderation and anti-abuse systems. LinkedIn is taking a more proactive approach by suppressing AI-slop rather than removing it. YouTube is taking the most robust action by taking down AI-slop channels entirely.

The signs are clear. Social media platforms are recognising the existential threat AI slops are to their businesses. They are starting to do something about it. The crackdown will intensify in the months ahead.

Do they really care about you?

Social media platforms are cracking down on AI-slop not because they care about their users’ experience. Make no mistake: they do not care about your experience.

All they care is to monopolise your human attention. As I wrote in my Google Trap book,

Basic economic theory says that the more abundant something is, the less valuable it becomes. Conversely, the scarcer something is, the more valuable it becomes. Today, information is dirt cheap, while attention is incredibly expensive. So, anyone who knows the secret art of attracting sustained attention of the masses will become rich.

The fact is, they are agnostic about AI-slops. If AI-slop is more effective at monopolising your attention (i.e., addicting you), they will prefer AI-slop. They will only care about AI-slop if it starts to threaten their business model. For example, if AI-slops become such a turn-off to their users that they stop scrolling on social media, then the platforms will care.

However, while addictive AI slops are still good for their core business model of monetising your attention, there is a looming threat to their business model arising from another direction…

AI model collapse is their greatest nightmare

The parent companies of many social media platforms are deeply invested in AI. Their financial future depends on the success of their AI.

And here is the crux of their problem: they train their AI models on the user-generated content on social media. For example, Reddit licenses its user-generated content to Google to train its AI model. Google trains its AI model on some YouTube content. Google’s AI Overview trains on website content. Likewise, the same is true for Meta.

What happens if social media is flooded with synthetic content produced by AI? Then a phenomenon called AI ‘model collapse’ will happen. As I wrote on Inspiriting.com,

When more and more of the Internet are resorting to using generative-AI to mass produce content, than those AI tools will train their output on synthetic AI-generated content. This forms a positive-feedback loop- people use generative-AI to create content, which in turn is fed into training AI, which in turn is used to create even more AI-generated content.

There are credible research that as this positive-feedback loop continues will degrade in performance over successive generations. Eventually, you will get what is called “model collapse”, where AI is producing unreliable, untrustworthy and poor-quality content.

When AI model collapse happens, their AI products and services will degrade and deteriorate in quality so badly that they will not be able to sell them. Given that they have sunk an insane amount of capital into developing AI infrastructure, training, and capacity, model collapse will threaten to destroy their AI revenue, which will result in the popping of the Wall Street financial bubble (which is dominated by AI), which in turn will bring down the economy.

What should you do now?

If you are using generative AI to produce content on social media, you need to be careful. This is particularly relevant if your AI-generated content is perceived to be “AI slop”, which is lacking in effort, low in quality, or short in meaning, and produced in high volume. These are the types of content that the social media platforms will go after first.

We do not know how the coming AI purge will look like. But you can imagine that it will include restricted distribution, shadow-banning of your posts, take down of your posts, and perhaps even account deactivation. In fact, this is already happening on YouTube as AI-slop channels (with millions of subscribers) are being taken down. Given the skyrocketing unpopularity of AI among the general populace, social media platforms may resort to name-and-shame measures by visibly labelling posts that they believe are AI-generated.

What about high quality AI-generated content?

Believe it or not, there is indeed high-quality AI-generated content that you will not classify as ‘AI-slop’. For example, Chloe VS History on YouTube is an example of awesome videos that are generated by AI. The digital creator of this channel features a photorealistic vlogger who appears to time-travel through significant historical events. These types of videos entice users to stop, watch, and spend time on the social media platform. It is good for their core business model.

However, at the end of the day, high-quality AI-generated content is still synthetic content. Synthetic content brings forth the looming problem of model collapse. So, what is the compromise solution?

Social media platforms allow their creators to voluntarily label their content as AI-generated. They call it ‘transparency’. The truth is, social media platforms do not care about transparency to their users. But they care about transparency for their AI model training. This voluntary label informs their AI model training process to skip such videos. That way, the social media platforms can have the best of both worlds. They can still get to addict their users with content that happens to be synthetically generated by AI, while at the same time, prevent such content from being used to train their AI models.

So, if you create high-quality AI-generated content, make sure you voluntarily label it as AI-generated.

Next threat that no one is talking about: your human-generated content will be collateral damage

Most people despise AI-slop. I have seen many people cheering the coming crackdown on AI-slop because they are sick and tired of consuming useless bland content.

But I see a new threat that no one is talking about.

AI-slop detection

There is one problem: there is no foolproof and reliable way to detect AI-generated content. For example, AI detectors that you find on Google searches cannot reliably tell the difference between human-written and AI-generated text. Some of them give contradictory judgements.

As traditional algorithms cannot be used to detect AI-generated content, AI has to be used to do the job instead. As AI technology is moving extremely fast, the quality of AI-generated content is improving by the day. It is now already tricky for humans to detect AI-generated written content, which means they cannot train AI models to do that reliably.

Worse still, as humans inadvertently read more and more AI-generated content, they unwittingly start to write like AI.

It is a cat-and-mouse game. As generative AI changes and improves, AI-slop detection technology also changes and improves too. Both of them are always in a state of disequilibrium.

Metric that they are not telling you: False positives

LinkedIn proudly announced that

Early results are encouraging. In our initial testing, we’re correctly identifying generic content 94% of the time. That means this kind of content is far less likely to spread beyond someone’s immediate network, and you’ll see less of it from outside your network in your feed. Members have shared they are already seeing fewer of these types of posts in their feeds, and we expect that to continue over time.

The 94% that they quoted was the true-positive rate. That is, their detection algorithm can detect 94 out of 100 AI slops.

Is that impressive?

As an influencer on LinkedIn, I am more interested in their false positive rate. That is, among the AI-slops they have identified, how many of them turn out to be false identifications (i.e. human-written content)? I doubt they are measuring it.

Tug of war between true positives and false positives

The question is, how did LinkedIn achieve the impressive 94% true-positive rate? Perhaps they increased this true-positive rate by increasing the sensitivity of their AI-slop detection algorithm. But is there a price to pay for increased sensitivity?

Yes, there is always a trade-off between the true-positives and false-positives rates. The more sensitive the detection is, the more likely there is going to be false positives. To understand this trade-off, let me give you an analogy. For example, let’s say you are trying to detect whether someone is lying. To make sure you will catch that person lying, you can increase your paranoia (sensitivity). By increasing your paranoia, you will more likely catch a person lying. But your paranoia can also cause you to think that the person is lying when he/she is actually telling the truth.

Will they get it right?

Since even before AI came into the picture (in 2022), Big Tech has been using algorithms to go after bad actors in their services. Very often, innocent people are caught in their dragnet and have their accounts banned. Using AI in their detection algorithms is going to increase the likelihood of false positives. That means innocent people are going to get penalised.

This is already happening, and innocent people are sharing their horror stories:

• TheyWillBanYou.com – this is a community awareness project where innocent people share their stories about getting banned from their Google accounts, losing their memories (stored in Google Photos), emails, and data overnight.
• Reddit forum: Google Photos – you can read up more horror stories about innocent people getting their accounts banned.
• Reddit forum: eBaySucks – again, more horror stories about banned accounts and false accusations of ‘endangering’ the eBay community.
• My personal experience – LinkedIn’s algorithmI gave me a strike for ‘hate speech’ on LinkedIn for combatting misinformation and hate speech.
• Other people’s experiences – in this article, I gave an example of where Instagram banned an innocent user’s account.

You’ve got to be clear about this: Big Tech has a track record of penalising and banning innocent users in their dragnet to catch the bad actors in their system. Most of these innocent users have no recourse because they cannot escalate their case to a human within the Big Tech organisation.

How it is going to affect you?

As social media platforms increase their crackdown on AI slop, you have to be prepared to be falsely accused.

It is bound to happen.

When that happens, the consequences can be as serious as a complete account ban, to unexpected shadow-banning of your posts, or being named and shamed for something you have not done.

What you need to do?

As I wrote in Is your business a serf under social media’s feudalism?,

If you want to be the master of your destiny, then your business needs its castle, which is in the form of a website. Not only that, but you also need to have a strategy to make your website the focal point of your online presence. Rented land, in the form of a Facebook/Instagram page/group, cannot be the focal point of your online presence. You need to find a way to nudge and direct your leads, visitors and customers from the Facebook lord’s rented land towards your castle (website).

This is something I am preaching to myself too. That is why I am writing this article on my own website. I have made the deliberate choice not to publish it on Substack (see my story on how I had fired Substack), LinkedIn Newsletter, or on any social media platforms. I prefer to operate from my own castle than to till on rented land.

Ready to reclaim control of your online presence?

Let us discuss how your website can serve as the cornerstone of your digital autonomy. If you need to talk, please feel free to contact us here.

Please correct the marked field(s) below.

Thank you for Signing Up

Gain your UNFAIR ADVANTAGE!

Do you want razor-sharp, human-written insights from a straight shooter who cannot stand nonsensical, hogwash propaganda from the industry?

Then sign up to our Business Tech & Strategy Insights!

1,true,6,Contact Email,21,false,1,First Name,21,false,1,Last Name,2