How AI Broke the Internet

An AI-generated manga just hit number one in Japan’s biggest digital store.

And we’re not talking about a hidden masterpiece here. The reviews say it’s bad. Copy-paste characters, flat dialogue. The only thing going for it is the sex scenes. And it’s still number one.

The wildest part is what Mob Psycho’s editor said: “Readers don’t seem to care whether the artist is AI or not.” People clearly notice. They just don’t care either way.

But hold on, that’s just Japan. On TikTok, a fully AI-generated account, think Jersey Shore but with horny fruit, hit 3 million followers in 9 days.

9 days. A human creator can spend years reaching that number. These pieces of shit fruit got there in the time it takes your router to restart.

All this AI-generated garbage got a name: slop. Merriam-Webster, the oldest American English dictionary, named it word of the year for 2025. We went from “selfie” to “slop” in ten years. That sentence alone sums up the arc of the internet.

To give you a sense of scale: for 500 years, humans slowly built up an entire body of knowledge. Books, letters, articles, thoughts. 500 years. Then AI showed up, and in under 2 years it has written more content than humans have in our entire history.

74% of all new web pages are now AI-generated. Automated traffic grew 8 times faster than human traffic last year. And AI agents, basically bots that browse, buy, and transact on the internet on your behalf, grew 7,851% in a year. Not 78%. 7,851%. They’re not just reading the web anymore. They’re operating it.

Thinking about it, googling something feels weird to me now. It’s only been a couple of years and it already feels strange to type something into Google, get hundreds of links back, and have to click through them one by one looking for what you need. When you think about it, that’s a ridiculously tedious process. And we lived with it for years.

So of course, someone hands you a robot that scans all of that for you and serves up the answer pre-chewed. There’s no going back to googling after that.

That sounds convenient for you and me, but for news sites and blogs it’s a death sentence. When Google shows an AI answer, in that new section that shows up as Gemini suggestions, traffic to real sites drops 61%. And 75% of AI sessions end without the user visiting a single external site. The user asks, the AI answers, the user leaves.

The content creator isn’t the destination anymore. They’re just the quarry AI mines, and it usually doesn’t even credit them.

The internet is this flooded right now because the whole world just got handed a tool that turns two lines of text into working code. A few days ago some guy posted on X that he’d built an emulator for those old airport flip-board displays. He was selling it for $199.

The internet didn’t waste any time. Someone replied: “It’s absolutely ridiculous to suggest anyone pay $199 for something that probably took Claude Code 18 minutes to build.” And he did it. Within minutes he had the code generated, pushed it to GitHub, and made it public. He said the hardest part was recording the video.

That’s what we’re up against. Anyone, in 15 minutes, can copy whatever you built and give it away for free. The difficulty of making something on the internet just stopped existing.

And that’s just the tip of the iceberg. The Charlie Gard Foundation’s website, a charity for children with cancer, got turned into a site advertising illegal online casinos. You’d land on the page and all you’d see was “Best non-Gamstop casinos in Britain today.” Written by an author whose profile photo was AI-generated.

There’s a name for this too: parasite SEO. You latch onto a domain that already has reputation, fill it with AI-generated junk, and that’s it. You show up at the top of Google results immediately.

If that doesn’t show you the internet is collapsing, nothing will.

And don’t even get me started on deepfakes. Trump called his generals to check whether a video of a burning aircraft carrier was real. It wasn’t. The Commander in Chief of the United States couldn’t tell real from fake.

That’s the actual problem. The tells that used to give deepfakes away don’t exist anymore. No more weird fingers, no more melting faces. We already crossed that line. And the worst part is that the tech to make these is accessible to anyone with a smartphone.

And now Meta, the biggest advertising company in the world, over $160 billion a year, 98% of it from ads, just launched traib v2. An AI model that predicts which parts of your brain light up when you watch something.

They put over 700 people through brain scanners, showed them movies and podcasts, and recorded more than a thousand hours of brain activity. With that data they trained a model that can now simulate those responses without needing the scanner at all. They basically cloned the brains of 700 people digitally.

Meta already knew what content hooked you based on screen time and clicks. traib v2 is the next step: knowing what literally activates your brain. Now pair that with Instagram and millions of people mass-producing garbage content, optimized not just to get a click, but to fire off specific parts of your head.

This isn’t academic research. This is the engine behind the most addictive content that has ever existed.

So what can we do about it? Honestly, this isn’t the kind of video where I hand you a solution. This is just what the internet is now.

In the end, the only thing that will keep mattering is who you already trust online. Not the platforms, not the content. The people behind them.

Someone who already earned your trust over time. That’s it. There’s nothing else. Thanks for watching.