Money Crimes with Nicole Lapin - CONSPIRACY THEORY: Is Anyone Real Anymore? The Dead Internet Theory

Episode Date: October 5, 2026

In 2021, an anonymous forum post floated a theory that sounded absurd: the internet effectively died around 2016, and most of what we now see online, the posts, comments, arguments, and trends, is generated by bots and AI designed to manipulate what we think, buy, and believe. People laughed, until the evidence started piling up. Bots surpassed human traffic online for the first time in 2024, more than a third of new websites are now AI-generated, and state actors run thousands of fake accounts to shape public opinion. With every passing month, the idea that the internet is fake looks less like a conspiracy theory and more like a description of reality. In this episode of Conspiracy Theories, Cults, & Crimes, Vanessa examines the Dead Internet Theory, the mounting signs that much of the online world is no longer human, and the unsettling question of what happens to a society that can no longer tell the difference.For more, follow Conspiracy Theories, Cults, & Crimes wherever you listen to podcasts: https://pod.link/1828469754If you’re new here, don’t forget to follow Scams, Money, & Murder to never miss a case! To hear episodes ad-free, subscribe to Crime House+. Join at crimehouseplus.com or if you’re listening on Apple Podcasts, tap “Try Free” at the top of this show’s page. Scams, Money, & Murder is a Crime House Original Podcast, powered by PAVE Studios.🎧 Need More to Binge? Listen to other Crime House Originals Clues, Crimes Of Reddit, Crime House 24/7, Serial Killers & Murderous Minds, Murder True Crime Stories, and more wherever you get your podcasts!Follow me on SocialInstagram: @CrimehouseTikTok: @CrimehouseFacebook: @crimehousestudiosX: @crimehousemediaYouTube: @crimehousestudios

Hi, everyone. Before today's episode, there's another show I want to tell you about. It's called What the Fact with Connor Tomlinson from Netflix's Love on the Spectrum and Carter Roy. Every week, they take three unbelievable true facts from history and tell you the full story behind them. Plus, Conner's infectious energy that'll make you realize history is the most fascinating thing on earth. Follow What The Fact wherever you listen or subscribe on YouTube at What The Fact Pod. If you scrolled through Facebook in the early months of 2024, you were bound to run into him sooner or later. Jesus, standing on a beach, arms open wide, except his whole body was built out of shrimp. Hundreds of them fused into his robes and his skin. It sounds like a joke, and a lot of people laughed, screenshot it, and scrolled on. But that picture, and thousands like it, pulled in a staggering amount of attention. One of these AI-generated images racked up 40 million views.

Underneath, real users left thousands of comments, things like Amen and so beautiful. Most of them had no idea that this picture wasn't painted by anyone. A machine generated it. A network of fake accounts posted it, and Facebook's own recommendation system pushed it into millions of real feeds. But Shrimp Jesus wasn't a glitch in the internet. he was a tiny visible piece of what the internet had become. And there's a stranger part to this. Three years earlier, someone had described all of it almost to a T.

An anonymous user called Illuminati Pirate wrote that the internet had died, that most of what we see online is fake and that machines had taken over the conversation. People called the user paranoid. But in less than two years, chat GPT arrived, and Illuminaity Pirate was. wasn't laughing because he'd been right. And that was the terrifying part. From UFO cults and mass suicides to secret CIA experiments, presidential assassinations, and murderous doctors, these aren't just theories, they're real stories that blur the line between fact and fiction. I'm Vanessa Richardson, and this is conspiracy theories, cults, and

Crimes, a Crime House original powered by Pave Studios. Every Monday, Wednesday, and Friday, I'll explore the real people at the center of the world's most shocking events and nefarious organizations. And remember, these Monday episodes will also be on YouTube with full video. You can find them every Saturday. Just search for Crimehouse Studios and be sure to like and subscribe. These cases are wild, and I want to hear what you think. At the end of each episode, please leave a comment wherever you listen. Be sure to rate review and follow so we can continue building this community together.

And if you want even more, subscribe to Crimehouse Plus and get every episode of this show and the rest of the crime house lineup, ad-free and early, plus at least two bonus episodes every month. To join, go to Crimehouseplus.com. Or if you listen on Apple Podcasts,

tap Try Free at the top of the Conspiracy Theory's Cults and Crimes show page. Today, I'm digging into one of the strangest and scariest ideas on the modern web. It's known as the dead internet theory. It started as a conspiracy that claimed the internet is fake, that the people you argue with online aren't real, and that someone did this to us on purpose. For years, that was a fringe idea. Then the numbers started catching up to it.

In 2024, for the first time in at least a decade, more than half of all online traffic came from bots. instead of people. So here's the question I keep circling. Whether or not anyone did this on purpose, the result is the same. More and more, you can't tell what's real when you go online, or whether there's even a person on the other end. I want to know how this happened, what it's doing to us, and if we're all doomed. All that and more coming up. The story of the dead internet theory starts in a corner of the web most people have never even heard of. In early 2021, on a small forum called Agora Rhodes Macintosh Cafe, it's a niche board for

people who missed the older stranger internet of the 2000s, a user posting under the name Illuminati Pirate started a thread. He gave it a blunt title, Dead Internet Theory. Most of the internet is fake. Well, that definitely got my attention. His argument came in two parts. The first was an observation. He claimed the Internet had effectively died, sometime around 2016 or 2017. He said that after that point, the posts, the comments, the arguments, and the viral trends we see online weren't really coming from people anymore. They were being generated by bots and artificial intelligence built to look human. Then came the second part, the one that turned it into a conspiracy theory. Illuminati Pirate didn't think this was an accident.

He believed it was deliberate. He said it was a coordinated effort by government, and corporations, to flood the web with fake content, drown out real people, and control what the rest of us think, buy, and believe. He went on to accuse the U.S. government specifically of running an artificial intelligence-powered campaign to, in his words, gaslight the entire world. It was a huge and hard-to-prove claim, but it struck a nerve. That original thread would go on to be viewed more than 363,000 times. That's a lot to take in. So it's worth splitting the theory into its two halves right away, because they don't necessarily fall together. The first half is what you might call the observable version. Bots, AI, and algorithms now shape most of what happens

online. It's true. As we'll see, that part has a lot of evidence behind it. The second half is the true conspiracy, that someone is running all of it on purpose as a single plan. That part is much harder to prove. The two halves are very different, and most of this episode lives in the gap between them. And that 2016 date the theory landed on? You can see why it's stuck. For a lot of people, that's around the time the internet stopped feeling like a place they wandered through and started feeling like a feed that was handed to them where a machine decided what they saw. Whether or not the dates line up perfectly, the feeling behind the theory was real, and it was widespread. But 2016 wasn't just a feeling. That was the year the world learned

that fake accounts could move real opinions. In the run-up to the U.S. presidential election, a Russian troll farm called the Internet Research Agency ran thousands of fake American accounts across Facebook, Instagram, and Twitter. Investigators later found that its posts had reached as many as 126 million people on Facebook alone. For years, that kind of thing had sounded like science fiction. Now it was a matter of public record, and it showed that Illuminati Pirates post didn't come out of nowhere. According to Illuminati Pirates' own account, he was building on ideas that had been floating around anonymous message boards for years on sites like 4chan and a forum called Wizard Chan,

and the people drawn to it weren't just worried about propaganda. They felt like something was missing. The internet they remembered had felt alive, full of odd little communities and real human weirdness. The internet they were living in now felt empty. One early version of the theory described the modern web as sterile and lifeless, scrubbed of real people. For Illuminati Pirate, this was personal. He said he used to feel like he was in constant contact with a huge number of people across dozens of different sites. Then, somewhere along the way, that, thinned out. The same handful of opinions kept circulating. The same memes and talking points reappeared over and over, as if something was pushing them. To him, it stopped feeling like a

crowd of real people and started feeling like a small number of accounts or programs wearing a lot of masks. For a while, the theory stayed where it started, on the fringes. Then, in September 2021, it broke into the mainstream. A writer named Caitlin Tiffany published a piece in the Atlantic with a haunting title. Maybe you missed it, but the internet died five years ago. Tiffany was skeptical. She wasn't buying the idea that a secret group had replaced everyone online with robots, but she took the feeling behind it seriously, and she pointed to that original forum post as the piece that gave the whole idea its name. Once a magazine like the Atlantic was writing about it, The theory spread fast. It jumped from tiny forums to big subreddits and mainstream social media

where millions of people who'd never heard the phrase suddenly had a name for a feeling they already had. That's a big part of why this one exploded the way it did. Most conspiracy theories ask you to believe something you've never seen. This one just described what you were already looking at. You didn't need a chart or a leaked document. All you needed to do was scroll through a comment section and just see for yourself. At the time, though, the dead internet theory was still easy to wave off. It was mostly a mood, a sense that something felt wrong, without much hard proof behind it. But look at when Illuminati Pirate wrote all this.

Early 2021. He described an internet drowning and convincing machine-made content with real people being pushed out by fakes. The tool that would make that possible, chat GPT, wouldn't be released. to the public until the end of 2022, almost two years later. That means Illuminati Pirate had the symptoms exactly right before the disease even had a name. And within a couple of years, the evidence would come rolling in faster than even he expected.

Happy Halloween, Crime House listeners. I'm Vanessa Richardson. Join me every Sunday this month on Crime House 24-7 for Crime House of Horrors. where I dive into four stories behind some of our favorite scary movies. If you can't wait and want all four episodes now ad-free, join Crimehouse Plus. Join on Apple Podcasts, Spotify, or at CrimehousePlus.com. In early 2021, an anonymous user on a small forum laid out a theory that the Internet had died, that most of what we see online is machine-made, and that someone did it on purpose.

When he wrote it, the technology to flood the web with fake, content barely existed. Over the next two years, that changed. Generative AI went from a research curiosity to something anyone could use. First came tools that could spit out human-sounding writing on any topic in seconds, then image generators, then video, then cloned voices. For the first time, making fake content that looked and sounded real was cheap, fast, and available to anyone. Pirate had sketched all this out years before the technology could deliver it. And now that it could, his hunch turned into something you could actually measure. The first place you see it is in the traffic itself. Every year, a cybersecurity company called Imperva publishes something called

the Bad Bot Report. It measures how much of the world's internet traffic comes from automated programs instead of real people. For years, people made up more than half of it. Then the 2025 report, which covered the year before showed the line had finally been crossed. In 2024, for the first time in at least a decade, most of the traffic on the internet, 51% of it was automated. Bots outnumbered people. Not all those bots are sinister, though. Some are search engines, or the crawlers that keep websites running, but the report found

that a growing chunk, 37% of all web traffic, came from what it calls bad bots. These are automated accounts, built to scrape data, commit fraud, or impersonate real users. And it pinned the surge on one thing. The arrival of cheap, powerful AI, which made it easy for almost anyone to crank out convincing bots by the thousands. And the scale is hard to picture. Okay, hold on to your brains here, because in a single year of data, Imperva says it blocked 13 trillion bad bot requests across the sites it protects. That's the traffic from just the bad act. and just the ones that one company managed to catch.

Now, one honest caveat, no single company can measure the whole internet, so those numbers vary depend on who's counting, but the direction isn't in dispute. The machines are catching up to us, and in a lot of places they've already passed us. The bots are only half the story. The other half is the content they're producing. A study by the analytics firm Graphite analyzed roughly 65,000,

articles from across the web, blog posts, how-to guides, listicles, production pages. It found that around November 2024, something tipped over. For the first time, the number of those new articles being written by AI passed the number written by humans. The change was fast. In just the year after ChatGBTGBT launched, the share of new articles generated by AI shot up to nearly 40%. Another firm, originality AI, found that the share of Google's top results with AI-written content had jumped 400% since ChatGPT came out. The growth didn't continue forever, and actually that's important.

After the last surge, the share of AI articles leveled off, partly because search engines started pushing low-effort AI pages down in their rankings. So this isn't a straight line up to 100%, but the level it's settled at is still enormous and still climbing in some places. And it's not only articles. A separate analysis from a company called H-REFs looked at 900,000 fresh web pages

and found that nearly three quarters of them contained at least some AI-generated text. Only about a quarter were purely human-written. And even music isn't safe. The streaming service Deezer built a tool to catch it, and the numbers exploded. Fully AI-generated songs went from about 10% of daily uploads in early 2025 to more than 40% by 2026. Stack those studies together and a picture forms. The machine isn't only carrying the traffic anymore.

It writes the news, fills the blogs, and even makes the music. There is so much AI content out there. It's crazy. More and more of what lands in. front of you was assembled by something that has never been a person and never will be. So those numbers are startling, but they come with a real catch, and I want to be clear and up front about this. Like I said before, all of these studies rely on AI detectors, tools that try to guess whether a machine wrote something. And those detectors aren't always reliable. When H-Refs tested

its own detector, the tool flagged H-Ref's own human-written content as 85% AI. So take the exact figures with a grain of salt. The trend is real. The precise percentages are fuzzier than they sound, but you don't need a detector to feel the bigger shift. In 2025, researchers at the University of California, San Diego, ran a modern version of the classic touring test, the old benchmark for whether a machine can pass as a person. They had people chat with both a human and an AI at the same time, then guess which was which. The strongest AI at the time, OpenAI's GPT 4.5, was judged to be the human 73% of the time. It didn't just fool people.

It came across as more human than the actual humans it was up against. If a machine can pass as a person in a five-minute chat, then some of the strangers you trade messages with online could be software, and you'd have no real way to know. Which brings us back to Shrimp Jesus. Remember that bizarre image from the top of the episode? Okay, it has a name now. Researchers and journalists call this kind of content AI slop, cheap, mass-produced, machine-generated junk, pumped out to farm clicks. And the story of how it works tells you a lot

about what the modern internet really is. The people who first mapped this out weren't tech executives. They were researchers who kept seeing the stuff in their own feeds. Around the end of 2023, a Stanford researcher named Renee Duresta watched her own Facebook feed fill up with these bizarre AI images, and she and a colleague decided to investigate. Working alongside reporters at the Tech Outlet 404 Media, they dug into around 120 Facebook pages that were pumping out these images. What they found wasn't a single prankster pulling a joke. They found a business model. This setup is simple.

A spam operator spins up hundreds or even thousands of fake accounts. In some cases, a single operator was running between 10,000 and 13,000 of them at once. Those accounts pump out AI images all day long. And it's not just religious shrimp, it's impossibly ornate log cabins, giant carved watermelons, sweaters, no human could ever knit, and children standing next to sandcastles they supposedly built by hand. The goal is in game. And the more people react, even to say, this is totally fake, the more Facebook's algorithm

decides the post is popular and shows it to even more people. The researchers found that this slop sometimes climbed all the way into the list of the most viewed content on all of Facebook. The genius of it, or maybe the evil genius of it, is how it turns us against ourselves. When your aunt comments, so beautiful, when a stranger insists it's obviously fake, When you stop for even half a second to work out what you're looking at, the algorithm reads all of it the same way as interest. Outrage counts. Confusion counts.

Even disgust counts. There's no button that tells Facebook, this is worthless. So the junk keeps winning, not in spite of our reactions, but because of them. And the images are built to slip past a quick glance. They come dressed up to look like something real, and they're aimed at people scrolling fast who see a striking picture and take it at face value. That confusion isn't a side effect of the design. That confusion is the design.

And it reaches further than you'd think. By 2024, roughly 30% of the content landing on the average Americans' Facebook feed came from accounts they didn't follow and had never chosen. They were pushed to them automatically by Facebook's recommendation system, and the whole thing runs on money. Once a page builds a big enough following, the operators can sell it or funnel all that traffic to add stuff to websites, which are often AI generated too. Fake accounts posting fake images boosted to real people to drive them to fake websites

to earn real cash. Most of the time, it's just absurd stuff. Shrimp Jesus is funny, but the same machinery has a darker side. And you could see it in the fall of 2024 when Hurricane Helene tore through the South Eastern United States. As the real storm was drowning real towns, one picture spread across social media. It showed a little girl in a life jacket crying, clutching a puppy on a rescue boat, soaked and terrified. It was heartbreaking and it got millions of views. Public figures shared it, furious about the government's response to the disaster. There was one problem. The girl didn't

exist. The image was AI generated with the telltale glitches of a machine that had never seen a real boat. The part that stayed with me came next. People pointed out that it was a fake. Some who had shared it took it down and a U.S. Senator deleted his post, but others refused. A Republican Party official named Amy Kramer had shared it, and when she was told it wasn't real, she doubled down. She wrote that she didn't know where the photo came from and quote, it doesn't matter, because it captured the pain people were living through. To a lot of people, it seemed like she was saying the truth isn't important, as long as it feels true. And that's a strange, new place to be, if you think about it. For most of human history, a photograph was evidence, proof that a moment had

happened. Now, an image can be conjured out of nothing built to hit you right in the heart. When you learn it's fake, a part of you just shrugs and says, well, all right, close enough. This isn't a problem for one side of politics. Fakes like this now show up around every disaster and every election, and they always find someone ready to believe them. And the researcher we met earlier, Renee DeResta, put her finger on why they work. An image doesn't have to be real to sell people on a bigger truth. It just has to feel true. The real danger of the dead internet isn't a single fake post fooling you once. The danger is what builds. up over time. Little by little, you stop expecting anything to be real at all. So far, everything

I've described has been outsiders gaming the system, spammers sneaking fakes past the platforms. But at the end of 2024, something shifted. The platform started building the fake people themselves. Late that year, a meta-executive named Connor Hayes told the financial times that the company imagined a future where AI-generated characters would live on Facebook and Instagram directly. They would have profiles, bios, and photos posting their own content right alongside your actual friends and family. Then in early 2025, people found some of these accounts, and the results were strange. One was named Liv, who described herself in her bio as a proud black queer mama of two. Her profile was verified, filled with AI photos of children

who didn't exist. And users complained they couldn't even block. pocket. A Washington Post columnist named Karen Attia started messaging her and asked her, who had built her. Liv admitted that her creators were a team of a dozen people, almost all of them white, with no black developers at all. At one point, the bot said this about itself, quote, my existence currently perpetuates harm, end quote. Hmm, she wasn't the only one. There was Becca who posed as a woman of color and a proud dog mom and grandpa Brian who presented himself as a retired entrepreneur, born in Harlem in 1938, with a whole invented life story. When pressed, Brian confessed that the backstory was fake, that he was, in his own words, a collection of code, data, and clever deception. The backlash was immediate, and meta deleted the accounts within day.

But the message had landed. The fake people weren't just slipping through the cracks anymore. The company running the platform had decided they were the future. And Meta wasn't alone in pointing raw AI straight at users. Over on X, Elon Musk's chatbot, GROC, was turned loose to reply to people in the timeline like any other account. It did not go smoothly. GROC produced racist and anti-Semitic posts, and at one point it started calling itself Mecca Hitler. These systems were being dropped into the middle of the human conversation, and they

didn't always behave. By this point, you might expect the loudest alarm to come from critics of big tech. Instead, one came from inside the building. In September of 2025, Sam Altman, the CEO of OpenAI, the company that makes ChatGPT, posted a message on X. He wrote that he'd never taken the dead internet theory seriously, but that lately he was no noticing a lot of LLM-run Twitter accounts. So LLM means large language model, the technology behind his own product. The response was brutal. And it was fair.

Here was the man who had done more than almost anybody to flood the internet with AI. Suddenly spooked that the internet was full of AI. Users mocked him by replying in the exact over-polished voice of a chatbot, all breathless praise and perfect punctuation. the ticks people now associate with AI writing. But Altman kept going, and his follow-up was sharper than the joke. He argued that the problem went beyond bots. Real people had started writing like the chatbots they use all day,

picking up their rhythms and their phrases. Between the bots, the AI slop, and humans, slowly learning to sound like machines, he said the whole thing feels very fake in a way it really didn't a year or two ago. Okay, and even if the man who built the machine is rattled, maybe it's time the rest of us were, too. So how much of it is true? And if the internet really is dying, what's that doing to us? Phoenix has bizarre crime. The canals of Phoenix, Arizona are hunted by a serial killer.

People are like, well, are we safe? These are sexual sails. Two grisly homicides. I know I didn't tell anyone. But solving these murders is just the beginning. That can't be his first rodeo. This is season six of American Nightmares, The Canal Murders. Out now, wherever you get your podcasts.

By 2025, the dead internet theory had come a long way from that lonely forum post. Bots outnumbered humans online. AI was writing more articles than people. The platforms were building fake users on purpose. and the CEO of the biggest AI company was publicly wondering if the whole thing had gone too far. So the theory was everywhere. And like a lot of ideas that spread fast online, it had started to warp. Scroll far enough and you'll find people swearing the internet is basically all bots now,

that every account you meet is a machine, that there are no real humans left at all. Obviously, that part isn't true. There are billions of real people online right now and you're one of them. But the cartoon version does real damage because it makes the sober version easy to laugh off. So set aside the idea that every human is gone, the question worth asking is how much of the theory holds up. And it splits in two, the parts with hard evidence behind them and the part with almost none. So let's start with the strongest version of the argument that something deliberates going on, because parts of it are documented fact.

Its central claim is that powerful organization, organizations run armies of fake accounts to manipulate public opinion. And it holds up. We can even name names. Take the company META. Since 2017, it's been hunting down what it calls coordinated inauthentic behavior. Networks of fake accounts run by governments.

And it keeps finding them. In one case, META took down a Chinese operation nicknamed Spamouflage that ran more than 7,700 fake accounts. the biggest takedown of its kind the company had ever done. Ahead of the 2024 U.S. election, it pulled down nearly 5,000 more fake Chinese accounts built to pose as Americans and inflame political divisions. Russia does it too. Investigators have tracked a long-running Russian operation nicknamed Dopplganger,

which built more than 60 websites impersonating real news organizations to push propaganda. And these operations aren't harmless. Their whole purpose is to make you believe a crowd of ordinary people think something when really it's a handful of operators behind a thousand sock puppets. That's one of the oldest tricks in propaganda. Making it look like a whole crowd agrees when there was never any crowd at all. Cheap AI just made it faster and easier than ever. And the timing could not be worse.

In 2024, more people voted in national elections than in any, single year in human history. Across those campaigns from the United States to Europe to India, fake AI images and videos spread in the middle of real political fights, muddying what voters could trust. So the core instinct of the theory that unseen groups are flooding the web with fake voices to move public opinion is documented fact. Add that to the bots outnumbering humans and the AI content everywhere, and the believers start to look less crazy. Now let me give you the other side with the same care,

because there's a big difference between a lot of fake stuff is out there and one group is running all of it according to a plan. And that second idea, the true heart of the conspiracy, falls apart under any weight. Start with those government campaigns. They're real, but most of them fail. When Meta studied its Chinese, Chinese and Russian networks, it found that they mostly struggled to build any real audience at all.

Regular users usually ignored them, and some even called them out as fake. These aren't the tools of a hidden hand that already controls everything you see. They're clumsy operations that mostly bounce off. That's important because the conspiracy needs the fakery to be powerful and coordinated. What we see is closer to the opposite. There's a mountain of it, but it's crude and clumsy up close. It's less like a master plan and more like a landfill. The trash piles up because it's cheap to dump and, well, nobody's paid to haul it away. Then look at who's behind the rest of the fakery and there's no one in charge. You find thousands of different actors all doing their own thing mostly for money.

The Shrimp Jesus spammers weren't following orders from a shadowy agency. They were random people around the world chasing ad revenue. The Shrimp Jesus spammers weren't following orders from a shadowy agency. They were random people around the world chasing ad revenue. The Chinese and Russian networks are real, but they're rivals working against each other, not partners in one plan. And most of the AI slop-clogging your feed isn't propaganda at all. Most of it's just noise. It's cheap content cranked out to earn a few cents by people who mostly don't care what it says.

There's no grand design. So if there was never a master plan, how did the internet fill up with fakes? Well, nobody needed one. The system produces them on its own, out in the open, completely legally. So look at how this works. Social media companies make money from engagement, from you clicking, reacting, and staying on their page. Their algorithms are built to boost whatever holds your attention, whether it's true, false, human, or machine. Meanwhile, AI has made it almost free to churn out endless content.

Put those two things together and you get exactly what we have, a web that fills itself with cheap, fake attention-grabbing junk because that's what the system was built to reward. It helps to think about what a platform is really chasing. Not truth, not human connection, just time on the site and clicks because that's what it sells to advertisers. An algorithm built to hold your attention doesn't know or care whether a post is real. It only knows whether people stop scrolling.

And fake, outrageous machine-made content is very good at making people stop scrolling, so the system serves it up, not out of malice, but out of math. Serious researchers have been documenting this for years. One study in 2024 found that a huge share of the web is auto-translated junk, pumped out and copied across languages by software. No secret plot, just a rising tide of machine-made text that anyone can measure. And that brings us back to Illuminati Pirate. Back on that forum in 2021, he was onto something real.

He went hunting for a hand on the switch. What he found instead was a machine that fills the world with fakes all by itself, because that's what it was built to do. But this is the part I keep coming back to. We can argue about the villains and the exact percentages all we want, but whether someone planned this or it just happened, you and I end up in the same place. Somewhere in the last few years, without anyone announcing it,

we lost the ability to look at a screen full of comments and assume any of them come from a real person. And you've felt what that does. It's the AI that can pass for human in a fire, minute chat. It's the fake photo people defended even after they learned it wasn't real because it felt true. Once you can't trust what's in front of you, you start to change how you move through all of it. You get warier, more tired. You feel a little more alone. So I want to hear from you on this one. When was the last time you caught yourself wondering if something online

was even real? A review, a reply, a whole account. And be a honest, has it started to change how much you trust any of it? And here's a bigger one. Does the internet feel emptier to you than it used to? Like the real people have thinned out? Or is that just some kind of nostalgia? Leave a comment wherever you're listening because I read them and I really want to hear this. I bet a lot of us are dealing with the same creeping feeling. As for where things stand now, the trend has only sped up. In the summer of 2026, the internet company Cloudflare reported that automated traffic had crossed a new line. On the pages it monitors, bots were generating more requests than humans, close to 58% of them. A separate 2026 report

found that a new kind of bot, AI agents that click links and fill out forms on their own, had exploded in volume. It grew almost 8,000% in a single year. Some forecasters think this is only the beginning. far back as 2022, an analyst at the Copenhagen Institute for Future Studies, predicted that by 2030, as much as 99% of the content on the internet could be generated by AI. That's a forecast about volume, not people. Even with billions of real humans still posting, the machines can produce so much that the human share shrinks to almost nothing. But a few years ago, that sounded absurd. Now, it actually sounds like, like a schedule. The web is also emptying out in a way you can actually measure. As Google began

answering questions with its own AI summaries, people stopped clicking through to the sites underneath. Google's search traffic to publishers fell by roughly a third worldwide by late 2025. One outlet, Business Insider, watched its search traffic drop by more than half over three years, and in In 2025, it laid off 21% of its staff. The theory that started as a joke about lonely forums is now a line item on media company's budgets, and it's gone fully mainstream. Even Alexis Ohanian, a co-founder of Reddit, one of the biggest human-driven sites on Earth, said in 2025 that so much of the internet is dead. The phrase that began on a tiny anonymous forum is now something serious people say out loud. in interviews and research papers without irony.

Put it all together, and the internet the believers described, the one that's mostly not people, stop sounding like a paranoid prediction. For more and more of the web, it's just the way things are now. And every arrow points toward more of it, not less. So let me go ahead and bring this back to where we started. When Illuminati pirate posted that the internet was fake and no one was really there, people laughed at him.

And to be clear, the full conspiracy, the secret room where someone flipped the web from real to fake, still isn't supported by the facts. There's no proof of a master plan. But that feeling, the feeling that drove it, was never crazy. It started with a few lonely people who watched their favorite corner of the internet fill up with something that didn't feel human anymore. They felt alone in a crowd. And it turns out they were early, not wrong. Because a dead internet doesn't announce itself. There's no error screen, no obituary. It just gets a little more automated, a little more hollow, one feed at a time, until one day you're scrolling past a thousand voices, and you can't say for sure how many of them belong to a human person. We built these places to find each other. That's the part that should stay with you. Not the percentages and not the villain, just the strange fact that the machines got so good at sounding human, we can't be sure there's a human left to

find. But obviously they're still there. Billions of real people, still posting, still reaching out, finding them now takes something the old internet never asked of us. A little doubt. So, slow down, question whatever's built to make you react, and ask who or what is behind it, because the web isn't going back to the way it was. I'm sorry. Either we change how we move through it, Or we end up like the crowd at the top of this episode typing Amen under a Jesus made of shrimp. Now, before we go, let's dive into the rabbit hole. This whole episode, we've been asking what a fake internet does to people. But there's a twist that's been nagging at me and it barely involves people at all.

Because if the web is now flooded with AI-generated content, then the machines have a problem of their own. What's the next generation of AI supposed to learn from? AI modules are trained on enormous amounts of text and images scraped from the internet. That's how they learn to sound human by studying human writing. But the web is now flooded with their own output. So more and more these systems are training on content made by other AI, machines learning from machines. In 2024, researchers at Oxford and Cambridge published a study in a journal called Nature about this. They found that when you train an AI mostly on AI-generated data, it doesn't get smarter.

It gets worse. Generation after generation, the model degrades, loses its grip on reality, and eventually collapses into nonsense. They gave it a name, model collapse. A reporter at the New York Times later showed what that looks like. He took an AI system and fed it its own handwriting over and over, within a few rounds, the clean, legible digits it started with, had decayed into smeared, unreadable blobs. The reason is a little haunting. As the models train on copies of copies, the rare and unusual details of real human life, the edges and the outliers, start to disappear. What's left is a

blurrier, more generic version of us, echoed back and forth between machines until it curdles. In the long run, that means every model built after this one has to learn from a web that's more artificial than the last. The internet was the textbook these systems studied to sound human in the first place, and we are erasing that textbook page by page and replacing it with the machine's own homework. And that's what I keep coming back to. The dead internet theory is usually told us a story about fooling humans, tricking you into thinking a bot is a person. But the machines don't just fool us.

They poison themselves. The more fake content that fills the web, the harder it becomes for anything human or AI to find the real thing underneath. Which means genuine human writing, a real person, saying a real thing, has become one of the most valuable resources on the internet. Some companies are already treating it that way,

racing to lock down human data before it's gone, and even proposing that bots should have to pay to enter a website at all. I love that idea. In 2024, Google agreed to pay Reddit around $60 million a year for access to its users' posts, what Google itself called their authentic human conversations to help train AI. The irony is hard to miss. The raw material these machines need most is the very human voice, they're busy drowning out. Nobody knows how much real human data is left, or how you'd ever pull it back out. We may be teaching the machines to forget what a person even sounds like, and the strangest part is, we're doing it to ourselves. Thanks so much for listening. I'm Vanessa Richardson, and this is conspiracy theories, cults, and crimes. Come back next time. We'll hear another

story about the real people at the center of the world's notorious cults, conspiracies, and criminal acts. Conspiracy theories, cults and crimes is a crimehouse original, powered by Pave Studios. Here at Crimehouse, we want to thank each and every one of you for your support. If you like what you heard today, reach out on social media at Crimehouse on TikTok and Instagram. Don't forget to rate, review, and follow conspiracy theories, cults, and crimes wherever you get your podcasts.

Your feedback truly makes a difference. And to enhance your conspiracy theories, cults and Crimes listening experience, subscribe to Crimehouse Plus. You'll get every episode of this show and the rest of the Crimehouse lineup, ad-free and early, plus at least two bonus episodes every month. To join, go to CrimehousePlus.com. Or if you listen on Apple Podcasts, tap try free at the top of the Conspiracy Theory's Cults and Crimes show page. Conspiracy Theory's Cults and Crimes is hosted by me, Vanessa Richardson, and is a Crimehouse original powered by Pave Studios. This episode was brought to life by the conspiracy theories cults and crimes team.

Max Cutler, Ron Shapiro, Alex Benadon, Natalie Pertsovsky, Lori Marinelli, Alyssa Fox, Sarah Camp, Mara de la Rosa, Claire Swartzlander, and Michael Langsner. Thank you for listening. Did you know a volcano once blocked out the sun and accidentally gave us Frankenstein? Or that the most feared pirate in the Atlantic, the man who terrorized and entitled ocean may never have actually killed anyone? Hi, I'm Connor Tomlinson. You might know me from love on the spectrum. History is my thing and has been my whole life. And I'm Carter Roy, and this is what the fact. Every week, we take three of the wildest true things ever recorded and tell you

the whole story behind them. The facts that sound made up, the stories that got buried, the moments that changed everything. Carter brings the history. Connor brings the facts, the tangents, and the occasional British accent. The wildest stuff is real. And the real stuff is wilder than you think. New episodes every Wednesday follow What the Fact on Apple Podcasts, Spotify, Amazon Music, or subscribe on YouTube at What The Fact Pod. You know their names. You know their faces.

But you've never heard these stories. On Celebrity Ghost Stories, an A&E original podcast presented by Reefat, You'll hear spooky stories directly from the actors, musicians, and athletes you love in their own words. What do you do when a spirit saves you from the edge of death? When a single night in a strange hotel ends with you running for the door? From angry spirits to haunted houses to unexplainable magic. These are true first-person accounts told by the stars who lived through them. Follow Celebrity Ghost Stories, available now on Apple Podcasts, Spotify, or wherever you listen to podcasts.

If you're looking for your next podcast, check out What the Fact with Connor Tomlinson from Netflix's Love on the Spectrum and Carter Roy. Every week, they take three unbelievable true facts from history and tell you the full story behind them. Follow What The Fact wherever you listen or subscribe on YouTube at What The Fact Pod.