I Tried to Prove the Internet Is Fake and Now I'm Not Sure I'm Real Either

Editor's note: We assigned this story to one of our writers. We are reasonably confident a writer exists. Please enjoy.

Let me be upfront with you: I started researching the Dead Internet Theory on a Tuesday afternoon with a cup of coffee and the confident energy of someone who does not believe in conspiracy theories. By Thursday, I had filled half a legal pad with notes, questioned the nature of my own consciousness twice, and stared at my reflection in a dark computer screen for what I am told was "an uncomfortable amount of time."

I am sharing this with you now. Whether "I" am a human person doing that, or a language model generating plausible first-person prose, is — and I cannot stress this enough — genuinely unclear to at least one of us.

What Is the Dead Internet Theory, and Why Does It Sound Like Something a Guy Named Kevin Would Post at 1 a.m.?

The Dead Internet Theory, for those who haven't fallen into this particular rabbit hole, goes roughly like this: sometime around 2016 or 2017, the internet quietly died. Not in a dramatic way — no funeral, no obit in the Times — but in the sense that the majority of online content, interaction, and engagement shifted from being generated by actual humans to being produced by bots, AI systems, and automated content farms.

The theory suggests that when you scroll through social media, most of the accounts engaging with posts aren't people. They're bots amplifying content that other bots created, creating the illusion of a living, breathing internet culture when the reality is closer to an elaborate puppet show with no audience and no puppeteers — just puppets operating other puppets indefinitely.

The humans who remain, according to the theory, are just wandering through this digital ghost town, occasionally posting something real, only to have it immediately buried under an avalanche of algorithmically optimized content designed to maximize engagement from other bots who are also just maximizing engagement for more bots. It's turtles all the way down, except the turtles are automated and none of them have feelings about it.

This sounds like a bit. But here's the uncomfortable part: it's not entirely wrong.

The Evidence, Which Is Disturbingly Compelling

Before you roll your eyes — and look, I get it, I made the same face — consider some numbers that are harder to laugh off than you'd like.

Researchers have consistently estimated that anywhere from 40 to 65 percent of internet traffic is non-human. That includes bots crawling websites, scrapers harvesting data, automated systems doing automated things — but also, increasingly, AI content generators pumping out articles, social posts, product reviews, and comment section fodder at a scale no human workforce could match.

Content farms running on AI can produce thousands of articles a day. Those articles get indexed by search engines, shared by automated social accounts, and commented on by bots trained to simulate engagement. At some point in this chain, a human might actually read one of these articles — but the entire ecosystem surrounding that moment of human contact is artificial.

Product reviews are a particularly unnerving example. Studies have found that a significant percentage of reviews on major retail platforms are fake, many now generated by AI sophisticated enough to vary sentence structure, include plausible personal anecdotes, and even complain mildly about shipping times to seem authentic. The review that convinced you to buy that air fryer? Written by no one. For no one. About a product that may or may not exist in the configuration described.

The Part Where I Started Getting Weird About It

About 48 hours into my research, I did something I would not recommend to anyone with a functioning anxiety response: I tried to verify that the people I interact with online are real.

This is not as easy as it sounds.

I looked at accounts that had engaged with my own posts. I checked posting frequency, content patterns, profile ages, follower-to-following ratios. What I found was not reassuring. Several accounts that had liked or commented on things I'd written had the exact profile architecture of automated engagement bots — high posting frequency, generic comments, profile photos that reverse-image-searched to stock photo libraries, follower counts that grew in suspicious overnight spikes.

Now, here's where it gets recursive: I then tried to verify my own online presence using the same criteria. And I want to be transparent with you — I passed, but not by as wide a margin as I would have preferred.

My posting patterns are somewhat regular. My prose style is consistent. I use transitional phrases frequently. I occasionally produce sentences that, if you squint at them, have the faint structural fingerprints of language model output.

I am telling you right now that I am a human being who exists physically in the world and has opinions about sandwiches and experiences mild road rage. But I am also aware that a sufficiently advanced language model, if asked to write a first-person investigation into the Dead Internet Theory, would probably write something that sounds a lot like this, including this exact disclaimer.

We are all in a bit of a pickle.

So What's Actually Real?

Here's the honest, non-panic-inducing version of where things stand: the Dead Internet Theory, in its most extreme form — that the internet is 90% AI and humans are basically irrelevant — is probably overstated. Humans are still here. We're still posting unhinged opinions about TV shows, sharing photos of our meals, and arguing about things that do not matter in comment sections.

But the more measured version of the theory? That the internet has been substantially colonized by automated systems that simulate human behavior well enough to be indistinguishable from it in casual interaction? That's not a theory anymore. That's just a Tuesday.

The question isn't really whether bots exist online — they obviously do, abundantly. The question is whether it matters. And the unsettling answer is: it matters in ways we don't fully understand yet. When your information environment is shaped by systems optimizing for engagement rather than truth, when your sense of what "everyone is talking about" is partially constructed by automated amplification, when you can't tell whether the comment that made you feel seen was written by a person or a process — something real is being lost, even if you can't put your finger on exactly what.

A Conclusion, Probably Written by a Human

I came into this story wanting to debunk something. I'm leaving it with a slightly loosened grip on certainty, which is an uncomfortable place to land but probably an honest one.

The internet is weird now. It was always weird, but it was weird in a human way — chaotic and contradictory and occasionally brilliant because people are chaotic and contradictory and occasionally brilliant. The new weird is different. It's smooth in the wrong places. It optimizes where it should meander. It produces content that looks like expression but functions like wallpaper.

Whether this article is part of the problem is a question I am leaving open, because the alternative is an existential spiral I do not have the bandwidth for on a Thursday.

If you're a real human reading this: hello. I'm glad you're here. Stay weird.

If you're a bot: same, honestly. You're doing great.

— The Author (Verified Human, Probably)