The Dead Internet Theory Is Becoming Reality

Have you ever heard of the "dead internet theory"? It's the claim that most of the content drifting around the web isn't made by people at all, but by bots. When it first surfaced on an online forum in 2021, it was dismissed as a conspiracy theory. Look around today, though, and you start to wonder whether it wasn't right after all.

According to a report from the security firm Imperva, 51% of all web traffic in 2024 was generated by bots. Malicious bots alone accounted for 37% of it, and that figure has climbed for six straight years. Bots are now more active than humans. In other words, more than half of the content we see is machine-made. You can no longer be sure that the article you're reading, or the comment you just "liked," was actually written by a person. The volume of AI-generated content is growing exponentially, and its quality has risen to the point where it's hard to tell apart from human writing.

On YouTube, a phenomenon called "the Inversion" is reportedly already taking shape. Fake views have grown so numerous that there's concern the algorithm may start treating the fake as the default and misclassifying the views of real humans. It's a world where the real and the fake have traded places.

Meta's Startling Patent: AI Resurrection of the Dead

In December 2025, Meta was granted a patent. What it describes is technology that uses a large language model to simulate a user's activity even when that user is away from social media—on an extended break, say, or deceased. Put simply, it means an AI running a dead person's account on their behalf: liking posts, leaving comments, even sending DMs, and—going further still—simulating video calls.

The patent document lays out this logic:

When a user has died and can no longer return to the social network,

the impact on other users is far more severe and permanent.

And so, the reasoning goes, to soften that impact, an AI should step in and play the part of the deceased.

Meta CTO Andrew Bosworth is listed as the patent's lead inventor, and the company has said it "has no plans to actually develop it." Still, it's hard to simply let this pass. Researchers at the Hebrew University and Leipzig University judge that the patent could mark a turning point for "AI resurrection." Earlier cases of AI resurrection were small, one-off projects run by families or startups, but Meta's patent envisions building posthumous simulation directly into the platform's own infrastructure.

The reason such technology emerges at all is simple. It's about engagement. Fewer users means less ad revenue.

Facebook has long been turning into a graveyard of forgotten accounts, endless ads, and birthday wishes that never get a reply. At the same time, its feed is filling up with harmful AI-generated content. AI is moving into the spaces people have abandoned. And the activity that AI generates becomes data in turn, training the next generation of AI. An infinite loop.

When Quantity Overwhelms Quality

But hold on—it's worth stepping back for a moment. What is the essence of all of this? Here's how I see it:

The "quantity" of content has begun to overwhelm its "quality."

The volume of content AI produces in a single day easily surpasses what a human could create in a lifetime. Blog posts, social media updates, comments, reviews, even news articles. The problem is that they all resemble one another, because they're produced by models of the same structure drawing on the same training data.

This is where the concept of "model collapse" comes in. It's the phenomenon in which feeding AI-generated content back into AI for training causes the quality of the output to steadily degrade. Repeat the copy of a copy long enough and the original's character eventually disappears—much the way a document loses its sharpness when you photocopy a photocopy, then copy that, and copy it again.

Three Kinds of Content That Survive the Dead Internet

So what kind of content survives in this dead-internet era? I see three.

First: Content Born of Experience

AI assembles training data into plausible-sounding text. But it can't manufacture a story that comes from actually living through something.

"I made this mistake while running a publishing house, and here's how I fixed it" is fundamentally different from an AI-generated piece titled "10 Things to Watch Out for When Running a Publishing House." The former has the context of failure, it has emotion, it has the specifics of a real situation.

AI learns patterns, but it doesn't have experiences. That difference is bigger than you'd think.

Second: Content With a Point of View

People interpret the same facts differently, don't they? That difference in interpretation is what creates a piece's value.

AI excels at synthesizing existing viewpoints and delivering the most average answer. But an original lens—"this is how I see this phenomenon"—one that's sometimes even provocative, can come only from a human being.

Content that doesn't converge on the average. That's the content that survives in the age of AI.

Third: Content Built on Accumulated Trust

In the dead-internet era, the scarcest resource isn't information—it's trust. "Does the person who wrote this actually know the field?" "Has this information been verified?" Only content that can answer questions like these will survive.

That's why I think the author's name matters more than ever. A brand with a real person behind it grows more important. We're moving toward an era in which a single piece written by one trustworthy person is worth more than thousands of anonymous pieces made by AI.

What Publishing Means in the Age of AI

A book carries the author's name in print, the verification of having passed through a publisher's editorial process, and the substance of existing in physical form. It has a fundamentally different trust structure from the anonymous, source-unknown AI content drifting across the internet.

There's a caveat here, though: the form of a book doesn't automatically guarantee trust. As books cranked out by AI in a single month begin to flood the market, trust in the book as a medium is itself growing precarious. That's exactly why editorial intent, an author's authenticity, and creative originality matter all the more.

In a world where Meta has patented a way to revive a dead person's account with AI, what we humans can do is clear. We do the things only a living person can do. We actually experience things, organize the thinking that comes out of that experience, and express it in our own words.

What Only Humans Can Do: Making Meaning Within Context

An era in which bots have crossed half of all internet traffic. An era in which a patent for letting AI run the social media accounts of the dead has been registered, so that you can no longer tell whether a living person wrote what you're reading. If you make content in this era, you'll likely have no choice but to come back to one question.

"Am I writing as a living person?"